Looking ahead just a short time, Apple could hold a powerful position in AI where it most makes sense: deployment.

Not only will the company offer up its own AI models for the kind of tasks millions use ChatGPT to do today, but it will provide more sophisticated on-device agentic models to help users get things done through Siri AI.

Apple also offers limited capacity for more complex tasks through Private Cloud Compute, and, in partnership with the likes of Google in the US and Alibaba in China, the company is giving users a trusted conduit through which to access even more sophisticated AI services. 

Deeply deployable

Critics can say it took Apple a long time to get to this point, but they also seem to think the company has finally got the mix right with its series 27 operating systems. Arriving late to a party doesn’t mean you won’t shine once you get there.

Apple is also coming up the inside lane around frontier AI, with iterative OS and hardware enhancements that mean its devices become increasingly effective for Edge AI use cases, on device — no cloud service required.

The company appears to be digging down into those use cases. Mark Gurman at Bloomberg recently predicted that future M7 Ultra Macs will support as much as 1.5TB RAM, making these systems more than capable of running full weight frontier models in people’s offices, colleges, and homes. 

While that does assume the AI-flationary memory market can supply that much RAM at prices humans can afford, it is also true that people are already running AI clusters using off-the-shelf Mac minis networked over Thunderbolt cables. It’s no stretch to believe this will continue to be the case, and that it will even broaden as the power/performance offered at the high end grows.

What’s wrong with good enough?

When combined with open AI stacks, particularly newly emerging varieties, Apple’s platforms should become leading contenders for private AI services and edge AI. Many business users will leap at the chance to offer their workers powerful, self-hosted, private AI services using one or more daisy-chained Mac Studios or Mac minis. The recent craze in deployment of both Macs to support OpenClaw instances shows they already are.

Ultimately, these different slices of momentum mean I agree with investor Jason Calacanis that Apple is in position to apply a great deal of pressure on OpenAI and Claude just by putting models on their devices. 

It’s also worth thinking about how people use AI today. How many of the queries made in the world right now constitute relatively simple tasks that could be transacted by on-device AI, such as the emerging new version of Apple Intelligence or even smaller LLM models running on device? You can even run PrismML’s 1-bit, 27-billion parameter Bonsai on an iPad using the Locally app, and that’s in the here and now.

What happens? Pretty soon you’ll find people recognize that they can already run the vast majority of their AI-augmented workflows using services they have on their existing device or can access on their on-prem Mac set-ups. And, of course, as people get used to running small tasks locally and larger tasks on premises, the actual space in which they need to turn to cloud-based frontier models will erode. That’s even as companies like PrismML work towards slimming down full-weight models so they don’t need to run on a server at all.

“It’s going to be wild when people have unlimited tokens on their desks,” said Calacanis in a podcast round table discussion.

Who has the most to lose?

The current incarnations of AI felt like they came from nowhere. Most people weren’t aware of the technology until returning to work after the 2022 holiday season. Since then, the industry has proliferated with dozens of competing models, most recently including powerful but affordable frontier models such as Qwen and Kimi.ai.

These models aren’t necessarily all as good as one another, but in many cases for much of what we do, we’ll find them to be good enough. That’s an existential crisis for some, as industry observers now think the inevitable pricing pressure means some services might have over-invested in capacity before finding any way to turn a profit.

Those profit-seeking services are the ones with the most to lose as Apple extends its hardware advantage, democratizing AI access for all while providing platforms suitable for edge AI, on-premises AI, private AI, and even AI access using third-party services. (The need for the latter will shrink as the capabilities of the former get better.)

Cupertino rising

What does this all mean? While the industry remains young, it is already fragmenting. And striding through the dust of that process comes Apple, equipped with the hardware, software, and approach to build its business even as the enterprise of first mover AI services erodes.

You can follow me on social media! Join me on BlueSky,  LinkedIn, Mastodon and subscribe to my daily Apple-related news summaries at The Core.

Healthy software companies typically don’t suddenly eliminate one-fifth of their workforce, but monday.com is doing just that as it bets on flatter teams, AI agents, and customer implementation expertise as the winning combination in the AI era.

Monday.com co-founder and co-CEO Eran Zinman today announced the “very difficult decision” to reduce the AI work platform company’s global workforce by about 20%, or 620 people.

The move has nothing to do with increasing margins or replacing humans with AI, he insisted in his post on LinkedIn; rather, it’s a calculated decision to trim down and hone the company’s focus as AI becomes integral to day-to-day workflows.

“This is not a distress signal; it is a deliberate reset, disclosed with its price attached,” said Sanchit Vir Gogia, chief analyst at Greyhound Research. “The industry has quietly swapped the meaning of productivity, and this filing is the clearest exhibit yet.”

A ‘significant opportunity’ in technology

In a SEC filing this week, monday.com said its restructuring plan reflects the “ongoing transformation of its product, marketing, and go-to-market strategy.” The move is intended to support a “leaner, more focused operating model” as the company continues to invest in its AI-driven strategy.

Zinman noted in his post that the company has shifted to “doing the work with AI and not just managing it,” and is focused on building environments where “people and AI agents [work] together in one workspace.”

In recent months, monday.com has evolved its products, strategy, and the way it serves its customers, and Zinman contended that “the organization we built for our previous chapter is not the organization that fits the new AI era.” Monday.com needs to “execute more decisively,” take on new challenges, and quickly respond to market changes, he said.

“We have never seen such a significant opportunity in software, driven by such exciting technology,” Zinman noted. He emphasized that the reduction is not to replace people with AI, nor to improve margins; the “vast majority” of savings will be reinvested into talent, products, and AI.

The restructuring will result in a “flatter organization” with fewer management layers and smaller, more autonomous teams, and monday.com also has a new go-to-market model, Zinman explained. Customers expect “deeper implementation support” as they deploy AI, and the company will work more closely with customers, increase its on-site presence, create new roles, and “adapt many existing ones.” In its SEC filing, the company said it expects to continue hiring in “key strategic areas” throughout 2026.

Workers will be expected to work better, “not harder,” Zinman noted. He pointed to several past examples where work could have been done in a few days, but instead took many months with “multiple meetings and endless friction.”

“This wasn’t people’s fault and everyone was frustrated by this,” he said. “Our new org changes ownership to allow people to make decisions and move fast.”

A spokesperson for monday.com declined to comment further on the staff reductions.

Monday.com’s key market advantages

Monday.com certainly isn’t struggling; the company expects 19% to 20% year-over-year growth in 2026.

“Companies in that position do not restructure because they must,” Greyhound’s Gogia noted. “They restructure because they have decided to become something else.”

Melody Brue, VP and principal analyst at Moor Insights & Strategy, pointed out that organizational redesign is important for real AI transformation, but while it can signal confidence to the market, it can still be “devastating” to humans.

While the company looks as though it’s trying to do right, that ultimately remains to be seen, she said. “There are often hidden internal bruises that can surface long after layoffs.”

Monday.com’s advantage is in its “structured substrate,” Gogia noted; its boards, permissions and typed workflows give agents something firmer to act on than just documents and chat history. The company highlights its natively built agents that can be configured by any team member, as well as connectors with Claude, Microsoft Copilot, and ChatGPT, and dedicated routes for external agents to authenticate and operate.

“For some time, the sharper enterprise question has been shifting from who has an agent to who owns the governed runtime in which an agent can safely act,” he said. “Structured work is a serious claim on that runtime.”

But parts of monday.com’s agent estate remain in staged release, and its product is ultimately “mid-transition,” Gogia pointed out; its agent builder carried a beta label as recently as March,. Also, the company’s pricing model changed in May to a hybrid model charging for seats as well as mandatory AI credits. And, while its AI-powered no-code builder monday vibe passed $1 million in annual recurring revenue within two and a half months, monday.com has not released subsequent outcomes, usage volumes, or attach rates.

Further, there’s an element of “gravity” with its competitors, he observed. Asana is reorganizing teams around agents, Atlassian is wiring agents into the developer estate, and others are simply bundling them into their offerings: Microsoft is doing so across the productivity stack, and ServiceNow across enterprise operations, each with identity and procurement built in.

“Their pull is strongest exactly where monday.com wants to grow, in the largest accounts, where control-plane depth and administrative reach decide the deal,” said Gogia.

Actions for the near-term

Going forward, buyers should focus on operating risk, not headline risk, Moor’s Brue noted. In practice, that’s continuity of service, roadmap consistency, and strength of enterprise support. Productivity should be valued as better outcomes per unit of organizational effort, not mere activity.

“It should be a measure of how much smoother, faster, and more effective the operating model becomes when AI is built into the work,” said Brue.

Gogia noted that strain surfaces first in customer service, and monday.com’s attention is being redistributed. The company’s annual report disclosed that its focus is now concentrated on the largest accounts, with support for medium-sized clients moved to an AI-first and human-supported model.

During the first month of the transition, buyers should track named account continuity and escalation times, he advised. By the first quarter, keep an eye on whether credit governance and admin controls mature on schedule, and if the roadmap beyond the AI estate keeps pace. By the half-year mark, determine whether promised implementation depth is producing outcomes or “simply more billable engagement.”

Support tiers should be enumerated in writing before renewal, and buyers should contract for “side exits,” Gogia emphasized, with overage pricing fixed in advance, the right to pause consumption, and portability for workflows and agent configuration “if the relationship sours.” Finance should also insist on monthly consumption reporting by capability. Further, integration efforts, partner dependency, and change management should be considered first-class costs of the agent era, “not as afterthoughts to a license.”

“A license was a known cost,” said Gogia. “A meter is a behavior, and behavior is harder to forecast than headcount.”

This article originally appeared on CIO.com.

OpenAI has launched Presence, an enterprise service for deploying voice and chat agents that can resolve customer and employee requests, potentially automating some work now handled by frontline support teams.

The agents can answer questions and operate IT systems, and enterprises can decide what actions the agents may take and when they should seek human approval for actions or transfer a case to a human.

OpenAI is already using Presence internally for its English-language phone support channel, where it verifies callers and uses account information to complete approved actions. The company said the system resolves 75% of inbound issues without human assistance.

Another OpenAI service, Codex, can be used to monitor agents and suggest updates or improvements to processes. In OpenAI’s own tests, suggestions from Codex helped reduce handoffs to humans by 15 percentage points over 10 days, it said. Presence also includes simulation and evaluation tools that allow companies to test an agent before deployment. The tests assess whether it reaches the correct outcome, follows company policy, and hands a case to an employee when required.

OpenAI intends each Presence deployment to deal with one kind of task, for example billing issues, insurance claims, or employee IT service requests, with agents getting only the knowledge and system access required for that task.

Presence is not a self-service product: Enterprises will have to sign up for the limited availability program, with integration performed by OpenAI or selected global systems integrators.

Companies exploring or testing Presence include Spanish bank BBVA, which is evaluating the service for everyday banking support in Mexico, and Japanese technology group SoftBank, which is using it in trials involving Japanese-language customer interactions. Australian insurer IAG is assessing whether the technology can help it respond to surges in customer demand during severe weather events.

Workforce impact

OpenAI’s announcement did not address the potential effect of Presence on employment. But its claimed automation rate raises questions about how the technology could affect staffing in customer service and other support functions.

Pareekh Jain, CEO of Pareekh Consulting, said CIOs should regard the 75% figure as evidence that the technology can work, rather than as a benchmark that every enterprise can expect to reach.

Jain said OpenAI’s deployment benefits from being built around the company’s own products and data. Large enterprises may achieve lower automation rates because they must contend with fragmented legacy systems, uneven knowledge bases and more complex compliance demands.

“Most organizations should expect lower initial automation levels that improve over time as the AI agent is refined,” Jain said.

The first workforce effect is more likely to be slower hiring than immediate layoffs, according to Tulika Sheel, senior vice president at Kadence International.

“The roles most exposed are likely to be repetitive, high-volume functions such as frontline customer support and routine back-office processing,” Sheel said. “However, I would expect the first impact to be on hiring and team growth rather than immediate large-scale job cuts. Over time, enterprises may redesign roles around AI-assisted workflows, with humans focusing more on complex cases, escalation, and relationship management.”

Jain said Tier-1 support agents handling predictable queries would face the most exposure. Broader reductions would become more likely only after companies reorganize their operations around the technology.

However, Lian Jye Su, chief analyst at Omdia, said Presence is unlikely to increase the threat of job displacement because companies have used similar customer-support automation from vendors such as Genesys, NiCE, Five9 and AWS for years.

Enterprises are more likely to use Presence alongside employees, with AI handling routine requests while people remain responsible for work requiring judgment and empathy, Su said.

Cost and operational risks

Analysts said CIOs should examine whether Presence can maintain resolution quality as usage grows, since fewer human handoffs could leave employees dealing with a more difficult mix of cases.

“The key question is not simply how many tasks AI can handle, but whether it can handle them reliably at scale,” Sheel said.

The financial case will depend partly on the cost of connecting Presence to existing systems and maintaining the controls needed to govern its use, according to Jain. “Often the biggest cost of enterprise AI is not tokens but integration and governance,” Jain added.

Companies will need to determine what systems and data the agents can access, monitor their performance, and audit the actions they take. Those investments could offset early savings.

Su said the complexity of enterprise IT will make it difficult for OpenAI to automate entire workflows on its own. Enterprises will still need to work with other technology providers and human employees, while CIOs will favor systems that can be audited and integrated with existing infrastructure.

Jain said the economics could improve if companies use the same integrations and governance controls across additional workflows.

This article first appeared on CIO.

AMD executives took to the stage at its Advancing AI 2026 event in San Francisco today to detail the company’s next generation of AI infrastructure solutions, from Instinct MI455X AI accelerator GPUs and 6th Gen EPYC “Venice” CPUs, to Pensando networking, ROCm.AI software and its Helios rack-scale platform that ties it all together.

AMD has been working towards rack-scale AI system solutions for years. Its ZT Systems acquisition last year added valuable engineering talent and intellectual property that is now finally bearing the real fruits. Its Helios AI platform is a major platform evolution for AMD, with shipments scheduled to begin in the second half of this year (which is here and now).

The announcements at Advancing AI show how the company has engineered its AI platform solutions for large reasoning models, sustained inference and agentic workflows. These workloads pressure memory capacity, data movement, networking and CPU orchestration. AMD’s approach is to keep as much data close to the compute engines as possible and move it more efficiently throughout the system, but there’s deeper nuance here that’s obvious versus AMD’s chief rival, NVIDIA.  

AMD’s MI455X targets the AI memory wall

The Instinct MI455X GPU is the compute engine that fuels the Helios rack, and the first GPU based on AMD’s new CDNA 5 architecture. Built with a modular mix of 2nm and 3nm chiplets, it carries 432GB of HBM4 and 23.3TB/s of peak memory bandwidth.

Compared to AMD’s current MI355X, the MI455X offers 1.5 times the memory capacity, up to 2.9 times the peak memory bandwidth and up to four times the peak matrix performance with MXFP4 and MXFP8 data types, which are lower-precision numerical formats designed to accelerate AI processing while reducing memory demands. With MXFP6 (6-bit floating point), performance is rated at up to twice that of MI355X.

AMD also shared some actual, measured internal results using production silicon. The company claims MI455X delivers 3.8 times higher FP8 decode performance, 3.5 times more measured FP4 compute performance and between 2.5 and 3.5 times more networking bandwidth than MI355X, depending on the transfer path tested. Those figures provide more context than just numerical specifications, though they remain AMD-provided comparisons that will need independent validation.

AMD

The architectural choices behind the numbers are important. Reasoning models and long context windows require sizeable KV caches for maintaining AI attention states, while mixture-of-experts models frequently move large amounts of data across accelerators. MI455X should let more model data, activation states and cache remain local. New dedicated IP in hardware can transfer data while the GPU continues processing, and expanded cache and multicast capabilities are designed to reduce redundant data movement to further improve efficiency.

The aforementioned lower-precision formats can also raise throughput and reduce memory use, but model developers still have to determine where they can be applied without unacceptable accuracy loss.

AMD’s Helios rack takes aim at Vera Rubin

Dave Altavilla

Helios is AMD’s primary rack-scale competitor to NVIDIA’s Vera Rubin platform. Each liquid-cooled rack combines 72 MI455X GPUs, 18 single-socket Venice host CPUs and Pensando networking technologies.

In its most complete, premium configuration, AMD rates Helios for 2.9 exaflops of low-precision AI compute, with 31TB of aggregate HBM4 capacity, 1.7PB/s of memory bandwidth, 260TB/s of bidirectional scale-up bandwidth and 43TB/s of scale-out bandwidth.

These are formidable figures, but they are technical specifications rather than actual application benchmarks. The more consequential development is AMD’s move from collections of eight-GPU servers to a 72-GPU shared-memory domain. Models too large for one node can operate across the rack without treating every exchange as a scale-out networking transaction, which benefits large-model inference as well as training.

AMD uses UALink over Ethernet, or UALoE, for an open standard scale-up fabric. Each MI455X provides 3.6TB/s of bidirectional scale-up bandwidth, while the complete rack delivers all-to-all connectivity through a single switch layer. AMD also claims six times more scale-out bandwidth per GPU than MI355X when MI455X is configured with three Pensando Vulcano 800 AI NICs.

While open standards give cloud providers more control over suppliers and system design, AMD and its partners now have to prove those components can deliver the predictable performance, reliability and deployment experience customers expect from a tightly controlled, more vertically integrated platform.

Finally, AMD designed Helios with automatic rerouting around failed links, virtual rack partitions, tray-level serviceability and rack-wide power, cooling and health monitoring. Major hyperscalers and potentially large-scale enterprise customers will likely key in on these capabilities, which can affect the availability, total cost and consistency of the AI services they consume.

Kind of like cowbell, AMD Venice gives agentic AI more CPU

AMD

AMD’s agentic CPU messaging regarding its upcoming Venice-based EPYC processors is mostly marketing speak, but the underlying requirement is very real. An AI agent can invoke retrieval, databases, security checks, code execution and other tools before a GPU generates a response. Running many agents concurrently increases the amount of conventional compute requirements surrounding the accelerators.

Venice scales to 256 Zen 6 cores with support for 512 threads, 16 memory channels, up to 1GB of L3 cache per socket, along with PCIe 6.0 and CXL 3.1 connectivity. AMD is also offering several Venice configurations for other applications, including general-purpose servers, high-frequency workloads, GPU hosts and high-density CPU sandbox systems used to execute agent tools.

Treating the CPU solely as a GPU host understates its role. Gateways, tokenization, vector search, databases and short-lived code execution stress different mixes of per-core performance, thread count, memory bandwidth and I/O. Specifically, AMD’s internal testing shows Venice significantly outperforming its current EPYC 9965 Turin CPU across five parts of the agentic AI pipeline, including gateway processing, context assembly, vector search, enterprise applications and short-lived tool execution. Individual gains vary by workload, but AMD details the overall generational improvement at up to a 1.7 times lift. As with the MI455X figures though, these comparisons come from AMD and will require independent validation.

Pensando networking and ROCm software advance

Keeping GPUs fed with data and coordinating traffic across racks directly affects utilization and operating costs. In fact, GPU utilization is a pretty sad state of affairs currently for some of the major frontier model providers.

As such, Pensando networking has become central to AMD’s roadmap. Helios can connect each MI455X to as many as three 800Gbps Vulcano AI NICs, while Salina DPUs handle front-end networking and infrastructure services.

On the software side, which is an equally critical component, AMD also introduced ROCm.AI, an AI-assisted development layer due to arrive in August. It includes reusable skills for coding agents, simplified management and Hyperloom, which can profile workloads, tune serving configurations, modify kernels and validate results.

These tools address two persistent AMD challenges: developer efficiency and ease of use, and software tuning. Automated optimization still has to produce repeatable gains without creating hard-to-maintain code, however. And while ROCm has progressed significantly over the last few years, NVIDIA’s CUDA retains an advantage in maturity, tooling and developer familiarity.

Customer commitments underscore rack-scale confidence

AMD now has commitments that give its MI450 generation and Helios considerably more weight. Meta and OpenAI have announced multi-generation agreements composed of up to 6GW of AMD compute capacity, with initial 1GW deployments planned for the second half of 2026.

Oracle plans a 50,000-GPU public cloud cluster beginning in the third quarter, while Microsoft will deploy Helios for Azure AI inference. Finally, just before the AMD event, Anthropic announced a strategic partnership for up to 2 Gigawatts of AMD-fueled AI compute, with its first gigawatt expected online in the first half of 2027.

Commitments of this scale reflect confidence in more than just MI455X performance. These customers are evaluating the complete architecture, including Venice CPUs, Pensando networking, ROCm software, rack integration, serviceability and AMD’s ability to deliver and execute across multiple product generations.

There is some financial alignment behind the agreements as well. AMD issued OpenAI performance-based warrants and committed to investing up to $5 billion in Anthropic. That context matters when evaluating these deals as market validation, but these planned deployments are substantial nonetheless and put Helios on a much stronger foundation as it begins shipping.

AMD expands its robotics and embedded foundation

AMD also expanded its physical AI portfolio, building on credible traction from its Xilinx-derived Kria adaptive system-on-modules and embedded technologies that are already powering robotics, machine vision and industrial automation applications.

The new Ryzen AI Embedded X100 combines up to 16 Zen 5 CPU cores, integrated Radeon graphics, a second-generation NPU and as much as 128GB of unified LPDDR5X memory shared across its compute engines. To me this looks a lot like a repackaging and optimization of the company’s Strix Halo platform, but with specific optimizations for the embedded space. Regardless, AMD is pairing X100 with the Kria AI Robotics Developer Platform, which includes a System Module or SOM, and a new Robotics Partner Network spanning hardware, software and platform providers.

Samples began shipping in June, with full production expected in the fourth quarter. This broader objective is to give developers a path across AMD x86 CPUs, GPUs, NPUs and FPGAs for real-time autonomous systems, rather than requiring them to assemble those hardware engines and software components independently.

Execution for AMD is now the test

AMD has assembled a credible platform for the burgeoning agentic AI market that’s blowing up currently with no signs of stopping. MI455X addresses memory and data movement, Venice handles dense agentic CPU workloads, Pensando networking connects global system resources, and ROCm.AI addresses software complexity. Finally, Helios assembles these components into a true competitive threat for NVIDIA’s latest Vera Rubin platform.

AMD’s open architecture may appeal to customers seeking supplier choice, but openness must also translate into reliable deployments, competitive total cost and software that does not require a significant rip-up. NVIDIA enters this cycle with a stronger ecosystem and far more rack-scale deployment experience. The true test will be how easily and reliably customers can integrate, operate and maintain these AMD solutions at scale.

As it stands, AMD now has major customers and a clearly defined architecture with systems engineering expertise behind it. Delivering Helios on schedule and showing that its performance claims translate into a real production workload throughput advantage and total cost of ownership gains will determine how much the competitive gap narrows. And of course, this is in a market that is clamoring for ever-more compute resources with a seemingly insatiable demand for AI services and capacity. That’s an environment for big iron success. Now AMD just has to deliver optimized, turnkey AI platforms. This is far easier said than done, but time will soon tell as deployments take shape this year.

This article is published as part of the Foundry Expert Contributor Network.
Want to join?

Foto: I Believe I Can Fly – shutterstock.com

Enterprise Architecture (EA) Tools unterstützen Unternehmen und Organisationen dabei, mit ihren IT-Strategien die Geschäftsziele optimal zu unterstützen. Sie sorgen ebenfalls dafür, dass Unternehmen ihre Roadmaps für die digitale Transformation geordnet vorantreiben können. EA Tools bieten dafür unter anderem Collaboration-, Reporting-, Testing- und Simulationsfunktionen. Mit deren Hilfe lassen sich Modelle implementieren, die Geschäfts- und IT-Prozesse gezielt verbessern.

Um die beste Lösung für Ihr Unternehmen zu finden, sollten Sie zuerst prüfen, ob sich das jeweilige Tool mit Ihrem Technologie-Stack integrieren lässt. Anschließend gilt es abzuwägen, ob die Informationen, Diagramme und Tabellen, die die Software zur Verfügung stellt, für das Unternehmen auch einen echten Nutzwert haben.

Empfehlenswerte Enterprise-Architecture-Tools

Nachfolgend finden Sie einen Überblick über die wichtigsten Enterprise-Architecture-Tools – in alphabetischer Reihenfolge. Sie stellen einen Mix aus Visualisierungs-, Collaboration- und Project-Management-Funktionen bereit und unterstützen eine Vielzahl von Enterprise Architecture Frameworks.

Ardoq

Nachdem zuerst über einfache Formulare Informationen von Usern, Entwicklern und sonstigen Stakeholdern im Unternehmen eingesammelt wurden, lässt sich mithilfe von Ardoq ein digitaler Zwilling der gesamten Organisation erstellen. Der Ansatz setzt also darauf, die Menschen, die in ihren Rollen mit den verschiedensten Systemen arbeiten, realistisch in ihrer Arbeitswelt abzubilden.

Jede Mitarbeiterin und jeder Mitarbeiter im Unternehmen kann später von den Netzwerkvisualisierungen und Datenfluss-Diagrammen profitieren, um seine eigene Rolle optimal zu unterstützen und den Arbeitsplatz immer wieder anzupassen und zu modernisieren. Das Tool lässt sich mit den wichtigsten Cloud-Plattformen integrieren. Es bietet eine API, die individuelle Anpassungen in allen wichtigen Programmiersprachen (Python, C#, Java, etc.) ermöglicht.

Wichtigste Use Cases:

• “Architektonischen Stress” bei Lastspitzen simulieren, falls größere Veränderungen bevorstehen;

• Verstehen, wie verändertes Nutzerverhalten neue Anforderungen generiert;

• Application Portfolio Management, um besser strategisch zu planen.

Atoll Group SAMU

Das EA-Tool SAMU macht die Enterprise Architecture sichtbar, indem es tiefe Verknüpfungen zwischen On-Premises-Systemen, dem Cloud-Layer und Tools für das Business Process Management aufzeigt. Das Tool der Atoll Group bietet vielfältige Integrationsmöglichkeiten, zum Beispiel mit Monitoring-Tools (etwa Tivoli, ServiceNow), Configuration-Management-Datenbanken (zum Beispiel CA, BMC) oder Service-Organisations-Tools (BMC, HPE). Alle Informationen fließen in ein zentrales Datenmodell ein, das um den zusätzlichen Input der Stakeholder weiter angereichert wird.

Wichtigste Use Cases:

• Enterprise-Architektur visualisieren;

• strategische Planungsprozesse und Architektur-Reviews mit Informationen unterfüttern;

• mithilfe einer visuellen Verständnisgrundlage die Kommunikation verbessern.

Avolution Abacus

Dieses Tool erfasst die Breite und den Umfang der Unternehmensarchitektur mit Hilfe eines auf Diagrammen basierenden Dashboards. Die Integration mit gängigen Tools wie SharePoint, Excel, Visio, Google Sheets, Technopedia oder ServiceNow vereinfacht die Nutzung. Abacus wurde inzwischen auch um einen Machine-Learning-Layer ergänzt, der es Anwendern ermöglicht, ein Modell zu trainieren, das ihnen beispielsweise hilft zu erkennen, wer im Unternehmen für welches System verantwortlich ist.

Wichtigste Use Cases:

• die IT für das gesamte Unternehmen “öffnen”, um ein allgemeines Verständnis der Datenflüsse zu erzeugen;

• umfassendes Enterprise Modeling, um eine Roadmap für künftige Entwicklungen zu erstellen;

• Business-Metriken tracken, die mit der Unternehmens-Performance zusammenhängen.

BOC Group ADOIT

ADOIT soll Teams dabei unterstützen, Ressourcen zu verwalten, Bedarfe vorherzusagen und Assets zu tracken. Dazu mappt das Tool jedes System oder Softwarepaket mit einem Objekt. Die Datenflüsse zwischen den Systemen werden in Beziehungen umgewandelt, die von diesen Objekten mithilfe eines anpassbaren Metamodells erfasst werden. Geschäftsprozesse können auf ähnliche Weise über ein gut integriertes Begleitprodukt namens ADONIS modelliert werden. ADOIT ist Web-basiert und lässt sich auch mit Tools wie Atlassian Confluence integrieren, um die Datenerfassung und -entwicklung zu beschleunigen.

Wichtigste Use Cases:

• ein unternehmensweites Modell erstellen, das bei sämtlichen Teammitgliedern ein Verständnis über den Stack schafft – und wie man diesen verbessern kann;

• vollständiger Zugriff auf EA-Daten über eine Mobile-Anwendung;

• bei Fusionen und Übernahmen den Tech-Bereich durch genaues Asset-Mapping orchestrieren.

Bizzdesign Hopex

Nach der Übernahme von Mega International zählt die Hopex-Plattform zum Portfolio von Bizzdesign. Sie soll dabei unterstützen, Unternehmensanwendungen zu modellieren und dabei ein Verständnis der von ihnen unterstützten Geschäfts-Workflows schaffen. Dabei liegt ein Schwerpunkt auf den Bereichen Data Governance und Risikomanagement. Hopex basiert auf Microsoft Azure und stützt sich auf eine Reihe offener Standards wie GraphQL und REST Queries, um Informationen aus Komponentensystemen zu sammeln. Das Reporting ist mit den Office-Tools von Microsoft sowie mit grafischen Lösungen wie Tableau und Qlik integriert.

Wichtigste Use Cases:

• datengestützte Erkenntnisse herbeiführen, um Cloud- und Anwendungsbereitstellung zu steuern;

• akkurate Nutzungsmodelle erstellen, um Architekturanforderungen zu verstehen;

• eine Bedarfsschätzung mit Umfragen und anderen Tools vornehmen, um für die Zukunft zu planen.

Bizzdesign Horizzon

Das Tool dient dazu, Business Workflows und den zugrundeliegenden Tech-Stack zu modellieren. Dazu bietet Horizzon ein Graph-basiertes Modell, das Daten von sämtlichen Stakeholdern einsammelt und diese an eine Analytics-Engine weitergibt. Im Ergebnis entstehen Diagramme, die den aktuellen Systemzustand widerspiegeln. Wichtige Schwerpunkte dieses Tools sind Change Management und Zukunftsplanung: Horizzon ist nicht zuletzt dafür konzipiert worden, die Risiken eines Redesigns zu minimieren. Das Toolset unterstützt die wichtigsten Frameworks ArchiMate, TOGAF und BPMN. Neben Mega hat Bizzdesign im Januar 2025 auch den EA-Geschäftsbereich der Software AG – Alfabet – übernommen.

Wichtigste Use Cases:

• Vorhersage zukünftiger Anforderungen durch Predictive Modeling;

• Orchestrieren von Workflows auf der Basis der technischen und der Business-Architektur;

• Antizipieren von Risiken sowie Security- und Governance-Problemen durch die Modellierung von Datensicherheitsanforderungen.

Capstera

Das Tool von Capstera fokussiert darauf, die Business Architecture selbst abzubilden. Value und Process Maps helfen dabei, die Rollen der verschiedenen Unternehmensbereiche zu definieren und nachzuverfolgen. Dabei können im laufenden Prozess Verknüpfungen mit den zugrundeliegenden Softwarprodukten und Tools hinzugefügt werden.

Wichtigste Use Cases:

• Reports erstellen, die sich erst einmal mit der Business-Architektur selbst beschäftigen;

• Beziehungen zwischen Menschen, Abteilungen und Rollen analysieren;

• die langfristige strategische Planung vorantreiben.

Clausmark Bee360

Teammitglieder, die Clausmarks Flaggschiffprodukt Bee360 (früher Bee4IT) verwenden, wollen eine einfache “Single Source of Truth” über die Workflows im Unternehmen. Ziel ist es, verschiedenen betrieblichen Rollen intelligentere Entscheidungen zu ermöglichen. Das Modul Bee360 FM (Finanzmanagement) bietet etwa die Möglichkeit, Kosten nachzuvollziehen und zuzuordnen. Die Anwender können verschiedene solcher Module miteinander verknüpfen, um EAM, Finanzmanagement, Portfolio Management und Agile Planning nahtlos zu integrieren – bei maximaler Transparenz.

Wichtigste Use Cases:

• C-Suite-Ebene befähigen, Projekte zu managen und Assets zuzuweisen;

• präzise digitale Zwillinge entwickeln, um ein Verständnis über Datenflüsse zu schaffen und künftige Erweiterungen zu planen;

• integrierte Wissensdatenbank aufbauen, um alle digitalen Workflows zu tracken.

EAS

Das Essential-Paket von EAS (Enterprise Architecture Solutions) nahm als Open-Source-Projekt seinen Anfang und hat sich inzwischen zu einer kommerziell verfügbaren Cloud-Lösung weiterentwickelt. Das Tool erstellt ein Metamodell, das die Interaktionen zwischen Systemen und Geschäftsprozessen beschreibt. Ebenfalls enthalten sind Pakete, um gängige Business Workflows wie Datenmanagement oder DSGVO-Compliance zu tracken.

Wichtigste Use Cases:

• den technischen Reifegrad der eigenen Architektur evaluieren;

• Sicherheit und Governance durch besseres Asset Tracking optimieren;

• wachsende Systemkomplexität kontrollieren und managen.

OrbusInfinity

Orbus Software hat Anfang 2025 die Akquisition seines Konkurrenten Capsifi abgeschlossen. Der Anbieter stellt mit OrbusInfinity eine Enterprise-Transformation-Plattform auf KI-Basis zur Verfügung, die schnellere, bessere Entscheidungen, Kosteinesparungen und Risikominimierung verspricht. Architecture-Teams sollen mit Hifle von OrbusInfinity mit einer Vielzahl von Stakeholdern interagieren können, um eine “digitale Blaupause” ihres Unternehmens zu generieren, die eine einheitliche Sicht auf das aktuelle und künftige Geschäft realisieren soll. Diverse Drittanbieter-Tools lassen sich außerdem mit der Plattform integrieren, darunter etwa von Microsoft, Flexera, ManageEngine oder ServiceNow.

Wichtigste Use Cases:

• Stakeholder-Management;

• Enterprise-Landschaften visualisieren;

• Entscheidungsfindung und Datenanalyse automatisieren.

Planview Enterprise One

Planview bietet eine ganze Reihe von Produkten, mit denen Unternehmen Teamwork, Prozesse und die Enterprise Architecture nachvollziehen können. Die Enterprise Tools sind in drei Kategorien unterteilt: strategisches Portfolio-Management, Produktportfolio-Management und Projektportfolio-Management. Im Zusammenspiel entstehen hardware- und Software-übergreifende Layer, die rollenbasierte Perspektiven für Führungskräfte und Teammitglieder eröffnen. Das Toolset integriert mit gängigen Ticket-Tracking-Systemen wie Jira, um Workflow-Analysen und Reports zu erstellen. Inzwischen hat Planview nach einer Übernahme neue Tools in sein Portfolio integriert, die früher unter den Namen Daptiv, Barometer und Projectplace bekannt waren.

Wichtigste Use Cases:

• eine langfristige, strategische Vision für die Architekturentwicklung aufbauen;

• Entwicklungsarbeit auf Projektebene tracken und in eine beliebige Strategie integrieren;

• mit Fokus auf die Customer Experience und die Produktstruktur den Change vorantreiben.

QualiWare Enterprise Architecture

Das Enterprise Architecture Tool von QualiWare ist Teil einer größeren Sammlung von Modellierungswerkzeugen, die darauf abzielt, sämtliche Geschäftsprozesse zu erfassen. Beispielsweise ist es möglich, einen digitalen Zwillinge zu bauen, mit dem sich Customer Journeys nachvollziehen lassen. Qualiware hat diverse KI-Algorithmen integriert, um Dokumentation und Process Discovery zu optimieren.

Wichtigste Use Cases:

• ein kollaboratives Ökosystem für Business Manager aufbauen, das ein Verständnis von der Enterprise Architecture vermittelt;

• architektonische Designelemente erfassen, um ein Wissens-Ökosystem rund um den Stack aufzubauen;

• eine breite Beteiligung in Sachen Dokumentationserstellung und -überprüfung fördern.

Quest Erwin Evolve

Das Erwin Evolve Tool von Quest hat sich von einem Datenmodellierungs-Tool zu einem System für Enterprise-Architecture- und Geschäftsprozess-Modellierung weiterentwickelt. Um die Komplexität moderner, ineinandergreifender Softwaresysteme und der von ihnen gemanagten Geschäftsprozesse zu durchdringen, können Anwender auf benutzerdefinierte Datenstrukturen zurückgreifen. Das Web-Tool erstellt Modelle, rollenbasierte Diagramme und andere Visualisierungen, die in allgemein zugängliche Dashboards einfließen. Zum Paket gehört ein KI-basiertes Modellierungs-Tool, das Whiteboard-Skizzen integrieren kann.

Wichtigste Use Cases:

• einen digitalen Zwilling für die strategische Modellierung der Enterprise Data Architecture erstellen;

• Customer Journeys verstehen;

• Services und Systeme mit Application Portfolio Management tracken.

SAP LeanIX Enterprise Architecture Suite

Die Tool-Sammlung von LeanIX umfasst unter anderem Enterprise Architecture Management und andere Bereiche, die für Aufgaben wie SaaS– und Value-Stream-Management wichtig sind – etwa um Cloud-Deployments und darauf laufende Services zu tracken. Die Daten die dabei über die IT-Infrastruktur gesammelt werden, fließen in ein grafisches Dashboard ein. Das Tool ist eng mit wichtigen Cloud-Workflow-Tools wie Confluence, Jira, Signavio und Lucidchart integriert. Das ist für Teams von Vorteil, die diese Tools bereits nutzen, um ihre Entwicklungsstrategien zu planen und umzusetzen. Seit November 2023 ist LeanIX Teil von SAP.

Wichtigste Use Cases:

• Anwendungsmodernisierung und Cloud-Migration managen;

• Obsoleszenz von Software-Services evaluieren;

• Kosten kontrollieren und managen.

ServiceNow

Die Tool-Sammlung von ServiceNow lässt sich auf verschiedene Architekturtypen herunterbrechen, darunter Assets, DevOps, Security und Service. Die Tools katalogisieren die unterschiedlichen Hardware- und Softwareplattformen, um Workflows und Datenflüsse im Unternehmen abzubilden und zu verstehen. Ausführliche Reportings und detaillierte Dashboards ermöglichen Analysen, auf deren Grundlage Risiken minimiert und die Ausfallsicherheit der Systeme erhöht werden können.

Wichtigste Use Cases:

• Tracken von Assets, Services und Systemen, die das Unternehmen ausmachen;

• Governance-Themen, Risikobegrenzung, IT-Management und Security Operations werden in einer Plattform zusammengeführt;

• durch die Integration von CRM-Tools lassen sich auch kundenorientierte Services managen.

Sparx Systems

Um Teams und Projekte verschiedener Größe und Komplexität zu unterstützen, hat Sparx vier Versionen seines EA-Tools entwickelt. Allen gemeinsam ist eine UML-basierte Modellierung, mit der sich die Komponenten komplexer Systeme tracken lassen. Eine Simulations-Engine ermöglicht “War Gaming” und vermittelt ein Verständnis darüber, wie sich Fehler ausbreiten und kaskadieren können. Sparx stellt zudem eine Vielzahl von vorgefertigten Design Patterns bereit, um Teams bei der Modellierung zu unterstützen.

Wichtigste Use Cases:

• Nachfrage- und Lastveränderungen zur Prognose künftiger Anforderungen simulieren;

• (potenzielle) Probleme durch eine Verbindungs-Matrix im Auge behalten;

• Dokumentation erstellen.

Unicom System Architect

System Architect ist eines der Angebote aus Unicoms Team Blue. Es handelt sich um ein Tool, das ein Metamodell verwendet, um automatisiert so viele Daten wie möglich über die laufenden Systeme zu sammeln – manchmal auch durch ein Reverse Engineering von Datenflüssen. Dieses systemweite Datenmodell kann über benutzerdefinierte Dashboards Teammitgliedern aller Rollen zugänglich gemacht werden. Ein weiteres erwähnenswertes Feature: Die Ressourcenzuweisung lässt sich mit Hilfe von Simulationen optimieren.

Wichtigste Use Cases:

• Was-wäre-wenn-Fragen zum Architekturmodell stellen;

• ein Metamodell von Daten und Systemen aufbauen;

• Migrations- und Transformationspläne erstellen.

ValueBlue BlueDolphin

Dieses EA-Tool sammelt Daten auf dreierlei Art:

1. Es importiert Basisdaten auf der Grundlage standardgesteuerter Automatisierung (ITSM, SAM).

2. Es arbeitet mit den Dateiformaten von Architekten und Systemdesignern – etwa ArchiMate oder BPMN.

3. Es gibt Fragebögen an andere Stakeholder heraus, die auf anpassbaren Vorlagen basieren.

Die aufbereiteten Informationen werden in einer visuellen Umgebung bereitgestellt, die Auskunft über die historische Entwicklung von Systemen gibt.

Wichtigste Use Cases:

• systemweite Daten von internen und externen Stakeholdern automatisiert und formularbasiert erfassen;

• zukunftsorientierte Reportings erzeugen, um den Change zu überwachen und voranzutreiben;

• Kooperation und Zusammenarbeit durch offenes Data Reporting fördern.

(fm)

Dieser Artikel ist im Original bei unserer Schwesterpublikation CIO.com erschienen.

When it comes to building an effective AI stack, context is king and power isn’t everything it’s cracked up to be.

“Smaller, smarter, safer — this is a bet our company has taken in how we deploy AI internally,” said Ricky Thakrar, head of sales and account management at Zoho, provider of a suite of popular cloud-based software solutions for sales, marketing, and finance.

“I’m on the business side, and so decisions made by our CIO and IT folks affect me directly, and my teams’ workflows and processes,” he added.

Speaking to a room of tech leaders at the CIO 100 Leadership Live New York event last week, Thakrar explained that every company wants the speed of AI-generated work wedded to the quality of human work, even though these two are diametrically opposed. No amount of model upgrades or spend will close that gap, so the only way forward is to architect your way out. Thakrar encapsulated this idea in a simple formula:

• Smaller: Stop deploying maximum firepower on every task. Many tasks don’t need it.
• Smarter: The system around the model decides more than the model does.
• Safer: Verify at the point a mistake gets locked in, not just downstream of it.

He noted that organizations that win with AI won’t be those deploying the biggest, most powerful models or the most sophisticated architecture, but the ones that figure out that the model is the easy part and the right architecture is harder. That means understanding the hardest element, and the biggest differentiator, is building a human system that learns and compounds alongside agentic systems.

To get it right, organizations need to prioritize the context layer. The size of frontier models like the GPT series, Claude, and Gemini mostly exist to compensate for missing context, Thakrar explained. Without enough context, models need to be able to reason harder and infer more about what a user actually means because it doesn’t know the user’s account, process, or history. A rich context layer makes it possible for enterprises to run workloads on much smaller, lower-power models.

“The intelligence moves from the model into the architecture around it,” he said.

A steep learning curve

One of Zoho’s earliest AI agents was a churn management agent to help the account management team detect churn in customer subscriptions. So when a subscription became inactive, the agent would collect context from notes, meeting recordings, and Zoho’s data enrichment tool, then create a summary of reasons the account might have churned, and schedule a call.

“What happened was I got this churn agent a couple months later, already embedded in our CRM, and within a week my team no longer trusted that agent,” Thakrar said. “The reason is we forgot to collect one very key point.”

In Zoho’s CRM, when a customer buys a bundle of products, that bundle is represented as a single line item. That means the status of any products the customer may have previously purchased individually changes to inactive as they’re moved to the bundle. That’s not churn, but it was interpreted it that way. Zoho fixed it in the second version of the agent.

Then a new problem arose. Many potential customers first purchase Zoho products as pilots or sandboxes. As those customers move from pilot to live instance, they close down the pilot versions. And again, the CRM would record that as subscriptions going inactive.

“The trust deteriorates again because everyone got excited for version 2,” Thakrar said.

Sometimes, a certain product might not be the best fit for a customer and Thakrar’s team will suggest the customer move to another product. That’s deliberate churn, not a churn risk.

“You may have a similar story like this where the agent sounds so good, it’s going to do something quick and add value, but it’s missing context from the account managers, and there are so many more pieces we’re still building out,” Thakrar said. “It’s been almost a year and the problem I have is my team still doesn’t trust it. They’ll see [a message from the agent] and go out and do all the research anyway to make sure it gave the correct answer.”

The team is more on top of potential churn, though, but the promised productivity gains have yet to materialize because the agent has to earn back lost trust due to a lack of context.

“My goal for this year is having an AI-assisted customer journey from sales to account management where the handoff is clean, the context flows, and every piece of information we gather about a customer is weighed, identified, and coached so the sales team can close more deals,” he said.

Zoho’s early experience with agents has led to the idea that constrained, context-rich, deterministic architectures consistently outperform expensive models bolted onto fragmented systems. It all comes down to three pillars: routing, harness, and specialization.

Routing

Routing is about sending workloads to the proper model for the job, which entails providing enough context to a given task that a small, cheap model can handle it without the need for spare reasoning capacity to fill gaps.

Frontier models are expensive and companies can burn through a year’s budget worth of tokens in months. But most tasks can be handled by much smaller, more constrained models at a fraction of the cost.

“You don’t always have to pay the frontier guys for every task,” he said. “We’ve observed with some clients that we could save them 95% with a 3 billion parameter model.”

Harness

An AI agent harness is the software infrastructure scaffolding around an LLM that differentiates an agent from a chatbot. It’s what enables an agent to act on tasks rather than simply respond to prompts. A model reasons through a problem and decides what to do about it. The harness connects the model to the tools, systems, memory, guardrails, and execution environments required to perform the actions determined by the model. The term is frequently used more or less interchangeably with orchestration layer.

“It’s the process around the model, which matters way more than the model itself,” Thakrar said.

In benchmark tests, a superior harness on a less powerful model produces better results than an inferior harness on a much bigger model.

For the best results, Thakrar said, it’s essential to understand the deterministic and non-deterministic elements of a given workload, and build that into the architecture. Machines can read, organize, and validate, and they excel at deterministic tasks. Humans, on the other hand, are exceptional at non-deterministic tasks like judging, synthesizing, and deciding.

Those non-deterministic tasks in a process are the ideal point for AI agents to incorporate a human in the loop, what Thakrar calls human harness. He pointed to a stakeholder mapping agent Zoho built for sales as an example, which takes the context of an initial meeting and third-party enriched data like a LinkedIn profile, weighs probabilities, and makes an educated guess about the stakeholder map.

“The initial goal was just to eliminate that task completely from the human workflow,” he said. “The stakeholder map is done, it’s in the folder, and you can look at it.”

But the agent would struggle to capture nuance. The meanings of titles in organizations always vary, and the politics and dynamics of any given meeting can be difficult for an AI agent to discern. Rather than keep feeding the agent data to try to make it intelligent enough to make those determinations, it was simpler and more efficient for the agent to create a proposed stakeholder map and hand it over to a human who could make changes and explain why those changes were necessary.

Ultimately, Thakrar said the agent still saved human team members time because the stakeholder map was usually pretty close, and the corrections also helped the model grow smarter by adding richer context.

Specialization

Specialization is transitioning a process from testing on a frontier model to production on a much narrower, smaller model. Once you’ve proven that an agent can do a job well, you want to stop paying master-craftsman rates to keep doing that one job well.

Specialization is all about capturing your subject matter experts’ best judgement and pattern recognition to build an open-weight, open source, trained, and fine-tuned model that can be deployed in your own data center.

“The true enterprise bet is to keep that orchestration layer, which is your IP and knowledge, in house,” Thakrar said. “You don’t want to host that on someone else’s model. The goal of everyone in enterprise should be to run, train, and host their own models.”

The traditional methods of measuring success no longer tell the full story. Here’s what should replace them — and why.

In brief:

• AI is transforming software as a service (SaaS), and the old ways of keeping score no longer apply.
• Smart companies are evolving new metrics that provide deeper insight into how AI-native software is performing in a new marketplace.
• These changes impact everything from pricing to valuations.

The transformation of the software-as-a-service (SaaS) industry toward AI-native operating companies is rapidly changing the unit of value across the industry.

The traditional metric of seats — which measured access — is rapidly giving way to credits designed to measure work performed. This evolution is upending the industry in multiple ways, impacting everything from pricing to enterprise valuations.

While many companies still cling to seat-based metrics to measure growth, efficiency and durability, the future is likely to be one in which companies utilize a credit-centric metrics framework, with seats and outcomes as the bookends of a spectrum.

Why do software companies need new metrics?

Why the rethink, and why now? There are five major forces that are driving this shift:

1. The unit of value is changing. Seats measured who could access software, and credits measure what the software actually does. But in an AI-native world, agents don’t have seats; they have workloads. Over the past 18 months, every major SaaS platform has moved to some forms of credit or consumption unit.
2. The cost of goods sold (COGS) is exploding. AI inference adds real per-unit costs that scale with usage. In an AI-native world, software companies can’t scale to infinite users at near‑zero marginal cost as before.
3. Buying is moving up the org chart. AI-native applications shift purchasing to higher-level operators — such as line-of-business leaders or chief operating officers — which expands the market from software budgets to labor budgets. And because AI agents replace services as well as software, the total market opportunity is 3x to 10x larger than traditional SaaS.
4. Time to value (TTV) is collapsing. With AI-native tools, customers start seeing meaningful results in weeks rather than quarters. Onboarding and setup are fast, workflows are pre-built, and there’s no need for extensive customer success or professional services — dramatically reducing implementation time and costs.
5. Retention is bifurcating. AI forces clarity in a way that traditional SaaS couldn’t. Products that can provide value become even “stickier” and retain customers. Those that don’t churn faster. In an AI-native marketplace, the middle disappears.

How this shift is impacting pricing

Given how AI-native software is transforming the market, the shift to more variable pricing options is inevitable.

Seats won’t go away completely. Subscription pricing based on the number of users is stable and predictable and will continue to work for some customers. Tokens — the use of pass-through pricing for underlying compute — will fit those customers where the AI feature is commoditized or the buyer wants transparency into costs.

Credits will likely become the dominant architecture because they provide a simple metric for both customers and providers. The vendor sets the conversation ratio between credits and underlying compute, shielding the customer from inference cost details. Credits are easy to understand and can be packaged into annual contracts for multiple features and products.

Finally, the industry will likely see some move toward outcome-based pricing for results such as resolved tickets, recovered revenue or qualified leads. This strategy will mostly be limited to verticals where it is easy to prove AI impacted the result.

Where a software vendor sits on this spectrum is a signal of differentiation and pricing power. Credits are where most defensible AI-native businesses are landing because they balance customer predictability with vendor margin control.

How AI upends classic SaaS metrics

When SaaS was in its infancy, companies settled on key metrics designed to answer a small set of core questions. Are we growing? Are customers using the product? Are we retaining and expanding accounts?

But as AI upends software itself, it is also requiring companies to adopt new metrics to track success. These new metrics fall into three primary buckets, rebuilt around the pricing spectrum described earlier and the trend toward credits as the primary frame:

Revenue composition

• Committed credit annual recurring revenue (ARR) vs. burndown ARR: Measuring the credits sold on annual commitment vs. those consumed and replenished. This is the single most important split for valuation. Committed credits behave like subscription and burndown behaves like usage.
• Credit utilization rate: The percentage of purchased credits consumed per period. This is a leading indicator of renewal sizing.
• Credit burn velocity: How fast is a customer consuming their credits, and is that consumption increasing or decreasing quarter over quarter? This metric predicts expansion or contraction before it shows up in ARR.
• Effective price per credit: The real revenue per credit after discounts, overage and rollover, which can detect revenue leakage and help companies set smarter guide rails.

Margin reality

• Credit margin: The gross profit the company earns per credit after subtracting inference costs. This is the core economic unit for AI-native, usage-based businesses — the replacement for gross margin per seat used in SaaS.
• Inference-adjusted gross margin: By carving out AI inference costs separately in the P&L statement, you can see true AI margins, avoid hiding deterioration inside blended SaaS margins, and clearly distinguish AI economics from legacy SaaS economics.
• Compute leverage ratio: This metric measures how efficiently the business converts compute spend into revenue. It shows whether your AI margins are improving as you scale.
• AI-adjusted “Rule of 40”: This updated metric recalibrates the traditional growth and profitability benchmark to account for AI’s lower gross margins and variable inference costs, giving a more accurate picture of business health for AI-native companies.

Behavioral and value signals

• Time-to-first outcome: Replaces traditional onboarding metrics. Tracks how fast a customer reaches their first measurable result.
• Adoption: AI-native adoption is measured by workflow penetration and active agent density, not seat count. As AI replaces human-driven usage, the unit of adoption shifts from people to automated workflows and agents.
• Net credit retention (NCR): Credit-volume retention across the customer base, tracked separately from net recurring revenue to avoid price-change impact.

Along with these new metrics, the industry’s transformation is prompting companies to retire or recalibrate old SaaS measures, including per-seat ARR as a primary key performance indicator (KPI), traditional magic number calibrated to subscription dynamics, unadjusted Rule of 40, customer success metrics tied to human touchpoints, and blended gross margin without AI COGS carve-outs.

What does this mean for enterprise value calculations?

As the internal metrics of success change, so do the ways the investment community measures growth and long-term viability.

Increasingly, a company’s valuation multiple depends on whether its revenue behaves like committed subscription ARR or volatile usage ARR, and the commit‑to‑burndown ratio is the metric investors use to decide where the company fits.

For example, a business with 80% committed credit ARR could trade closer to subscription comps and one with 80% burndown could trade closer to usage comps even though both have the same types of customers. Being able to proactively explain the commit‑to‑burndown mix can help companies avoid undervaluation.

In addition, utilization is expected to replace net promoter scores and seat usage as the primary predictor of churn or expansion. Low utilization guarantees downsizing at renewal, so companies must track utilization cohorts the same way SaaS tracks logo retention cohorts today.

We’re also seeing an inversion of the operating model, with R&D and COGS moving up the P&L and sales and marketing (S&M) and customer success (CS) moving down or sideways. The net operating leverage profile is structurally different from classical SaaS, and the cost-to-scale curve looks different too.

Finally, credit margin engineering is a hidden value-creation lever. The gap between price per credit and cost per credit is set by the software vendor and can be optimized. Most operators have barely started managing this rigorously, and the ones who do will pull away on margin.

What this means for leaders, boards and investors

The shift from classic SaaS metrics to new AI‑native measures isn’t cosmetic. It represents the seismic change the industry is experiencing as AI matures and transforms products and organizations.

While these metrics — and perhaps others yet to be determined — may evolve over time, there is no doubt they are already changing how AI companies allocate capital, price products, incent sales teams, evaluate performance and communicate with investors.

It’s important to remember that SaaS metrics were practical tools for a specific era of software. As that era draws to a close, winning companies will choose new metrics that shape behavior and drive smart decision-making.

The views reflected in this article are the views of the author and do not necessarily reflect the views of Ernst & Young LLP or other members of the global EY organization.

This article is published as part of the Foundry Expert Contributor Network.
Want to join?

Over the past few years, I have learned that basic prompts produce inconsistent, hallucination-prone results that no executive would trust in production. What turned the tide was my move to advanced prompting techniques. These weren’t theoretical experiments; they became a practical foundation for reliable, measurable outcomes. I want to share the techniques that consistently delivered the biggest gains in my projects, complete with real before-and-after examples, copy-paste templates, lessons from failures and guidance on when to evolve beyond prompting to agentic systems.

Why advanced prompting still matters in enterprise settings

Sophisticated prompting remains essential for control, reliability and compliance. If you “ask nicely” and hope for the best, you need deterministic behavior, auditable reasoning and minimal risk of hallucination. Here’s what worked for me.

1. Chain-of-Thought (CoT) and its variants: Unlocking step-by-step reasoning

The problem: Models would jump to conclusions on complex analysis tasks, especially involving data interpretation or multi-step logic.

What I did: I started explicitly instructing the model to “think step by step” and show its reasoning.

Before (basic prompt): “Analyze last quarter’s sales data and recommend three actions.”

After (CoT prompt):

“You’re a senior business analyst. Analyze the following sales data step by step: [data]. First, identify the key trends. Second, calculate the rates and anomalies. Third, link findings to business context. Finally, recommend the three prioritized actions with expected impact. Explain your reasoning at each step.”  

Results: Accuracy and depth improved dramatically.

Variants that worked well: Self-consistency. I ran the same CoT prompt multiple times and took the majority consensus. This reduced variability significantly.

Template you can use:

You are [expert role]. Solve this problem by thinking step by step.

[Task or question]

For each step:

1. State your observation or calculation.

2. Explain the implication.

3. Proceed only when confident.

Final answer in this format: [structured output]

2. Tree-of-Thoughts (ToT): Exploring multiple reasoning paths

For truly complex decisions such as resource allocation or risk assessment, linear CoT isn’t enough. Tree-of-Thoughts lets the model generate and evaluate multiple branches.

Example: I was helping a client evaluate three potential vendor platforms for an AI deployment. A standard prompt gave a superficial comparison. With ToT

Prompt Snippet:

Explore three different reasoning paths for selecting the best vendor platform:

Path 1: Focus on cost and scalability.

Path 2: Focus on security, compliance and integration.

Path 3: Focus on innovation and long-term roadmap.

For each path, evaluate pros/cons against our requirements [list].

Then, compare the paths and recommend the strongest overall option with justification.

Outcome: The model surfaced nuanced trade-offs (e.g., one vendor had superior security, but higher integration cost).

When to use: Strategic planning, troubleshooting or scenarios with high uncertainty and multiple viable approaches.

3. ReAct (Reason+ Act) and prompt chaining: Moving toward agentic behavior

One of the biggest leaps I have noticed comes from combining reasoning with tool use and chaining prompts.

ReAct example: (used in data analytics workflow)

You are an AI analyst with access to tools. For the query below:

1. Reason about what information you need.

2. Choose the appropriate tool or action.

3. Observe the result.

4. Repeat until you can answer confidently.

Query: [user request]

In practice, I chained this with retrieval tools. One automated quarterly compliance reporting; the system reasoned about required data, pulled relevant records, validated them, and generated the reports.

4. Meta-prompting and self-reflection: Letting the model improve itself

Use the model to refine its own prompt. This is a huge time-saver.

You are an expert prompt engineer. Improve the following prompt for clarity, structure and effectiveness with [target model]. Make it more precise while preserving intent.

Original prompt: [paste]

Provide the improved version and explain your changes.

Self-reflection loops (asking the model to critique its own output and revise) are a game-changer for content generation and code-review tasks.

5. Multimodal and structured output techniques

With vision-enabled models, I started combining text with images (e.g., uploading architecture diagrams or dashboards).

Tip from experience: Be extremely specific in describing what the models should focus on.

Best practices I learned the hard way

• Start simple, then layer complexity: Over-engineered prompts from Day One usually backfire.
• Model specific tuning: Some models respond better to XML delimiters; others to explicit reasoning.
• Evaluation and versioning: Treat prompts like code if you track versions and run automated evals.
• Security guardrails: Always include instructions against prompt injections and respect data boundaries.
• When to stop prompting: For repetitive, high-stakes workflows, move to full agents or an orchestration framework.

Final takeaways for technical leaders

Advanced prompt engineering has now become a core competency for anyone responsible for enterprise AI outcomes. Start by picking one technique and apply it rigorously to a real business problem. Document before/ after and you will notice why it’s worth mastering.

The field continues evolving towards more automated and agentic systems, but the ability to precisely direct AI reasoning remains foundational.

This article is published as part of the Foundry Expert Contributor Network.
Want to join?

OpenAI has launched Presence, an enterprise service for deploying voice and chat agents that can resolve customer and employee requests, potentially automating some work now handled by frontline support teams.

The agents can answer questions and operate IT systems, and enterprises can decide what actions the agents may take and when they should seek human approval for actions or transfer a case to a human.

OpenAI is already using Presence internally for its English-language phone support channel, where it verifies callers and uses account information to complete approved actions. The company said the system resolves 75% of inbound issues without human assistance.

Another OpenAI service, Codex, can be used to monitor agents and suggest updates or improvements to processes. In OpenAI’s own tests, suggestions from Codex helped reduce handoffs to humans by 15 percentage points over 10 days, it said. Presence also includes simulation and evaluation tools that allow companies to test an agent before deployment. The tests assess whether it reaches the correct outcome, follows company policy, and hands a case to an employee when required.

OpenAI intends each Presence deployment to deal with one kind of task, for example billing issues, insurance claims, or employee IT service requests, with agents getting only the knowledge and system access required for that task.

Presence is not a self-service product: Enterprises will have to sign up for the limited availability program, with integration performed by OpenAI or selected global systems integrators.

Companies exploring or testing Presence include Spanish bank BBVA, which is evaluating the service for everyday banking support in Mexico, and Japanese technology group SoftBank, which is using it in trials involving Japanese-language customer interactions. Australian insurer IAG is assessing whether the technology can help it respond to surges in customer demand during severe weather events.

Workforce impact

OpenAI’s announcement did not address the potential effect of Presence on employment. But its claimed automation rate raises questions about how the technology could affect staffing in customer service and other support functions.

Pareekh Jain, CEO of Pareekh Consulting, said CIOs should regard the 75% figure as evidence that the technology can work, rather than as a benchmark that every enterprise can expect to reach.

Jain said OpenAI’s deployment benefits from being built around the company’s own products and data. Large enterprises may achieve lower automation rates because they must contend with fragmented legacy systems, uneven knowledge bases and more complex compliance demands.

“Most organizations should expect lower initial automation levels that improve over time as the AI agent is refined,” Jain said.

The first workforce effect is more likely to be slower hiring than immediate layoffs, according to Tulika Sheel, senior vice president at Kadence International.

“The roles most exposed are likely to be repetitive, high-volume functions such as frontline customer support and routine back-office processing,” Sheel said. “However, I would expect the first impact to be on hiring and team growth rather than immediate large-scale job cuts. Over time, enterprises may redesign roles around AI-assisted workflows, with humans focusing more on complex cases, escalation, and relationship management.”

Jain said Tier-1 support agents handling predictable queries would face the most exposure. Broader reductions would become more likely only after companies reorganize their operations around the technology.

However, Lian Jye Su, chief analyst at Omdia, said Presence is unlikely to increase the threat of job displacement because companies have used similar customer-support automation from vendors such as Genesys, NiCE, Five9 and AWS for years.

Enterprises are more likely to use Presence alongside employees, with AI handling routine requests while people remain responsible for work requiring judgment and empathy, Su said.

Cost and operational risks

Analysts said CIOs should examine whether Presence can maintain resolution quality as usage grows, since fewer human handoffs could leave employees dealing with a more difficult mix of cases.

“The key question is not simply how many tasks AI can handle, but whether it can handle them reliably at scale,” Sheel said.

The financial case will depend partly on the cost of connecting Presence to existing systems and maintaining the controls needed to govern its use, according to Jain. “Often the biggest cost of enterprise AI is not tokens but integration and governance,” Jain added.

Companies will need to determine what systems and data the agents can access, monitor their performance, and audit the actions they take. Those investments could offset early savings.

Su said the complexity of enterprise IT will make it difficult for OpenAI to automate entire workflows on its own. Enterprises will still need to work with other technology providers and human employees, while CIOs will favor systems that can be audited and integrated with existing infrastructure.

Jain said the economics could improve if companies use the same integrations and governance controls across additional workflows.

I have worked alongside enterprise technology for more than 30 years and watched AI evolve from a lab experiment into the modern boardroom’s core focus. However, the last few years of implementing AI alongside our customers have delivered our most profound reality checks.

The initial hype has faded, leaving CIOs to drive real enterprise value. Based on my experience implementing Google, OpenAI and Anthropic technologies, here are the fundamental, technology-agnostic lessons every leader must anchor their strategy around.

AI as a leadership multiplier

The most common tactical error we see is treating AI as an isolated technology project. What I have observed among our customers is that true success does not come from organizations that define a standalone “AI strategy,” but rather from those leaders that integrate AI into their business strategy.

When our customers isolate AI and define an AI strategy, it inevitably treats it like a “technological toy” to experiment with. This approach yields fragmented, orphaned initiatives that fail to scale because they are fundamentally disconnected from their core corporate objectives. What I learned is that AI is not the ultimate destination; it is a powerful catalyst. We have replaced “What can AI do for our customers?” with a more strategic question, “How does AI accelerate their existing business goals?”

Think of AI like electricity. No modern corporation designs a standalone “electricity strategy.” Instead, all companies route it invisibly across the entire organization to illuminate offices, power production lines and drive communication. AI must be woven into the enterprise fabric in the exact same way, acting as an underlying utility that supercharges your existing operational model.

Integrating AI into the broader business strategy also dictates how we measure success. It forces a shift away from short-term tech vanity metrics and anchors the technology into a long-term roadmap.

When AI remains trapped within the IT department of our customers, we notice that it is relegated to a mere “software experiment.” To become a true competitive advantage, we observed that AI requires intense cross-functional orchestration. This perspective does not diminish the merit of the technical team; their expertise is fundamental for establishing the architecture, data governance and tools your enterprise requires. However, while IT builds the foundational infrastructure, it lacks the organizational authority to decide what should be built on top of it. Only the CEO or the owner of the company can step in to ensure AI leaves the “toy project” phase and integrates into the DNA of the organization.

The requirement for top-down, executive ownership stems from three critical realities observed in the field:

• Silo-smashing and data collaboration: True enterprise AI is data-hungry and that data lives across disparate business lines, finance, operations, marketing and customer service. Only the CEO possesses the cross-functional authority to demand that data silos be dismantled.
• Cultural transformation and fear mitigation: AI triggers widespread anxiety over job displacement across all industries and hierarchies. When relegated to an “IT project,” resistance spikes as teams view it as a threat to their livelihoods. When I saw the CEO lead this cultural shift directly is when I noticed the best results.
• C-Suite education and strategic alignment: The mandate for AI capability cannot just be delegated downward; the transformation must begin at the very top. I have conducted more than 70 presentations for the Board of Directors and C-Level teams. These people need to be actively educated not on technical code, but on specific business use cases, return on investment (ROI) frameworks and how AI resolves core organizational bottlenecks.

PwC’s data found that only 12% of enterprises have achieved both cost and revenue benefits from AI. Those elite 12% succeeded precisely because their CEOs embedded AI extensively across strategic decision-making and cross-functional workflows. AI is simply too disruptive and too critical to be left exclusively in the hands of technical experts. If AI is not on the CEO’s weekly agenda, it is fundamentally missing from the company’s true strategy.

AI as a new operational framework

Traditional IT systems have operated on strict algorithmic certainty: if you input a specific set of data, the system executes an immutable line of code and guarantees the same, predictable output every single time.

AI completely breaks this paradigm. Because modern AI is built on probabilistic models, it does not execute static formulas; instead, it predicts the most likely correct response based on mathematical probabilities. This means that AI solutions carry an inherent, small percentage of uncertainty and variability. A prompt entered today might yield a slightly different, though contextually valid, output tomorrow.

Executive leadership and organizational cultures must be actively educated to accept and navigate this fundamental shift. Traditional quality assurance frameworks for software are designed for a 100% success rate. Applying this rigid standard to AI will paralyze your initiatives, keeping 80% of your projects trapped eternally in the pilot phase. This happened to us in a food and beverage company in Latin America a couple of years ago. After this experience, we started to include conditions in our contracts that tolerate statistical margins of error and still define the project as a success.

In terms of cost calculation, we had to teach CIOs and business managers to forget the monthly subscription model for AI and learn to manage the primary unit of exchange in modern AI: the token.

To understand AI costs, executives must understand how large language models process data. AI models do not read full words; instead, they break text, images or code down into “pieces” called tokens. As a baseline, every 100 words process as approximately 130 to 140 tokens. Because the major AI providers use the token as their currency, your business is billed dynamically based on the exact volume of tokens consumed by every query submitted (input) and every response generated (output).

Many leaders believe AI costs are fixed due to flat-rate enterprise tiers ($25–$30/user). This is a temporary illusion. These venture-capital-subsidized rates mask true operational costs and come with dynamic usage limits. Modeling long-term ROI on them guarantees a severe budget shock when true consumption pricing takes over.

The solution is not to halt AI adoption; doing so means losing your competitive edge. Instead, the cost per token must cease to be treated as a technical footnote relegated to the IT department. It must be elevated to a core business variable.

Risks in the AI adoption model

Since the beginning of the AI boom, I have seen all our customers making a critical tactical error that could cost them heavily in the medium term: they are focusing only on operational efficiency (reducing costs with AI).

I have observed that an alarmingly high percentage of companies remain trapped in pilot phases focused exclusively on short-term cost reduction. Bain & Company’s global Automation and AI Pathfinder Survey found that the largest share of companies measuring their AI initiatives (exactly 40%) realized cost reductions of 10% or less, heavily missing their internal targets. Our customers are putting too many resources and effort into marginal financial gains and in doing so, they are jeopardizing their most valuable assets: service quality, resilience and customer trust.

Utilizing AI solely to slash headcount or cut operational corners is a dangerous trap that introduces severe field liabilities. A financial service organization in Latin America announced that they saved $1 million in customer support by replacing humans with AI chatbots. However, the mid-term reality revealed a different story: a damaged brand reputation due to AI errors and an influx of frustrated clients fleeing because the automated system cannot handle special cases.

Putting a company on an extreme AI diet might make it look leaner on next quarter’s financial statement, but over-indexing on cost-cutting will ultimately leave the business too weak to compete when market dynamics shift. We are now inviting our customers to change the question from “How much money will AI save us?” to “How will we leverage AI to exponentially increase the long-term value of our enterprise?”

Deploying enterprise AI is a marathon, not a sprint, and the terrain changes with every mile. The organizations that thrive in this next era will be those that transition from fascination to discipline, treating AI not as a magic bullet for immediate savings, but as a core capability that demands rigorous governance, architectural foresight and cultural maturity. Navigating this shift requires moving past the theoretical hype and anchoring decisions in raw, field-tested reality.

As we continue to deploy these technologies across industries, the blueprint for success is being rewritten in real time. Let’s keep this conversation going as we map out the future of business intelligence together.

This article is published as part of the Foundry Expert Contributor Network.
Want to join?

Model Context Protocol (MCP), the emerging standard for connecting AI models to external tools and enterprise data, is undergoing its biggest architectural overhaul yet.

The latest release candidate, scheduled for release on July 28, removes protocol-level sessions in favor of a stateless architecture, a change which industry experts say is intended to make MCP easier to deploy across standard cloud infrastructure as enterprises move AI pilots into production.

“The session-based model made sense when MCP servers were local processes on a developer’s laptop. In production, it became an operational tax,” said Muskan Bandta, cloud associate at ZopDev.

“When your infrastructure team asks whether MCP services can scale like other cloud applications, the answer used to be ‘not quite.’ With the move to a stateless architecture, the answer is now yes,” Bandta added.

Earlier versions of the protocol maintained information about every client connection, meaning servers had to keep track of each session throughout an interaction. While that approach worked well for local development, it complicated deployments across multiple servers because requests often had to be routed back to the same machine, limiting scalability and making MCP a less natural fit for modern cloud architectures.

“Under the new stateless design, every request contains the information needed for any available server to process it independently. Applications that need to maintain context across multiple requests can still do so, but developers must now manage that state explicitly rather than relying on the protocol itself,” she said.

This transition to a stateless design goes beyond simplifying infrastructure by fundamentally changing how AI applications manage and share context across tools, according to Amit Jena, AI development manager at IT consulting firm Kanerika.

Instead of keeping application state hidden inside protocol sessions, the new design makes it explicit, allowing AI models to access, reason over, and pass that information between tools, giving developers greater control over how context is preserved and shared across tools, Jena said.

It should also make AI workflows more portable, resilient, and easier to orchestrate across distributed environments, he said.

MCP’s new features

Other changes to MCP include the addition of a Multi Round-Trip Requests (MRTR) mechanism that changes how AI agents request additional information they need to complete a task.

Instead of relying on a persistent connection between the client and server throughout the interaction, the new mechanism lets the server request additional input through a standard request-response exchange before continuing the task, Jena said.

Routable transport headers, another addition, enable API gateways and other networking infrastructure to identify and route MCP requests without inspecting their contents.

They reduce processing overhead, lower latency, and let enterprise teams enforce routing, rate-limiting and security policies more efficiently using existing API management infrastructure, Jena said.

MCP is also getting an updated authorization framework built around OAuth 2.1 and OpenID Connect; interactive MCP Apps; and deterministic caching of tool and resource listings to improve LLM prompt-cache hit rates, potentially saving on token costs.

Rebuilding the trust boundary

The MCP release steering committee also decided to deprecate some legacy features, including Roots, Sampling, Logging, the older HTTP+SSE transport and Dynamic Client Registration, although these will continue to work in this version and any other released over the next year.

The deprecation of Sampling is likely to have the biggest impact because it changes who is responsible for interacting with foundation models, said Jena.

“Sampling let MCP servers invoke the LLM through the client, which meant the server had a callback path into the model without owning that connection. Deprecating it means rebuilding that trust boundary,” Jena said. “Your server now calls the model provider directly. That changes your network architecture, your auth model, and depending on how you’ve built cost attribution, your billing flow.”

The year-long transition period will be enough for teams to audit their sampling dependencies now, said Jena: “The risk is that teams who haven’t implemented sampling themselves won’t know if a third-party MCP server they’re depending on uses it.”

Updated MCP SDKs

To accompany the protocol update, there are updated MCP SDKs for Python, Typescript, Go, and C#. These support both the old and new protocol versions, so new clients can continue communicating with older servers, while updated servers will also support older clients, reducing the risk of immediate disruptions.

That backward compatibility should make the transition largely incremental, except for enterprises that built custom infrastructure around MCP’s earlier session-based architecture, Bandta said.

Identifying and auditing those session dependencies may not be easy, Jena warned.

“Session management complexity tends to be hidden across multiple layers — the gateway config, the deployment scripts, the monitoring dashboards. The code change is small; finding everywhere the assumption lives is what takes time,” he said.

This article first appeared on InfoWorld.

Four months of Claude Code — that’s all it took for Uber to burn through its entire annual budget for AI. Token after token, engineers embraced the platform with few control mechanisms tying costs to outcomes. The result was a budget runaway and a clear case study in how limited oversight snowballs into an AI blowout.

This is a phenomenon I like to call “shadow tokens” — AI credits paid for by the company but largely invisible to decision-makers. Too many engineers have the final say over how much they consume and, therefore, what it costs. This all-you-can-eat attitude is part of the reason why Microsoft is reportedly winding down many internal licenses across key engineering teams and why one in five organizations is missing its AI spend forecast by more than 50%.

And the trend is only accelerating. By 2028, Gartner predicts that AI coding costs (driven by this kind of ungoverned consumption) will be as much per developer as the salary companies pay that person.

LLMs and agents introduce a new class of variable cost that scales with behavior rather than headcount, putting enterprises on the hook for tools that balloon with workload. I don’t see this as enterprises overspending because they’re reckless — it’s down to a lack of managerial oversight, budget alignment that demands a proven return on investment, and engineer education on how much is too much.

Going forward, CIOs need to thread the AI needle between governance that encourages transparency and reasonable spend without stifling innovation.

When shadow tokens result in real costs

The issue is that AI isn’t a traditional line item. Previously, enterprise leaders onboarded software-as-a-service (SaaS) with a good idea of the total cost. An allocated software seat or annual contract was a known quantity. The cloud added some variation (with fluctuations depending on hosting size), but instances were still modelable. AI flips this status quo on its head — the unit of consumption is behavior and the cost is exponential.

And these specifics aren’t immediately apparent at pilot. Tools can appear inexpensive in controlled experiments yet unpredictably scale depending on session length, context window size, model selection and whether agents run in parallel. This is the fallacy of the $20-per-seat enterprise plan — tokens are charged separately at API rates with no ceiling. The final dollar value of any session is set by factors that finance can’t always model in advance, particularly when these decisions usually rest with the engineers themselves.

According to Deloitte, only 21% of organizations deploying agents have a mature governance model, a real concern because they’re token-eating machines. This is what was happening at Uber — Claude Code in agentic mode was autonomously reading codebases, planning changes across dozens of files and opening pull requests. Each step quickly adds up, with Anthropic’s own documentation noting that agents consume approximately seven times as many tokens as standard sessions.

This is shadow IT and shadow AI, evolved. This time, however, many leaders approved the tool in question without guardrails governing consumption. AI hype adds fuel to the fire and normalizes long sessions. Uber’s CTO, for example, described a company-wide shift toward “agentic software engineering” with employees “who are quietly experimenting, quietly shipping and quietly pushing things forward”. This is an exciting way to test the limits of what’s possible, certainly, but it’s also a position that goes a long way to explaining how the company spent its annual AI budget by April.

Shifting the culture from usage to yield

Engineers haven’t done anything wrong here. In fact, they’re adopting and experimenting as instructed, with Uber creating leaderboards and ranking users by token consumption. More use led to a better ranking, reflecting a culture that lauds new ways of doing things. This behavior is known as “tokenmaxxing,” and its principal knock-on effect is shadow tokens — quantity-over-quality processes that leaders struggle to control until they’re fully realized in the budget. Of course, if management treats adoption metrics as performance metrics, then engineers can’t be blamed for using more tokens. The tension is that the teams driving adoption aren’t the ones managing spend.

None of this is meant to dismiss AI’s productivity possibilities and potential return on investment. Developers save 3.6 hours per week, achieve 60% higher pull request throughput and cut onboarding time in half with automation. Meanwhile, Uber shared that roughly 11% of live backend updates were written by agents with no human in the loop. However, these wins aren’t the problem — it’s that too many teams aren’t connecting input to output. I’ve spoken to admins who discovered their token spend had tripled in a single quarter after using heavier models or accidentally doubling up on agentic applications. Nobody knew until the financial damage was done.

Automation needs to happen sustainably with an eye on the bottom line. In my view, a much better metric for achieving this is AI yield — the measurable business or engineering output generated per dollar spent on tokens. Otherwise, without a feedback loop, even genuinely productive teams are flying blind.

Stopping token waste before an AI blowout

Creating that throughline between AI investment and token consumption starts with established financial metrics. This is possible via maximum spend limits (dictated by spend tagging, workload tiering and cost-per-output benchmarks) per team or project. Then, any additional allocation requires approval, closing the loop between the engineers spending the tokens and the leaders paying for them. AI isn’t cheap and teams should demonstrate a bang for their buck.

This is something we do with our engineering team at Hexnode. Resource allocation for Claude Code and Cursor is tied directly to ROI rather than letting consumption run open-ended. Given the pay-as-you-go nature of these tools, a firm usage limit per team offers simple but essential control.

Similarly, there’s room to apply some of the governance principles IT uses for device management. Things like policy enforcement, role-based access, real-time monitoring and automated alerts can flag usage behavior in advance. Uncovering such insights at the token layer works to identify power users and prevent excessive spending.

We also need to encourage cultures that praise outputs that actually achieve efficiency. AI applications that result in shipping faster, reducing rework and cutting review cycles are gains that should be celebrated. If your company hosts leaderboards, frame unnecessary token burn as wasteful rather than valuable. The organizations creating healthier consumption habits work with their engineers to understand not just how to use AI, but what responsible use looks like and what it costs.

This is a conversation teams need to have now. Anthropic just ended flat-rate pricing for programmatic workloads from June 15. Now, agents, continuous integration pipelines and automated workflows draw from a dedicated monthly credit pool billed separately from the subscription. Once that pool is exhausted, agent tasks either stop entirely or overflow to extra billing. Work can either get very expensive or grind to a halt for teams that aren’t prepared.

Getting a grip on shadow tokens means better rules and tools connecting spend to outcomes. Only by building the financial and cultural infrastructure that encourages sustainable adoption can leaders see what they’re spending, connect it to what they’re getting and course-correct before the costs become a crisis. Ultimately, shadow tokens are only invisible if we choose not to look.

This article is published as part of the Foundry Expert Contributor Network.
Want to join?

In August 2025, 126 malicious packages landed in the npm registry. Even after the community caught the initial wave, 80 of these hidden backdoors remained actively listed.

That was enough. Over 86,000 downloads. Malicious code in PhantomRaven, packages running in the production systems of Fortune 500 companies worldwide. And throughout the entire window, not a single EDR/XDR alert.

This happened because the attack surface has expanded to a layer EDR/XDR was never designed to see: VS Code extensions, local MCP servers, and rogue AI coding assistants that inherit your engineers’ valid credentials to steal data at machine speed.

To eliminate this structural vulnerability, Palo Alto Networks acquired Koi, an AI-native developer security product engineered for proactive, precision enforcement. Below we compiled a 2026 CISO checklist you can use to audit your environment and see how Koi automates each defense from day one.

#1. Gain real-time visibility into shadow AI & extensions

Your existing asset management tracks binaries and installers, but it cannot see local VS Code extensions, MCP servers, or ad-hoc Python scripts running on developer endpoints. This visibility gap was recently exposed by the MaliciousCorgi campaign, where two marketplace extensions with 1.5 million combined installs silently harvested every file a developer opened. Neither triggered any detection because they were not binaries, not executables, not anything your inventory was built to flag. To counter this, Koi closes the gap by analyzing what extensions actually do after installation, exposing hidden data-harvesting channels running inside your active workspace.

#2. Distinguish between human and autonomous agent behavior 

When a rogue AI agent exfiltrates your proprietary source code, it uses a developer’s valid credentials during normal working hours, making the session look entirely legitimate to standard XDR baselines. Moving beyond static permission lists, Koi deploys behavioral profiling within the workspace runtime. By actively intercepting unauthenticated background tasks and blocking unauthorized file-system reads, it stops automated data exfiltration in real time.

#3. Establish guardrails for automated package updates on endpoints

Developers prioritize speed, often allowing software packages to auto-update on their endpoints the moment a new version appears. Attackers weaponize this supply chain vulnerability, as seen in the May 2026 Team PCP attack where 3,800 GitHub repositories were compromised in just 36 minutes via poisoned auto-updates. Securing agentic endpoints against these rapid breaches requires behavior-based inspection within the active workspace context. Koi operates at this layer by providing safe deployment buffers that automate version cooldowns, blocking bleeding-edge updates until they are vetted. By continuously auditing process creation within the IDE runtime, Koi instantly drops unauthorized remote connections before malicious payloads can exfiltrate credentials from the endpoint.  

#4. Enforce principle of least privilege for AI agents

AI coding assistants inherit the privileges of whoever deployed them. In practice, that means read access to production databases, write access to core repositories, and access to every secret in environment files and configuration directories. To restrict this excessive access, Koi applies dynamic sandboxing directly to AI agent processes at the kernel level. It enforces a strict zero-trust boundary that segregates sensitive workspace vectors, preventing agents from pulling data outside their approved scope without interrupting developer workflows.

#5. Maintain continuous endpoint posture management

Signature-based scanning only stops known threats. Sophisticated repository attacks often arrive as functional, high-rated software that carries no known bad signature. Koi’s research into the DarkSpectre campaign found eight browser extensions, all carrying “featured” badges from Google and Microsoft, installed by over 8 million users, silently harvesting every conversation from ChatGPT, Claude, and Gemini in the background. Koi addresses this by operating upstream: scanning marketplace listings every hour, using LLM-driven code analysis to compare what software promises against what its code does, sandboxing it, and scoring the risk before it ever reaches the endpoint.

Summary

Securing the modern enterprise is no longer about patching individual gaps. As AI agents redefine the workforce, Agentic Endpoint Security (AES) is now a strategic imperative for every CISO. By establishing a mandatory control plane for the AI-native workspace, AES ensures that your organization can scale engineering velocity without ever compromising enterprise integrity. 

Ready to secure the future of your software stack? See how Koi Agentic Endpoint Security delivers complete visibility, risk scoring, and real-time prevention across every endpoint in your enterprise.

Agentic AI has advanced from the promises-and-pilots phase of 2025 to reality and rollouts in 2026. In the process, agentic AI is transforming virtually every aspect of customer relationship management (CRM), the platform that manages sales, marketing, and customer service.

“Last year, everybody was dipping their toes into the water,” says Keith Kirkpatrick, research director at The Futurum Group. This year, agentic AI has built momentum from the boardroom down, with companies recognizing that having an AI strategy is imperative. “They feel like if they don’t embrace it now, their competitors will.”

Harry Datwani, a principal at Deloitte Digital, adds that enterprise CRM customers have transitioned from “proof of concept” to “scale and execution.”

“Across sales, service, marketing, even in the commerce space, enterprises are really using AI and agentic,” he says.

“CRM in 2026 is undergoing a structural shift, not just an incremental evolution,” says Forrester analyst Kate Leggett, noting that AI is becoming a core part of CRM infrastructure, not just a feature or an add-on. According to Forrester data, around 70% of companies are already using AI in their CRM systems, she says.

Here are the hot AI-driven trends in CRM this year.

CRM becomes an action hero

CRM platforms have traditionally served as passive, static systems of record. Now, agentic AI is transforming CRM into a powerful, real-time solution that can act autonomously.

“Organizations that rethink CRM as a real-time, AI-powered system of action — and embrace agentic AI to handle complex, unpredictable work — are better positioned to deliver exceptional customer experiences,” says IDC analyst Neil Ward-Dutton. “This approach not only enhances satisfaction and loyalty but also drives operational efficiency and business agility.”

Forrester’s Leggett says that AI-powered CRM platforms have advanced from simple data capture to real-time decision-making and execution. Standard capabilities include next-best action recommendations, call summaries, automated updates, generated emails, knowledge creation, predictive forecasting, and deal scoring.

She adds that AI agents can now execute workflows, such as routing cases, sending follow-ups, and updating records (with human oversight). They can also handle end-to-end service and sales tasks autonomously, including case resolutions and sales development activities.

Agentic drives workforce changes

AI use in CRM systems is also impacting workforce strategies.

“We used to hire for deep expertise,” says Constellation Research analyst Liz Miller. “AI has commoditized expertise because I can take all that data from my CRM and train my AI models to go deep, to know everything about any product I’ve ever sold, from what has worked, what hasn’t, every price, every sale.”

Now, instead of hiring candidates with deep expertise, organizations are looking for candidates who can go wide. “I can train a model to have deep expertise. What I can’t train for is experience, because experience is what happens when a person has gone broad across a lot of different scenarios and faced complexity across that broad scenario,” says Miller.

For example, AI systems can automate many aspects of marketing, Miller notes, but there’s no substitute for creativity: people who can interrogate the data and come up with innovative marketing campaigns that connect with customers.

Terence Chesire, group vice president of ServiceNow CRM and industry workflows, says that organizations are using agentic AI to free up team members from repetitive, lower-value activities. Those employees have now moved to higher-level roles “where they’re working on transformational deals rather than just building a spreadsheet.”

“That’s what we’re seeing as super-exciting as organizations not just free up people, but the speed and effort reduction and the friction reduction in what they can do,” he adds.

Data layer takes center stage

AI’s promise to deliver actionable customer and marketing intelligence has placed even greater emphasis on the importance on sound data management practices for CRM.

“The light bulb has flashed on very brightly for our clients,” says Deloitte’s Datwani. “Everyone is talking about AI agents, but your ability to really extract value is inextricably linked to the quality of your data and the ability to make that data accessible. What we’re finding is that despite large investments over time our clients still have fragmented data. And so, they are data rich and insight poor.”

The good news, says Datwani, is that AI agents themselves can help clean up and organize data. And vendors such as Snowflake and Databricks, along with the traditional CRM powerhouses, are offering powerful data analytics solutions. “Everyone is battling for that data layer,” Datwani says.

Forrester’s Leggett adds that CRM platforms are converging with customer data platforms (CDPs), real-time event streams, and external data sources to create connected customer data networks. These real-time, connected data models can help organizations deliver hyper-personalization at scale.

Agentic ushers in pricing complexity

The shift from license- or subscription-based pricing to an outcome or consumption pricing model has the potential to help CIOs tie their CRM costs to specific business metrics, such as the number of customer service calls resolved per hour. But it has also introduced a new level of complexity when it comes to budgeting for CRM costs.

For example, Chesire says ServiceNow’s CRM pricing plan starts with a baseline subscription model, and on top of that, customers get a certain number of AI tokens per user and can buy additional tokens as AI usage ramps up.

Meanwhile, Salesforce has rolled out pay-per-resolution pricing with its recently unveiled AI Help Agent and last month acquired usage-based billing specialist m3ter. Oracle is also piloting outcome-based AI pricing.

All these approaches undercut the predictability of the subscription model, which will complicate CIOs’ cost calculus, Deloitte’s Datwani says. “Now, as you start to think about consumption and tokens, costs might look different. As folks are opening up the architecture with things like headless CRM, what will the cost model look like for API calls or MCP server calls? So, there’s many more variables,” he adds.

The rise of multi-agent orchestration

To act autonomously, agents need to access multiple data sets and software platforms seamlessly. As a result, the proliferation of agents, some embedded within specific vendor platforms and some created in-house, is going to require an orchestration layer, Futurum’s Kirkpatrick says.

He points out that organizations need to monitor and manage agents, enforcing the same type of policy-based access control that exists for people. Organizations also need to set limits on what domains a specific agent can get into, what types of data they can access, what lines can’t they cross.

Kirkpatrick predicts that a new class of orchestration tools will emerge, although it’s not clear whether that orchestration layer will be provided by the leading CRM vendors, hyperscalers, or third parties.

Datwani agrees. “The orchestration layer is an interesting area, where the traditional vendors are in on it, the hyperscalers are also offering it, and there are third parties. It’s my belief that there’s not going to be a clear winner.”

The interface becomes conversational

Enterprise users who have traditionally had to manually wrangle with CRM systems are likely to find the ability to employ voice commands using a natural language interface to be a game changer. For starters, a salesperson can say, “I have a meeting today with Customer X. Help me prepare.” The agent will collect relevant data, ingest it, and provide a summary with recommendations.

ServiceNow’s Chesire says voice-enabled CRM systems have an “almost magical” ability to record, transcribe, and understand the content of a call between a salesperson and a customer or potential customer. The system can then “build a quote” based on that conversation.

On the customer service side of the equation, AI-driven voice technology enables customers to speak to an AI agent, describe the problem using natural language, and get a response. The agent has the capability to, for example, solve a credit card dispute, order a replacement product, send out a service rep, or do whatever is needed to resolve the issue, says Chesire.

Beyond that, agentic technology is capable of understanding the underlying business process flaws that led to the product snafu, and make recommendations for ways to fix whatever led to the issue in the first place, he adds.

Agentic drives business process transformation

With the emergence of outcome-based pricing, organizations are taking a fresh look at how they measure the benefits of CRM systems. That conversation is leading to an even more important analysis of underlying business processes. Or, as Constellation’s Miller says, “The old adage of applying new technology to old processes only gets you more expensive old processes.”

“When we survey customers, we hear time and time again that the reason why they want to apply AI into their organizations is to foster exponential opportunity and exponential growth,” she says. “How do we get there with CRM has started to become the new conversation.”

According to Miller, AI systems breach the walls of siloed data and can take a fresh look at legacy workflows. They also don’t get sucked into turf wars between marketing and sales teams. As a result, they often recommend new actions that can lead to better processes. “I think it’s starting to happen. You’re starting to see applications where AI is beginning to accelerate decision-making and decision velocity,” she says.

“The next phase of maturity is going to be, how do we start to spread AI across our platforms so that we are seeing that holistic end-to-end relationship that we have always wanted to optimize. How do we thread that across platforms and across solutions. We’re starting to see organizations on the leading edge really start to pull those strategies together,” says Miller.

For CIOs, learning from the startup ecosystem has never been more critical.

As pressure mounts to transform business operations with AI and agentic systems, IT leaders should be looking to those on the AI vanguard for insights into the strategic and technical decisions necessary to launch, grow, and thrive in today’s AI-disrupted business environment.

So why not immerse yourself in Silicon Valley’s most famous firehose of hyper-accelerated fail-fast and dream-big culture by registering for TechCrunch Disrupt 2026?

Three packed days of 200-plus sessions across six stages will spark new ideas for reshaping your AI strategy, provide fresh perspectives on the architectural, workflow, and resource decisions involved in moving AI from pilots to scale, and give you a sneak peek of business disruptions to come.

Get 10% off your TechCrunch Disrupt pass with the exclusive code CIO10. 

This year’s TechCrunch Disrupt, held Oct. 13-15 at San Francisco’s Moscone West, will feature big-picture conversations on what’s next in AI; discussions on how AI agents are rewriting SaaS, enterprise workflows, software pricing, and security; and demonstrations of AI’s future across robotics, manufacturing, defense, and industrial operations; and more.

Over 10,000 attendees will hear from 250-plus startup founders, technology executives, and enterprise IT leaders about how the future of programming is being rewritten, what enterprise AI security requires, how startups are orchestrating workloads across models while managing cost and reliability at scale, why creating a safety culture is essential for AI deployment, and how startups are deciding what work humans should own versus what should be delegated to AI as they work to build hybrid teams without losing speed, accountability, or culture.

And of course, the rising tide of enterprise-focused startups will be there seeking to bring agentic systems to your business workflows, as well as vendors familiar to your enterprise IT portfolios, such as AWS, Google, and Databricks, and enterprise IT colleagues creating mutually beneficial partnerships with the startup community, such as American Express.

That’s not to mention TechCrunch Disrupt’s signature Startup Battlefield, in which 200 standout companies showcase their innovations to compete for a $100K equity-free prize. The battlefield will give CIOs a rapid-fire, broad view of what’s possible — and a possible early look at the next big enterprise player. After all, Dropbox, Trello, and Cloudflare, among others, roamed that same battlefield before the world knew their names.

And with M&A now an early-stage startup strategy for many from day one, TechCrunch Disrupt’s exhibition floor provides IT leaders not just an opportunity to discuss the nuts and bolts of innovation architecture or how an upstart product can enhance your workflows, but a chance to find your next innovation partner, or more.

Leading-edge startups are figuring out how to make AI work at scale. Shouldn’t you be?

Don’t miss your chance to experience TechCrunch Disrupt 2026. Book your pass today and use the exclusive code CIO10 to save 10% before prices increase.

In August 2025, 126 malicious packages landed in the npm registry. Even after the community caught the initial wave, 80 of these hidden backdoors remained actively listed.

That was enough. Over 86,000 downloads. Malicious code in PhantomRaven, packages running in the production systems of Fortune 500 companies worldwide. And throughout the entire window, not a single EDR/XDR alert.

This happened because the attack surface has expanded to a layer EDR/XDR was never designed to see: VS Code extensions, local MCP servers, and rogue AI coding assistants that inherit your engineers’ valid credentials to steal data at machine speed.

To eliminate this structural vulnerability, Palo Alto Networks acquired Koi, an AI-native developer security product engineered for proactive, precision enforcement. Below we compiled a 2026 CISO checklist you can use to audit your environment and see how Koi automates each defense from day one.

#1. Gain real-time visibility into shadow AI & extensions

Your existing asset management tracks binaries and installers, but it cannot see local VS Code extensions, MCP servers, or ad-hoc Python scripts running on developer endpoints. This visibility gap was recently exposed by the MaliciousCorgi campaign, where two marketplace extensions with 1.5 million combined installs silently harvested every file a developer opened. Neither triggered any detection because they were not binaries, not executables, not anything your inventory was built to flag. To counter this, Koi closes the gap by analyzing what extensions actually do after installation, exposing hidden data-harvesting channels running inside your active workspace.

#2. Distinguish between human and autonomous agent behavior 

When a rogue AI agent exfiltrates your proprietary source code, it uses a developer’s valid credentials during normal working hours, making the session look entirely legitimate to standard XDR baselines. Moving beyond static permission lists, Koi deploys behavioral profiling within the workspace runtime. By actively intercepting unauthenticated background tasks and blocking unauthorized file-system reads, it stops automated data exfiltration in real time.

#3. Establish guardrails for automated package updates on endpoints

Developers prioritize speed, often allowing software packages to auto-update on their endpoints the moment a new version appears. Attackers weaponize this supply chain vulnerability, as seen in the May 2026 Team PCP attack where 3,800 GitHub repositories were compromised in just 36 minutes via poisoned auto-updates. Securing agentic endpoints against these rapid breaches requires behavior-based inspection within the active workspace context. Koi operates at this layer by providing safe deployment buffers that automate version cooldowns, blocking bleeding-edge updates until they are vetted. By continuously auditing process creation within the IDE runtime, Koi instantly drops unauthorized remote connections before malicious payloads can exfiltrate credentials from the endpoint.  

#4. Enforce principle of least privilege for AI agents

AI coding assistants inherit the privileges of whoever deployed them. In practice, that means read access to production databases, write access to core repositories, and access to every secret in environment files and configuration directories. To restrict this excessive access, Koi applies dynamic sandboxing directly to AI agent processes at the kernel level. It enforces a strict zero-trust boundary that segregates sensitive workspace vectors, preventing agents from pulling data outside their approved scope without interrupting developer workflows.

#5. Maintain continuous endpoint posture management

Signature-based scanning only stops known threats. Sophisticated repository attacks often arrive as functional, high-rated software that carries no known bad signature. Koi’s research into the DarkSpectre campaign found eight browser extensions, all carrying “featured” badges from Google and Microsoft, installed by over 8 million users, silently harvesting every conversation from ChatGPT, Claude, and Gemini in the background. Koi addresses this by operating upstream: scanning marketplace listings every hour, using LLM-driven code analysis to compare what software promises against what its code does, sandboxing it, and scoring the risk before it ever reaches the endpoint.

Summary

Securing the modern enterprise is no longer about patching individual gaps. As AI agents redefine the workforce, Agentic Endpoint Security (AES) is now a strategic imperative for every CISO. By establishing a mandatory control plane for the AI-native workspace, AES ensures that your organization can scale engineering velocity without ever compromising enterprise integrity. 

Ready to secure the future of your software stack? See how Koi Agentic Endpoint Security delivers complete visibility, risk scoring, and real-time prevention across every endpoint in your enterprise.

For CIOs, learning from the startup ecosystem has never been more critical.

As pressure mounts to transform business operations with AI and agentic systems, IT leaders should be looking to those on the AI vanguard for insights into the strategic and technical decisions necessary to launch, grow, and thrive in today’s AI-disrupted business environment.

So why not immerse yourself in Silicon Valley’s most famous firehose of hyper-accelerated fail-fast and dream-big culture by registering for TechCrunch Disrupt 2026?

Three packed days of 200-plus sessions across six stages will spark new ideas for reshaping your AI strategy, provide fresh perspectives on the architectural, workflow, and resource decisions involved in moving AI from pilots to scale, and give you a sneak peek of business disruptions to come.

Get 10% off your TechCrunch Disrupt pass with the exclusive code CIO10. 

This year’s TechCrunch Disrupt, held Oct. 13-15 at San Francisco’s Moscone West, will feature big-picture conversations on what’s next in AI; discussions on how AI agents are rewriting SaaS, enterprise workflows, software pricing, and security; and demonstrations of AI’s future across robotics, manufacturing, defense, and industrial operations; and more.

Over 10,000 attendees will hear from 250-plus startup founders, technology executives, and enterprise IT leaders about how the future of programming is being rewritten, what enterprise AI security requires, how startups are orchestrating workloads across models while managing cost and reliability at scale, why creating a safety culture is essential for AI deployment, and how startups are deciding what work humans should own versus what should be delegated to AI as they work to build hybrid teams without losing speed, accountability, or culture.

And of course, the rising tide of enterprise-focused startups will be there seeking to bring agentic systems to your business workflows, as well as vendors familiar to your enterprise IT portfolios, such as AWS, Google, and Databricks, and enterprise IT colleagues creating mutually beneficial partnerships with the startup community, such as American Express.

That’s not to mention TechCrunch Disrupt’s signature Startup Battlefield, in which 200 standout companies showcase their innovations to compete for a $100K equity-free prize. The battlefield will give CIOs a rapid-fire, broad view of what’s possible — and a possible early look at the next big enterprise player. After all, Dropbox, Trello, and Cloudflare, among others, roamed that same battlefield before the world knew their names.

And with M&A now an early-stage startup strategy for many from day one, TechCrunch Disrupt’s exhibition floor provides IT leaders not just an opportunity to discuss the nuts and bolts of innovation architecture or how an upstart product can enhance your workflows, but a chance to find your next innovation partner, or more.

Leading-edge startups are figuring out how to make AI work at scale. Shouldn’t you be?

Don’t miss your chance to experience TechCrunch Disrupt 2026. Book your pass today and use the exclusive code CIO10 to save 10% before prices increase.

Cisco and AMD have expanded their partnership with a new package of hardware and security software that’s designed to help enterprise customers protect, deploy, and manage distributed AI resources.

During AMD’s Advancing AI event this week, Cisco’s president and chief product officer Jeetu Patel took to the stage during AMD CEO Lisa Su’s keynote to talk about how AI inference will be widely distributed and will require an architectural stack of software and tools that Cisco and AMD are partnering to develop.

The joint architecture combines AMD’s compact, high-performance Ryzen AI Halo hardware and a variety of Cisco networking, observability, governance, and security technologies. “AMD provides the deskside/local AI platform. At the foundation is AMD Ryzen AI Halo hardware, an isolated agent sandbox and the services needed for local-first inferencing, including model routing and token limits via AMD’s Semantic Router and local inference on Lemonade,” wrote Cisco’s Yash Sheth, senior director, engineering and research, in a blog post about the new package.

Ryzen AI Halo (pictured below) is designed to support local AI inference on an AI PC using its CPU, GPU, and XDNA neural processing unit (NPU), according to AMD. A resilient AI platform should continue delivering useful AI services even when connectivity is limited, models need to change, or workloads shift, AMD stated.

Cisco then wraps that platform in a secure harness that includes its Splunk Agent Observability plus Splunk Infrastructure Monitoring to provide full-stack observability, tracking agent behavior, tokenomics and compute operation, according to Sheth.

Cisco also brings its AI Defense for model and agent security; DefenseClaw for security policy enforcement, so guardrails are enforced directly on-device, within the agent harness; and Cisco Cloud Control offering a single pane of glass for unified policy and control, Sheth stated.

“To make deskside and local AI computing work at enterprise scale, every AI node must be treated as a secure, managed node in the enterprise network,” Sheth wrote.

“The need for token efficiency and data sovereignty is driving a new class of computing, deskside computing, with users and teams putting AI agents right by their sides,” Sheth wrote. “Inference is moving to a hybrid architecture with thousands of ambient deskside agents in an enterprise helping employees have 24×7 productivity. That’s an extraordinary opportunity. It’s also a brand-new operating challenge.”

As agentic AI moves from experimentation to real enterprise workflows, organizations need more than powerful endpoints. AI agents can run continuously and act on enterprise data, but create new requirements for network infrastructure, tokenomics, agent behavior, and security, according to a statement from AMD.

“Running more AI locally can help improve responsiveness, keep sensitive data closer to users, and reduce dependence on cloud-only approaches, but enterprises also need a way to monitor and manage these systems at scale. AMD and Cisco are addressing that gap by collaborating to pair high-performance local AI compute with the observability, governance, and control infrastructure needed for enterprises to deploy it responsibly,” AMD stated.

“By combining AMD Ryzen AI Halo systems and our broader local AI software capabilities with Cisco’s enterprise networking, observability and security technologies, we are helping customers deploy AI in a way that is performant, secure, observable and manageable at scale,” said Jack Huynh, senior vice president and general manager, computing and graphics group with AMD, in a statement.

A few other interesting statistics and trends cited in AMD CEO Su’s keynote include:

• AI adoption is accelerating across all industries, with agentic AI driving a surge in compute demand and shifting workloads from training to inference, which accounts for 60% of global AI compute capacity in 2026.
• AI is moving beyond the cloud, with edge and personal devices becoming critical for real-time, distributed intelligence.
• The AI accelerator market is projected to reach $1.4 trillion by 2030, nearly tripling previous forecasts, with GPUs expected to dominate but CPUs gaining new growth vectors due to agentic AI.
• Server CPU market is forecasted to grow over 50% to $200 billion by 2030, fueled by rapid agentic AI adoption and the need for massive CPU infrastructure.

Model Context Protocol (MCP), the emerging standard for connecting AI models to external tools and enterprise data, is undergoing its biggest architectural overhaul yet.

The latest release candidate, scheduled for release on July 28, removes protocol-level sessions in favor of a stateless architecture, a change which industry experts say is intended to make MCP easier to deploy across standard cloud infrastructure as enterprises move AI pilots into production.

“The session-based model made sense when MCP servers were local processes on a developer’s laptop. In production, it became an operational tax,” said Muskan Bandta, cloud associate at ZopDev.

“When your infrastructure team asks whether MCP services can scale like other cloud applications, the answer used to be ‘not quite.’ With the move to a stateless architecture, the answer is now yes,” Bandta added.

Earlier versions of the protocol maintained information about every client connection, meaning servers had to keep track of each session throughout an interaction. While that approach worked well for local development, it complicated deployments across multiple servers because requests often had to be routed back to the same machine, limiting scalability and making MCP a less natural fit for modern cloud architectures.

“Under the new stateless design, every request contains the information needed for any available server to process it independently. Applications that need to maintain context across multiple requests can still do so, but developers must now manage that state explicitly rather than relying on the protocol itself,” she said.

This transition to a stateless design goes beyond simplifying infrastructure by fundamentally changing how AI applications manage and share context across tools, according to Amit Jena, AI development manager at IT consulting firm Kanerika.

Instead of keeping application state hidden inside protocol sessions, the new design makes it explicit, allowing AI models to access, reason over, and pass that information between tools, giving developers greater control over how context is preserved and shared across tools, Jena said.

It should also make AI workflows more portable, resilient, and easier to orchestrate across distributed environments, he said.

MCP’s new features

Other changes to MCP include the addition of a Multi Round-Trip Requests (MRTR) mechanism that changes how AI agents request additional information they need to complete a task.

Instead of relying on a persistent connection between the client and server throughout the interaction, the new mechanism lets the server request additional input through a standard request-response exchange before continuing the task, Jena said.

Routable transport headers, another addition, enable API gateways and other networking infrastructure to identify and route MCP requests without inspecting their contents.

They reduce processing overhead, lower latency, and let enterprise teams enforce routing, rate-limiting and security policies more efficiently using existing API management infrastructure, Jena said.

MCP is also getting an updated authorization framework built around OAuth 2.1 and OpenID Connect; interactive MCP Apps; and deterministic caching of tool and resource listings to improve LLM prompt-cache hit rates, potentially saving on token costs.

Rebuilding the trust boundary

The MCP release steering committee also decided to deprecate some legacy features, including Roots, Sampling, Logging, the older HTTP+SSE transport and Dynamic Client Registration, although these will continue to work in this version and any other released over the next year.

The deprecation of Sampling is likely to have the biggest impact because it changes who is responsible for interacting with foundation models, said Jena.

“Sampling let MCP servers invoke the LLM through the client, which meant the server had a callback path into the model without owning that connection. Deprecating it means rebuilding that trust boundary,” Jena said. “Your server now calls the model provider directly. That changes your network architecture, your auth model, and depending on how you’ve built cost attribution, your billing flow.”

The year-long transition period will be enough for teams to audit their sampling dependencies now, said Jena: “The risk is that teams who haven’t implemented sampling themselves won’t know if a third-party MCP server they’re depending on uses it.”

Updated MCP SDKs

To accompany the protocol update, there are updated MCP SDKs for Python, Typescript, Go, and C#. These support both the old and new protocol versions, so new clients can continue communicating with older servers, while updated servers will also support older clients, reducing the risk of immediate disruptions.

That backward compatibility should make the transition largely incremental, except for enterprises that built custom infrastructure around MCP’s earlier session-based architecture, Bandta said.

Identifying and auditing those session dependencies may not be easy, Jena warned.

“Session management complexity tends to be hidden across multiple layers — the gateway config, the deployment scripts, the monitoring dashboards. The code change is small; finding everywhere the assumption lives is what takes time,” he said.

Model Context Protocol (MCP), the emerging standard for connecting AI models to external tools and enterprise data, is undergoing its biggest architectural overhaul yet.

The latest release candidate, scheduled for release on July 28, removes protocol-level sessions in favor of a stateless architecture, a change which industry experts say is intended to make MCP easier to deploy across standard cloud infrastructure as enterprises move AI pilots into production.

“The session-based model made sense when MCP servers were local processes on a developer’s laptop. In production, it became an operational tax,” said Muskan Bandta, cloud associate at ZopDev.

“When your infrastructure team asks whether MCP services can scale like other cloud applications, the answer used to be ‘not quite.’ With the move to a stateless architecture, the answer is now yes,” Bandta added.

Earlier versions of the protocol maintained information about every client connection, meaning servers had to keep track of each session throughout an interaction. While that approach worked well for local development, it complicated deployments across multiple servers because requests often had to be routed back to the same machine, limiting scalability and making MCP a less natural fit for modern cloud architectures.

“Under the new stateless design, every request contains the information needed for any available server to process it independently. Applications that need to maintain context across multiple requests can still do so, but developers must now manage that state explicitly rather than relying on the protocol itself,” she said.

This transition to a stateless design goes beyond simplifying infrastructure by fundamentally changing how AI applications manage and share context across tools, according to Amit Jena, AI development manager at IT consulting firm Kanerika.

Instead of keeping application state hidden inside protocol sessions, the new design makes it explicit, allowing AI models to access, reason over, and pass that information between tools, giving developers greater control over how context is preserved and shared across tools, Jena said.

It should also make AI workflows more portable, resilient, and easier to orchestrate across distributed environments, he said.

MCP’s new features

Other changes to MCP include the addition of a Multi Round-Trip Requests (MRTR) mechanism that changes how AI agents request additional information they need to complete a task.

Instead of relying on a persistent connection between the client and server throughout the interaction, the new mechanism lets the server request additional input through a standard request-response exchange before continuing the task, Jena said.

Routable transport headers, another addition, enable API gateways and other networking infrastructure to identify and route MCP requests without inspecting their contents.

They reduce processing overhead, lower latency, and let enterprise teams enforce routing, rate-limiting and security policies more efficiently using existing API management infrastructure, Jena said.

MCP is also getting an updated authorization framework built around OAuth 2.1 and OpenID Connect; interactive MCP Apps; and deterministic caching of tool and resource listings to improve LLM prompt-cache hit rates, potentially saving on token costs.

Rebuilding the trust boundary

The MCP release steering committee also decided to deprecate some legacy features, including Roots, Sampling, Logging, the older HTTP+SSE transport and Dynamic Client Registration, although these will continue to work in this version and any other released over the next year.

The deprecation of Sampling is likely to have the biggest impact because it changes who is responsible for interacting with foundation models, said Jena.

“Sampling let MCP servers invoke the LLM through the client, which meant the server had a callback path into the model without owning that connection. Deprecating it means rebuilding that trust boundary,” Jena said. “Your server now calls the model provider directly. That changes your network architecture, your auth model, and depending on how you’ve built cost attribution, your billing flow.”

The year-long transition period will be enough for teams to audit their sampling dependencies now, said Jena: “The risk is that teams who haven’t implemented sampling themselves won’t know if a third-party MCP server they’re depending on uses it.”

Updated MCP SDKs

To accompany the protocol update, there are updated MCP SDKs for Python, Typescript, Go, and C#. These support both the old and new protocol versions, so new clients can continue communicating with older servers, while updated servers will also support older clients, reducing the risk of immediate disruptions.

That backward compatibility should make the transition largely incremental, except for enterprises that built custom infrastructure around MCP’s earlier session-based architecture, Bandta said.

Identifying and auditing those session dependencies may not be easy, Jena warned.

“Session management complexity tends to be hidden across multiple layers — the gateway config, the deployment scripts, the monitoring dashboards. The code change is small; finding everywhere the assumption lives is what takes time,” he said.

This article first appeared on InfoWorld.

Four months of Claude Code — that’s all it took for Uber to burn through its entire annual budget for AI. Token after token, engineers embraced the platform with few control mechanisms tying costs to outcomes. The result was a budget runaway and a clear case study in how limited oversight snowballs into an AI blowout.

This is a phenomenon I like to call “shadow tokens” — AI credits paid for by the company but largely invisible to decision-makers. Too many engineers have the final say over how much they consume and, therefore, what it costs. This all-you-can-eat attitude is part of the reason why Microsoft is reportedly winding down many internal licenses across key engineering teams and why one in five organizations is missing its AI spend forecast by more than 50%.

And the trend is only accelerating. By 2028, Gartner predicts that AI coding costs (driven by this kind of ungoverned consumption) will be as much per developer as the salary companies pay that person.

LLMs and agents introduce a new class of variable cost that scales with behavior rather than headcount, putting enterprises on the hook for tools that balloon with workload. I don’t see this as enterprises overspending because they’re reckless — it’s down to a lack of managerial oversight, budget alignment that demands a proven return on investment, and engineer education on how much is too much.

Going forward, CIOs need to thread the AI needle between governance that encourages transparency and reasonable spend without stifling innovation.

When shadow tokens result in real costs

The issue is that AI isn’t a traditional line item. Previously, enterprise leaders onboarded software-as-a-service (SaaS) with a good idea of the total cost. An allocated software seat or annual contract was a known quantity. The cloud added some variation (with fluctuations depending on hosting size), but instances were still modelable. AI flips this status quo on its head — the unit of consumption is behavior and the cost is exponential.

And these specifics aren’t immediately apparent at pilot. Tools can appear inexpensive in controlled experiments yet unpredictably scale depending on session length, context window size, model selection and whether agents run in parallel. This is the fallacy of the $20-per-seat enterprise plan — tokens are charged separately at API rates with no ceiling. The final dollar value of any session is set by factors that finance can’t always model in advance, particularly when these decisions usually rest with the engineers themselves.

According to Deloitte, only 21% of organizations deploying agents have a mature governance model, a real concern because they’re token-eating machines. This is what was happening at Uber — Claude Code in agentic mode was autonomously reading codebases, planning changes across dozens of files and opening pull requests. Each step quickly adds up, with Anthropic’s own documentation noting that agents consume approximately seven times as many tokens as standard sessions.

This is shadow IT and shadow AI, evolved. This time, however, many leaders approved the tool in question without guardrails governing consumption. AI hype adds fuel to the fire and normalizes long sessions. Uber’s CTO, for example, described a company-wide shift toward “agentic software engineering” with employees “who are quietly experimenting, quietly shipping and quietly pushing things forward”. This is an exciting way to test the limits of what’s possible, certainly, but it’s also a position that goes a long way to explaining how the company spent its annual AI budget by April.

Shifting the culture from usage to yield

Engineers haven’t done anything wrong here. In fact, they’re adopting and experimenting as instructed, with Uber creating leaderboards and ranking users by token consumption. More use led to a better ranking, reflecting a culture that lauds new ways of doing things. This behavior is known as “tokenmaxxing,” and its principal knock-on effect is shadow tokens — quantity-over-quality processes that leaders struggle to control until they’re fully realized in the budget. Of course, if management treats adoption metrics as performance metrics, then engineers can’t be blamed for using more tokens. The tension is that the teams driving adoption aren’t the ones managing spend.

None of this is meant to dismiss AI’s productivity possibilities and potential return on investment. Developers save 3.6 hours per week, achieve 60% higher pull request throughput and cut onboarding time in half with automation. Meanwhile, Uber shared that roughly 11% of live backend updates were written by agents with no human in the loop. However, these wins aren’t the problem — it’s that too many teams aren’t connecting input to output. I’ve spoken to admins who discovered their token spend had tripled in a single quarter after using heavier models or accidentally doubling up on agentic applications. Nobody knew until the financial damage was done.

Automation needs to happen sustainably with an eye on the bottom line. In my view, a much better metric for achieving this is AI yield — the measurable business or engineering output generated per dollar spent on tokens. Otherwise, without a feedback loop, even genuinely productive teams are flying blind.

Stopping token waste before an AI blowout

Creating that throughline between AI investment and token consumption starts with established financial metrics. This is possible via maximum spend limits (dictated by spend tagging, workload tiering and cost-per-output benchmarks) per team or project. Then, any additional allocation requires approval, closing the loop between the engineers spending the tokens and the leaders paying for them. AI isn’t cheap and teams should demonstrate a bang for their buck.

This is something we do with our engineering team at Hexnode. Resource allocation for Claude Code and Cursor is tied directly to ROI rather than letting consumption run open-ended. Given the pay-as-you-go nature of these tools, a firm usage limit per team offers simple but essential control.

Similarly, there’s room to apply some of the governance principles IT uses for device management. Things like policy enforcement, role-based access, real-time monitoring and automated alerts can flag usage behavior in advance. Uncovering such insights at the token layer works to identify power users and prevent excessive spending.

We also need to encourage cultures that praise outputs that actually achieve efficiency. AI applications that result in shipping faster, reducing rework and cutting review cycles are gains that should be celebrated. If your company hosts leaderboards, frame unnecessary token burn as wasteful rather than valuable. The organizations creating healthier consumption habits work with their engineers to understand not just how to use AI, but what responsible use looks like and what it costs.

This is a conversation teams need to have now. Anthropic just ended flat-rate pricing for programmatic workloads from June 15. Now, agents, continuous integration pipelines and automated workflows draw from a dedicated monthly credit pool billed separately from the subscription. Once that pool is exhausted, agent tasks either stop entirely or overflow to extra billing. Work can either get very expensive or grind to a halt for teams that aren’t prepared.

Getting a grip on shadow tokens means better rules and tools connecting spend to outcomes. Only by building the financial and cultural infrastructure that encourages sustainable adoption can leaders see what they’re spending, connect it to what they’re getting and course-correct before the costs become a crisis. Ultimately, shadow tokens are only invisible if we choose not to look.

This article is published as part of the Foundry Expert Contributor Network.
Want to join?

I have worked alongside enterprise technology for more than 30 years and watched AI evolve from a lab experiment into the modern boardroom’s core focus. However, the last few years of implementing AI alongside our customers have delivered our most profound reality checks.

The initial hype has faded, leaving CIOs to drive real enterprise value. Based on my experience implementing Google, OpenAI and Anthropic technologies, here are the fundamental, technology-agnostic lessons every leader must anchor their strategy around.

AI as a leadership multiplier

The most common tactical error we see is treating AI as an isolated technology project. What I have observed among our customers is that true success does not come from organizations that define a standalone “AI strategy,” but rather from those leaders that integrate AI into their business strategy.

When our customers isolate AI and define an AI strategy, it inevitably treats it like a “technological toy” to experiment with. This approach yields fragmented, orphaned initiatives that fail to scale because they are fundamentally disconnected from their core corporate objectives. What I learned is that AI is not the ultimate destination; it is a powerful catalyst. We have replaced “What can AI do for our customers?” with a more strategic question, “How does AI accelerate their existing business goals?”

Think of AI like electricity. No modern corporation designs a standalone “electricity strategy.” Instead, all companies route it invisibly across the entire organization to illuminate offices, power production lines and drive communication. AI must be woven into the enterprise fabric in the exact same way, acting as an underlying utility that supercharges your existing operational model.

Integrating AI into the broader business strategy also dictates how we measure success. It forces a shift away from short-term tech vanity metrics and anchors the technology into a long-term roadmap.

When AI remains trapped within the IT department of our customers, we notice that it is relegated to a mere “software experiment.” To become a true competitive advantage, we observed that AI requires intense cross-functional orchestration. This perspective does not diminish the merit of the technical team; their expertise is fundamental for establishing the architecture, data governance and tools your enterprise requires. However, while IT builds the foundational infrastructure, it lacks the organizational authority to decide what should be built on top of it. Only the CEO or the owner of the company can step in to ensure AI leaves the “toy project” phase and integrates into the DNA of the organization.

The requirement for top-down, executive ownership stems from three critical realities observed in the field:

• Silo-smashing and data collaboration: True enterprise AI is data-hungry and that data lives across disparate business lines, finance, operations, marketing and customer service. Only the CEO possesses the cross-functional authority to demand that data silos be dismantled.
• Cultural transformation and fear mitigation: AI triggers widespread anxiety over job displacement across all industries and hierarchies. When relegated to an “IT project,” resistance spikes as teams view it as a threat to their livelihoods. When I saw the CEO lead this cultural shift directly is when I noticed the best results.
• C-Suite education and strategic alignment: The mandate for AI capability cannot just be delegated downward; the transformation must begin at the very top. I have conducted more than 70 presentations for the Board of Directors and C-Level teams. These people need to be actively educated not on technical code, but on specific business use cases, return on investment (ROI) frameworks and how AI resolves core organizational bottlenecks.

PwC’s data found that only 12% of enterprises have achieved both cost and revenue benefits from AI. Those elite 12% succeeded precisely because their CEOs embedded AI extensively across strategic decision-making and cross-functional workflows. AI is simply too disruptive and too critical to be left exclusively in the hands of technical experts. If AI is not on the CEO’s weekly agenda, it is fundamentally missing from the company’s true strategy.

AI as a new operational framework

Traditional IT systems have operated on strict algorithmic certainty: if you input a specific set of data, the system executes an immutable line of code and guarantees the same, predictable output every single time.

AI completely breaks this paradigm. Because modern AI is built on probabilistic models, it does not execute static formulas; instead, it predicts the most likely correct response based on mathematical probabilities. This means that AI solutions carry an inherent, small percentage of uncertainty and variability. A prompt entered today might yield a slightly different, though contextually valid, output tomorrow.

Executive leadership and organizational cultures must be actively educated to accept and navigate this fundamental shift. Traditional quality assurance frameworks for software are designed for a 100% success rate. Applying this rigid standard to AI will paralyze your initiatives, keeping 80% of your projects trapped eternally in the pilot phase. This happened to us in a food and beverage company in Latin America a couple of years ago. After this experience, we started to include conditions in our contracts that tolerate statistical margins of error and still define the project as a success.

In terms of cost calculation, we had to teach CIOs and business managers to forget the monthly subscription model for AI and learn to manage the primary unit of exchange in modern AI: the token.

To understand AI costs, executives must understand how large language models process data. AI models do not read full words; instead, they break text, images or code down into “pieces” called tokens. As a baseline, every 100 words process as approximately 130 to 140 tokens. Because the major AI providers use the token as their currency, your business is billed dynamically based on the exact volume of tokens consumed by every query submitted (input) and every response generated (output).

Many leaders believe AI costs are fixed due to flat-rate enterprise tiers ($25–$30/user). This is a temporary illusion. These venture-capital-subsidized rates mask true operational costs and come with dynamic usage limits. Modeling long-term ROI on them guarantees a severe budget shock when true consumption pricing takes over.

The solution is not to halt AI adoption; doing so means losing your competitive edge. Instead, the cost per token must cease to be treated as a technical footnote relegated to the IT department. It must be elevated to a core business variable.

Risks in the AI adoption model

Since the beginning of the AI boom, I have seen all our customers making a critical tactical error that could cost them heavily in the medium term: they are focusing only on operational efficiency (reducing costs with AI).

I have observed that an alarmingly high percentage of companies remain trapped in pilot phases focused exclusively on short-term cost reduction. Bain & Company’s global Automation and AI Pathfinder Survey found that the largest share of companies measuring their AI initiatives (exactly 40%) realized cost reductions of 10% or less, heavily missing their internal targets. Our customers are putting too many resources and effort into marginal financial gains and in doing so, they are jeopardizing their most valuable assets: service quality, resilience and customer trust.

Utilizing AI solely to slash headcount or cut operational corners is a dangerous trap that introduces severe field liabilities. A financial service organization in Latin America announced that they saved $1 million in customer support by replacing humans with AI chatbots. However, the mid-term reality revealed a different story: a damaged brand reputation due to AI errors and an influx of frustrated clients fleeing because the automated system cannot handle special cases.

Putting a company on an extreme AI diet might make it look leaner on next quarter’s financial statement, but over-indexing on cost-cutting will ultimately leave the business too weak to compete when market dynamics shift. We are now inviting our customers to change the question from “How much money will AI save us?” to “How will we leverage AI to exponentially increase the long-term value of our enterprise?”

Deploying enterprise AI is a marathon, not a sprint, and the terrain changes with every mile. The organizations that thrive in this next era will be those that transition from fascination to discipline, treating AI not as a magic bullet for immediate savings, but as a core capability that demands rigorous governance, architectural foresight and cultural maturity. Navigating this shift requires moving past the theoretical hype and anchoring decisions in raw, field-tested reality.

As we continue to deploy these technologies across industries, the blueprint for success is being rewritten in real time. Let’s keep this conversation going as we map out the future of business intelligence together.

This article is published as part of the Foundry Expert Contributor Network.
Want to join?

ISC2 has begun developing a vendor-neutral AI security certification aimed at cybersecurity professionals working to secure AI systems and manage emerging AI risks.

The nonprofit organization, best known for the CISSP certification, says it is seeking volunteers worldwide to help define the knowledge and skills that will shape the new credential. While ISC2 has not finalized the certification domains, the certification is expected to address both technical AI security and governance topics, with a pilot exam planned before the end of 2026.

According to Casey Marks, ISC2 chief operating officer, feedback from cybersecurity practitioners led ISC2 to conclude that AI security had grown beyond expanding AI content within existing certifications.

“AI has reached a tipping point,” Marks says. “AI no longer is just another tool; instead, it has fundamentally changed the cybersecurity practice itself.”

ISC2 already includes AI-related content in certifications including CISSP and CCSP, but Marks says practitioners have identified new responsibilities and risks that extend beyond those programs. “Enterprise security teams are currently grappling with significant knowledge gaps, particularly around securing model architectures against new vulnerabilities like prompt injection, data poisoning, and model inversion,” Marks adds.

Organizations are working to understand emerging governance frameworks, including the NIST AI Risk Management Framework and ISO/IEC 42001, while adapting traditional application security and security operations workflows to AI systems, he says.

ISC2 has not finalized the certification domains, but Marks says the organization expects the credential to address both technical controls and governance practices for securing AI systems and managing AI risk. The certification will use ISC2’s established certification development process, which relies on cybersecurity practitioners to define job roles, develop exam content, and validate competencies.

Marks says ISC2 will continue to update the certification through ongoing input from cybersecurity professionals, in addition to its regular certification review process.

The organization is also determining which professionals the certification will target. Marks says AI security responsibilities are emerging across security architecture, risk management, security operations, software development security, governance and compliance, communication and network security, and security assessment and testing. ISC2 says the certification will reflect how those roles are evolving.

For organizations that are building AI security programs now, Marks recommends using existing AI training resources, adopting established governance frameworks, creating cross-functional AI security working groups, and participating in the certification development process.

Marks says ISC2 expects AI knowledge to become part of most cybersecurity roles while a more specialized AI security discipline continues to develop. He says organizations will increasingly need professionals with foundational AI security knowledge, as well as specialists in areas such as adversarial machine learning, model architectures, and AI data pipelines.

Looking ahead, Marks says he expects AI security expertise to evolve into both a foundational skill for cybersecurity professionals and a specialized discipline of its own.

“At this time, we are seeing a hybrid evolution occurring in real time: AI security is simultaneously becoming a baseline expectation for all security roles, while also carving out a dedicated, highly specialized discipline,” Marks says.

Over the past year, I had the same conversation with almost every team starting an AI initiative. Should we build on Semantic Kernel, AutoGen or Foundry?

At first it felt like the most important architectural decision we’d make. Each framework had its own philosophy, each promised to be the foundation for enterprise AI, and picking the wrong one felt like an expensive mistake. I spent a lot of time helping teams weigh the trade-offs.

Looking back, I think we were asking the wrong question. I certainly was.

I watched teams spend months debating SDKs while the decisions that actually decided whether their applications survived production went unexamined. Some built elaborate orchestration layers for workflows that a few deterministic functions would have handled. Others avoided agent frameworks entirely and later found they’d designed themselves into a corner.

Then Microsoft settled it for us. It introduced the unified Agent Framework, quietly moved Semantic Kernel and AutoGen into maintenance mode, and the debate I’d spent months refereeing was suddenly over. Turns out the answer to “which of the three” was “none of the three, here’s a fourth.” The framework hit version 1.0 and general availability in April 2026, stable across .NET and Python.

What surprised me wasn’t the decision. It was how fast a debate that had eaten so much of our attention stopped mattering. Microsoft changed the menu.

It didn’t change the meal.

The framework was never the hard part

Framework selection dominated almost every early conversation I had about enterprise agents. Which SDK do we standardize on? Which orchestration model gives us the most flexibility? Which one is Microsoft actually betting on?

Fair questions. But after a year of watching these projects play out, I’ve slowly come around to a different view. Those weren’t the questions that decided anything.

The first question I ask now is much smaller. Does this thing actually need an agent?

It sounds obvious, and I still get it wrong sometimes. But it’s the mistake I see most. On one project, a team spent weeks designing a multi-agent workflow for a process that ran the same four steps every time: read a document, validate it, call an API, send a notification. The diagrams looked great. The system in production didn’t. A few well-tested functions would have been easier to build, easier to maintain and a lot easier to trust.

Part of this is just that “agent” has become the word everyone reaches for. Sometimes it’s the right call. Sometimes it’s a workflow we already knew how to build, wearing a newer label. An agent earns its complexity when it genuinely has to decide things you can’t predetermine, choosing between tools, adapting to what it finds, working out its own next step. If you already know every step, you have a workflow, and a workflow is usually the better engineering choice. The consolidation didn’t change that. It just made it easier to see.

What building production agents actually taught me

Once I stopped fixating on frameworks, the same three problems kept showing up. None of them had anything to do with the SDK.

Context beats model choice

Early on I spent a lot of time comparing models, the way you’d agonize over a restaurant menu and then order what you always order. Now I spend most of it thinking about context, which is far less fun and far more useful.

I’ve watched good models fail because they were handed too much, not too little. One team I worked with gave the model access to nearly every internal document they had on the theory that more information meant better answers. It went the other way. Responses got slower, less consistent and sometimes skipped right past the thing that actually mattered. When we cut the context down to only what the task needed, the quality jumped almost immediately. I didn’t predict that. It taught me to be suspicious of “just give it everything.”

The best agent systems I’ve worked on weren’t the ones with the biggest context windows. They were the ones careful about what reached the model, and when. That’s not something the framework hands you.

Failure is where the real work is

Most agent demos look great because they’re built around the happy path. Production doesn’t extend that courtesy.

I remember a project where everything held up in testing. Then a downstream API timed out after the agent had already completed several earlier steps. We couldn’t just restart, because part of the business process had already gone through. We ended up spending far more time on recovery logic than we ever spent on prompts. That project changed how I think about this work. The hard part was never getting the model to make a decision. It was making sure the system didn’t fall apart when reality refused to follow the script.

Tool calls fail partway through. APIs return inconsistent data. Models call the same tool over and over because the last answer wasn’t what they wanted. That’s not the exception; that’s a normal Tuesday. Whether you retry, roll back, pause for a human or push on with partial results is a judgment call, and no framework is going to make it for you.

Identity is the real security boundary

This one surprised me most. The moment an agent stops being a chatbot and starts touching real business systems, identity matters more than orchestration.

Every project gets to the same question eventually. Who is this agent actually acting as? The developer’s credentials? A service account? The user who asked? Get it wrong and you’ve built something autonomous running with more access than any single person should have, which is exactly the kind of thing that looks fine until an audit. The Agent Framework, like most modern tooling, makes it easier to wire agents to tools through standards like the Model Context Protocol. That helps. But where human approval belongs, what needs extra authorization, how much rope to give the thing, those are still yours to decide.

The surprises weren’t technical

Here’s what I didn’t see coming. The hardest part of last year wasn’t technical at all. It was organizational. The moment a team heard “agent,” expectations shifted under everyone’s feet. Business stakeholders started expecting full autonomy. Developers assumed the thing could reason its way through anything. People started designing for flexibility before we’d even agreed on what problem we were solving. The word did damage before any code did. I found myself spending as much time resetting expectations as I did discussing architecture.

Build for change, not for today’s winner

I don’t think the teams that struggled last year picked the wrong framework. Semantic Kernel was reasonable. AutoGen was reasonable. Foundry made sense for plenty of cases. I’d have signed off on any of them.

The ones that got hurt put all their eggs in one framework, treating it as the foundation of the whole system instead of as one more dependency. Microsoft provided a migration path. But teams that had tightly coupled their applications to framework-specific abstractions discovered that migrating and rewriting are not the same thing. That wasn’t Microsoft’s doing. It was their own architecture’s. The teams that moved easily had kept their business logic, prompts and orchestration loose enough to evolve independently of any one SDK. For them, the change was a manageable project, not a teardown.

For what it’s worth, nobody I work with is treating this as an emergency. Most are moving the smaller workloads first, watching how they behave and leaving the production-critical systems alone until they actually understand the new abstractions. That’s the right instinct. And I doubt this is the last consolidation we’ll see, the ecosystem is still young, frameworks will keep absorbing each other and over time the differences between them will be operational more than architectural.

I don’t regret the framework debates, honestly. They were reasonable at the time. What changed wasn’t Microsoft’s roadmap.

It was mine. Watching these systems run in production taught me that the framework is the easiest piece to swap out. Recovery logic, context management, security boundaries, the business workflow itself, those stay with you long after today’s SDK gets replaced by tomorrow’s.

So, Microsoft made one decision easier by turning three frameworks into one. Good. Five years from now we’ll be on different tools, and we’ll still be asking the same handful of questions.

Does this actually need an agent? Does it have the right context? Can it recover when something breaks, because something will? Is it acting as the right person?

Those questions outlast every rewrite. That’s where I’ve learned to put my effort.

Frameworks come and go. Good architecture has to survive all of them.

This article is published as part of the Foundry Expert Contributor Network.
Want to join?

The cybersecurity landscape has evolved beyond human scale. Today’s adversaries have replaced predictable, manual playbooks with machine-generated attack chains that can breach traditional controls in seconds. To bridge the gap, organizations must move past legacy, reactive controls and embrace a fundamentally different, AI-driven architecture: Agentic Endpoint Security (AES). 

AES represents a paradigm shift, moving security from a passive monitor to an active participant in the defense lifecycle. It provides the visibility and automated guardrails necessary to govern autonomous AI agents and agentic tools, ensuring that as your workforce scales with AI, your security posture remains unbreakable. 

With autonomous AI agents now capable of planning and executing multi-stage attacks at machine speed, the pressure on traditional security operations (SOC) has reached a breaking point. To survive this shift, the strategy is clear: we must fight AI with AI. 

Here is how AI-driven defense, pioneered by Cortex XDR and the era of Agentic Endpoint Security, is fundamentally rewriting the cybersecurity playbook.

1. From reactive patching to proactive prevention 

For decades, the industry lived in a “wait-and-see” mode waiting for a vulnerability to surface, waiting for a signature, and then rushing to patch the hole. But reactive methods just don’t hold up against modern “frontier” AI attacks that are constantly morphing. 

AI-driven defense changes the game by shifting to a prevention-first architecture. Rather than relying on historical signatures, modern platforms deploy localized, ML-driven analysis to evaluate the intent and behavior of an active process, stopping threats pre-execution. Cortex XDR leads with a strict prevention-first approach by using AI-driven local analysis and behavioral threat protection; the XDR agent stops sophisticated threats pre-impact and pre-execution. This proactive stance reduces the overall risk profile by blocking malicious chains of events in real time across network, process, file, and registry activity. 

2. Eliminating the “agentic blind spot” 

As we all rush to adopt generative AI and automated workflows, a new gap has appeared: the “agentic blind spot.” Adversaries are now targeting AI assistants and automated scripts to bypass defenses. Since these digital agents often have deep access to enterprise data, a compromise here lets attackers move completely under the radar. 

The new playbook requires securing this entire ecosystem. By combining the distinct capabilities of Cortex XDR and Koi Security, organizations can effectively close this gap. Koi Agentic Endpoint Security tracks everything from shell commands to prompts in real time, while Cortex XDR adds a layer of defense that identifies and neutralizes behavioral anomalies unique to these automated threats. 

3. Machine-speed detection and “attack storylines” 

When an attacker can move through your network in seconds, human-led teams can’t keep up. To make matters worse, most systems just flood analysts with low-quality, isolated alerts, leading to major burnout. 

AI-driven defense fixes the investigation process by automatically stitching separate data points into a single, high-fidelity “attack storyline.” Cortex XDR uses thousands of machine learning detectors across endpoint, network, and cloud sources to group related signals into one cohesive case. This reveals the full story of an attack, letting your analysts focus on fast remediation instead of digging through piles of data, reducing alert noise by up to 98%. 

4. Surgical and autonomous response 

The final piece of the puzzle is moving from manual remediation to autonomous action. AI-driven response lets your SOC handle threats in minutes, not hours. The platform can automatically revoke compromised tokens or isolate endpoints at machine speed. 

Cortex XDR delivers built-in enterprise-grade automation at no additional cost, providing over 120 out-of-the-box playbooks and 18 quick actions to handle up to 99% of incidents without manual intervention. Crucially, this level of automation requires an unbreakable foundation of agent resilience. To ensure the defense cannot be disabled by an adversary, Cortex XDR is certified in both the AVC EDR Detection and Anti-Tampering tests, successfully blocking all attempts to disable or modify the agent. 

Summary

The threat landscape is changing faster than ever, driven by AI-powered attackers who exploit even the smallest gaps. But you don’t have to stay on the defensive. By shifting to a proactive, AI-driven architecture like the one built into Cortex XDR, you can stop threats before they happen, secure your agentic workflows, and automate away the noise that leads to analyst burnout. 

The journey to a more resilient, AI-powered SOC doesn’t have to be daunting. With the right foundation in place, you’re not just keeping pace with the new threat landscape; you’re staying one step ahead. It’s time to move beyond the old manual playbook and embrace the future of security operations. 

To learn more about Palto Alto Networks, visit https://www.paloaltonetworks.com.

Networking is core to any operating system, and when it comes to Linux, it’s actually a combination of several key components. The Linux kernel handles the data plane, moving packets, and holding live device state. NetworkManager is the network configuration service, operating as the control plane, deciding what a device’s configuration should be.

NetworkManager 1.58 was released this week, following more than five months of development and 407 commits since version 1.56. The release covers three areas: expanded support for IPv6-only networks, a set of Wi-Fi management updates, and a round of security hardening.

IPv4 address exhaustion remains the pressure behind the first of those areas, pushing more networks toward IPv6-only operation every year.

“More networks, mobile carriers, cloud providers, and anyone squeezed by IPv4 exhaustion are running IPv6-only by default,” Chris Van Hoof, director of Linux engineering, platform enablement at Red Hat, told Network World.

Advancing IPv6-only support

Dual stack networking, running IPv4 and IPv6 in parallel, has been the default IPv6 transition strategy for years. Dual stack networking, however, has a structural problem in that it still requires an IPv4 address on every device, so it does nothing to relieve address exhaustion pressure.

An alternative model called IPv6-mostly addresses that gap. It is defined in RFC 8925, “IPv6-Only-Preferred Option for DHCPv4,” and lets capable clients drop IPv4 entirely while legacy hosts that still need it keep receiving it on the same network segment.

“NetworkManager can also now auto-signal RFC 8925’s IPv6-only-preferred option, telling the network a host is fine skipping an IPv4 lease entirely,” Van Hoof said.

For the traffic that still needs IPv4, NetworkManager 1.58 adds support for CLAT, short for customer-side translator. CLAT is the client-side half of 464XLAT, a mechanism defined in RFC 6877, “464XLAT: Combination of Stateful and Stateless Translation.”

464XLAT pairs CLAT on the endpoint, which performs stateless header translation, with a stateful NAT64 translator on the provider side, letting IPv4-only apps keep functioning on a network that has no IPv4 of its own.

“CLAT is the translation layer that lets legacy IPv4-only apps and services keep working on those networks without bolt-on middleware,” Van Hoof said.

Wi-Fi management updates

NetworkManager 1.58 also brings a set of changes to how the daemon handles Wi-Fi connections and configuration.

• Band selection: The band property of Wi-Fi connections now accepts a 6GHz value, and a Wi-Fi scan run through nmcli, NetworkManager’s command line tool, now shows each access point’s band as well.
• Credential handling: WPS credentials with a 64 character hex PSK are now accepted, matching what some access points return.
• Text interface improvements: nmtui, NetworkManager’s menu driven text interface, picked up several usability additions. A new device select button lets you choose a physical interface from a list instead of typing its name. The activation screen gained a rescan Wi-Fi button, and secret prompts now include a show password checkbox. There is also a share QR code option, mirroring the existing nmcli device wifi show-password command.

Security hardening

The release fixes vulnerabilities and tightens several defaults tied to DHCP handling and connection permissions.

• CVE-2026-10805: Hostnames and MUD URLs are now validated before being written to the dhclient configuration file, rejecting characters that could alter the config syntax.
• DHCPv4 client fix: An out-of-bounds read in the internal DHCPv4 client, triggerable by an on-link attacker with a malformed UDP packet, has been fixed.
• Router option validation: The internal DHCPv4 client now ignores DHCP option 3, the Router option, when a lease also contains option 121, the Classless Static Route option, following the recommendation in RFC 3442.
• Permission checks and deprecations: For private connections that restrict access to specific users, NetworkManager now verifies that the user can access the referenced 802.1X certificates and keys.

Tunneling and automation updates

Two smaller but practical additions round out this release: a new tunnel type for virtualized networks, and a fix that closes a gap in how NetworkManager’s state survives a reboot.

NetworkManager 1.58 also adds support for creating and managing GENEVE tunnel interfaces. GENEVE, short for Generic Network Virtualization Encapsulation, is a tunneling protocol that wraps Ethernet frames inside UDP packets, letting virtualized or overlay networks run on top of physical Layer 3 infrastructure. It shows up mainly in virtualization and cloud environments, where a hypervisor or container networking layer needs to build a virtual network segment across physical hosts. Previously, NetworkManager could not create or manage these interfaces directly.

The release also adds persisted managed state. NetworkManager tracks whether it is responsible for a given network device, a setting called its managed state. Until now, that setting reset on every reboot, so provisioning tools had to reapply it each time a system restarted. NetworkManager 1.58 lets the managed state survive a reboot when it is set through nmcli or the D-Bus API.

“It’s a small change but closes a real automation gap: Provisioning tools and cloud-init style workflows can set a device’s state once via D-Bus or nmcli and trust it survives a reboot, instead of reapplying config every time,” Van Hoof said.

AMD executives took to the stage at its Advancing AI 2026 event in San Francisco today to detail the company’s next generation of AI infrastructure solutions, from Instinct MI455X AI accelerator GPUs and 6th Gen EPYC “Venice” CPUs, to Pensando networking, ROCm.AI software and its Helios rack-scale platform that ties it all together.

AMD has been working towards rack-scale AI system solutions for years. Its ZT Systems acquisition last year added valuable engineering talent and intellectual property that is now finally bearing the real fruits. Its Helios AI platform is a major platform evolution for AMD, with shipments scheduled to begin in the second half of this year (which is here and now).

The announcements at Advancing AI show how the company has engineered its AI platform solutions for large reasoning models, sustained inference and agentic workflows. These workloads pressure memory capacity, data movement, networking and CPU orchestration. AMD’s approach is to keep as much data close to the compute engines as possible and move it more efficiently throughout the system, but there’s deeper nuance here that’s obvious versus AMD’s chief rival, NVIDIA.  

AMD’s MI455X targets the AI memory wall

The Instinct MI455X GPU is the compute engine that fuels the Helios rack, and the first GPU based on AMD’s new CDNA 5 architecture. Built with a modular mix of 2nm and 3nm chiplets, it carries 432GB of HBM4 and 23.3TB/s of peak memory bandwidth.

Compared to AMD’s current MI355X, the MI455X offers 1.5 times the memory capacity, up to 2.9 times the peak memory bandwidth and up to four times the peak matrix performance with MXFP4 and MXFP8 data types, which are lower-precision numerical formats designed to accelerate AI processing while reducing memory demands. With MXFP6 (6-bit floating point), performance is rated at up to twice that of MI355X.

AMD also shared some actual, measured internal results using production silicon. The company claims MI455X delivers 3.8 times higher FP8 decode performance, 3.5 times more measured FP4 compute performance and between 2.5 and 3.5 times more networking bandwidth than MI355X, depending on the transfer path tested. Those figures provide more context than just numerical specifications, though they remain AMD-provided comparisons that will need independent validation.

AMD

The architectural choices behind the numbers are important. Reasoning models and long context windows require sizeable KV caches for maintaining AI attention states, while mixture-of-experts models frequently move large amounts of data across accelerators. MI455X should let more model data, activation states and cache remain local. New dedicated IP in hardware can transfer data while the GPU continues processing, and expanded cache and multicast capabilities are designed to reduce redundant data movement to further improve efficiency.

The aforementioned lower-precision formats can also raise throughput and reduce memory use, but model developers still have to determine where they can be applied without unacceptable accuracy loss.

AMD’s Helios rack takes aim at Vera Rubin

Dave Altavilla

Helios is AMD’s primary rack-scale competitor to NVIDIA’s Vera Rubin platform. Each liquid-cooled rack combines 72 MI455X GPUs, 18 single-socket Venice host CPUs and Pensando networking technologies.

In its most complete, premium configuration, AMD rates Helios for 2.9 exaflops of low-precision AI compute, with 31TB of aggregate HBM4 capacity, 1.7PB/s of memory bandwidth, 260TB/s of bidirectional scale-up bandwidth and 43TB/s of scale-out bandwidth.

These are formidable figures, but they are technical specifications rather than actual application benchmarks. The more consequential development is AMD’s move from collections of eight-GPU servers to a 72-GPU shared-memory domain. Models too large for one node can operate across the rack without treating every exchange as a scale-out networking transaction, which benefits large-model inference as well as training.

AMD uses UALink over Ethernet, or UALoE, for an open standard scale-up fabric. Each MI455X provides 3.6TB/s of bidirectional scale-up bandwidth, while the complete rack delivers all-to-all connectivity through a single switch layer. AMD also claims six times more scale-out bandwidth per GPU than MI355X when MI455X is configured with three Pensando Vulcano 800 AI NICs.

While open standards give cloud providers more control over suppliers and system design, AMD and its partners now have to prove those components can deliver the predictable performance, reliability and deployment experience customers expect from a tightly controlled, more vertically integrated platform.

Finally, AMD designed Helios with automatic rerouting around failed links, virtual rack partitions, tray-level serviceability and rack-wide power, cooling and health monitoring. Major hyperscalers and potentially large-scale enterprise customers will likely key in on these capabilities, which can affect the availability, total cost and consistency of the AI services they consume.

Kind of like cowbell, AMD Venice gives agentic AI more CPU

AMD

AMD’s agentic CPU messaging regarding its upcoming Venice-based EPYC processors is mostly marketing speak, but the underlying requirement is very real. An AI agent can invoke retrieval, databases, security checks, code execution and other tools before a GPU generates a response. Running many agents concurrently increases the amount of conventional compute requirements surrounding the accelerators.

Venice scales to 256 Zen 6 cores with support for 512 threads, 16 memory channels, up to 1GB of L3 cache per socket, along with PCIe 6.0 and CXL 3.1 connectivity. AMD is also offering several Venice configurations for other applications, including general-purpose servers, high-frequency workloads, GPU hosts and high-density CPU sandbox systems used to execute agent tools.

Treating the CPU solely as a GPU host understates its role. Gateways, tokenization, vector search, databases and short-lived code execution stress different mixes of per-core performance, thread count, memory bandwidth and I/O. Specifically, AMD’s internal testing shows Venice significantly outperforming its current EPYC 9965 Turin CPU across five parts of the agentic AI pipeline, including gateway processing, context assembly, vector search, enterprise applications and short-lived tool execution. Individual gains vary by workload, but AMD details the overall generational improvement at up to a 1.7 times lift. As with the MI455X figures though, these comparisons come from AMD and will require independent validation.

Pensando networking and ROCm software advance

Keeping GPUs fed with data and coordinating traffic across racks directly affects utilization and operating costs. In fact, GPU utilization is a pretty sad state of affairs currently for some of the major frontier model providers.

As such, Pensando networking has become central to AMD’s roadmap. Helios can connect each MI455X to as many as three 800Gbps Vulcano AI NICs, while Salina DPUs handle front-end networking and infrastructure services.

On the software side, which is an equally critical component, AMD also introduced ROCm.AI, an AI-assisted development layer due to arrive in August. It includes reusable skills for coding agents, simplified management and Hyperloom, which can profile workloads, tune serving configurations, modify kernels and validate results.

These tools address two persistent AMD challenges: developer efficiency and ease of use, and software tuning. Automated optimization still has to produce repeatable gains without creating hard-to-maintain code, however. And while ROCm has progressed significantly over the last few years, NVIDIA’s CUDA retains an advantage in maturity, tooling and developer familiarity.

Customer commitments underscore rack-scale confidence

AMD now has commitments that give its MI450 generation and Helios considerably more weight. Meta and OpenAI have announced multi-generation agreements composed of up to 6GW of AMD compute capacity, with initial 1GW deployments planned for the second half of 2026.

Oracle plans a 50,000-GPU public cloud cluster beginning in the third quarter, while Microsoft will deploy Helios for Azure AI inference. Finally, just before the AMD event, Anthropic announced a strategic partnership for up to 2 Gigawatts of AMD-fueled AI compute, with its first gigawatt expected online in the first half of 2027.

Commitments of this scale reflect confidence in more than just MI455X performance. These customers are evaluating the complete architecture, including Venice CPUs, Pensando networking, ROCm software, rack integration, serviceability and AMD’s ability to deliver and execute across multiple product generations.

There is some financial alignment behind the agreements as well. AMD issued OpenAI performance-based warrants and committed to investing up to $5 billion in Anthropic. That context matters when evaluating these deals as market validation, but these planned deployments are substantial nonetheless and put Helios on a much stronger foundation as it begins shipping.

AMD expands its robotics and embedded foundation

AMD also expanded its physical AI portfolio, building on credible traction from its Xilinx-derived Kria adaptive system-on-modules and embedded technologies that are already powering robotics, machine vision and industrial automation applications.

The new Ryzen AI Embedded X100 combines up to 16 Zen 5 CPU cores, integrated Radeon graphics, a second-generation NPU and as much as 128GB of unified LPDDR5X memory shared across its compute engines. To me this looks a lot like a repackaging and optimization of the company’s Strix Halo platform, but with specific optimizations for the embedded space. Regardless, AMD is pairing X100 with the Kria AI Robotics Developer Platform, which includes a System Module or SOM, and a new Robotics Partner Network spanning hardware, software and platform providers.

Samples began shipping in June, with full production expected in the fourth quarter. This broader objective is to give developers a path across AMD x86 CPUs, GPUs, NPUs and FPGAs for real-time autonomous systems, rather than requiring them to assemble those hardware engines and software components independently.

Execution for AMD is now the test

AMD has assembled a credible platform for the burgeoning agentic AI market that’s blowing up currently with no signs of stopping. MI455X addresses memory and data movement, Venice handles dense agentic CPU workloads, Pensando networking connects global system resources, and ROCm.AI addresses software complexity. Finally, Helios assembles these components into a true competitive threat for NVIDIA’s latest Vera Rubin platform.

AMD’s open architecture may appeal to customers seeking supplier choice, but openness must also translate into reliable deployments, competitive total cost and software that does not require a significant rip-up. NVIDIA enters this cycle with a stronger ecosystem and far more rack-scale deployment experience. The true test will be how easily and reliably customers can integrate, operate and maintain these AMD solutions at scale.

As it stands, AMD now has major customers and a clearly defined architecture with systems engineering expertise behind it. Delivering Helios on schedule and showing that its performance claims translate into a real production workload throughput advantage and total cost of ownership gains will determine how much the competitive gap narrows. And of course, this is in a market that is clamoring for ever-more compute resources with a seemingly insatiable demand for AI services and capacity. That’s an environment for big iron success. Now AMD just has to deliver optimized, turnkey AI platforms. This is far easier said than done, but time will soon tell as deployments take shape this year.

This article is published as part of the Foundry Expert Contributor Network.
Want to join?

OpenAI has launched Presence, an enterprise service for deploying voice and chat agents that can resolve customer and employee requests, potentially automating some work now handled by frontline support teams.

The agents can answer questions and operate IT systems, and enterprises can decide what actions the agents may take and when they should seek human approval for actions or transfer a case to a human.

OpenAI is already using Presence internally for its English-language phone support channel, where it verifies callers and uses account information to complete approved actions. The company said the system resolves 75% of inbound issues without human assistance.

Another OpenAI service, Codex, can be used to monitor agents and suggest updates or improvements to processes. In OpenAI’s own tests, suggestions from Codex helped reduce handoffs to humans by 15 percentage points over 10 days, it said. Presence also includes simulation and evaluation tools that allow companies to test an agent before deployment. The tests assess whether it reaches the correct outcome, follows company policy, and hands a case to an employee when required.

OpenAI intends each Presence deployment to deal with one kind of task, for example billing issues, insurance claims, or employee IT service requests, with agents getting only the knowledge and system access required for that task.

Presence is not a self-service product: Enterprises will have to sign up for the limited availability program, with integration performed by OpenAI or selected global systems integrators.

Companies exploring or testing Presence include Spanish bank BBVA, which is evaluating the service for everyday banking support in Mexico, and Japanese technology group SoftBank, which is using it in trials involving Japanese-language customer interactions. Australian insurer IAG is assessing whether the technology can help it respond to surges in customer demand during severe weather events.

Workforce impact

OpenAI’s announcement did not address the potential effect of Presence on employment. But its claimed automation rate raises questions about how the technology could affect staffing in customer service and other support functions.

Pareekh Jain, CEO of Pareekh Consulting, said CIOs should regard the 75% figure as evidence that the technology can work, rather than as a benchmark that every enterprise can expect to reach.

Jain said OpenAI’s deployment benefits from being built around the company’s own products and data. Large enterprises may achieve lower automation rates because they must contend with fragmented legacy systems, uneven knowledge bases and more complex compliance demands.

“Most organizations should expect lower initial automation levels that improve over time as the AI agent is refined,” Jain said.

The first workforce effect is more likely to be slower hiring than immediate layoffs, according to Tulika Sheel, senior vice president at Kadence International.

“The roles most exposed are likely to be repetitive, high-volume functions such as frontline customer support and routine back-office processing,” Sheel said. “However, I would expect the first impact to be on hiring and team growth rather than immediate large-scale job cuts. Over time, enterprises may redesign roles around AI-assisted workflows, with humans focusing more on complex cases, escalation, and relationship management.”

Jain said Tier-1 support agents handling predictable queries would face the most exposure. Broader reductions would become more likely only after companies reorganize their operations around the technology.

However, Lian Jye Su, chief analyst at Omdia, said Presence is unlikely to increase the threat of job displacement because companies have used similar customer-support automation from vendors such as Genesys, NiCE, Five9 and AWS for years.

Enterprises are more likely to use Presence alongside employees, with AI handling routine requests while people remain responsible for work requiring judgment and empathy, Su said.

Cost and operational risks

Analysts said CIOs should examine whether Presence can maintain resolution quality as usage grows, since fewer human handoffs could leave employees dealing with a more difficult mix of cases.

“The key question is not simply how many tasks AI can handle, but whether it can handle them reliably at scale,” Sheel said.

The financial case will depend partly on the cost of connecting Presence to existing systems and maintaining the controls needed to govern its use, according to Jain. “Often the biggest cost of enterprise AI is not tokens but integration and governance,” Jain added.

Companies will need to determine what systems and data the agents can access, monitor their performance, and audit the actions they take. Those investments could offset early savings.

Su said the complexity of enterprise IT will make it difficult for OpenAI to automate entire workflows on its own. Enterprises will still need to work with other technology providers and human employees, while CIOs will favor systems that can be audited and integrated with existing infrastructure.

Jain said the economics could improve if companies use the same integrations and governance controls across additional workflows.

OpenAI has launched Presence, an enterprise service for deploying voice and chat agents that can resolve customer and employee requests, potentially automating some work now handled by frontline support teams.

The agents can answer questions and operate IT systems, and enterprises can decide what actions the agents may take and when they should seek human approval for actions or transfer a case to a human.

OpenAI is already using Presence internally for its English-language phone support channel, where it verifies callers and uses account information to complete approved actions. The company said the system resolves 75% of inbound issues without human assistance.

Another OpenAI service, Codex, can be used to monitor agents and suggest updates or improvements to processes. In OpenAI’s own tests, suggestions from Codex helped reduce handoffs to humans by 15 percentage points over 10 days, it said. Presence also includes simulation and evaluation tools that allow companies to test an agent before deployment. The tests assess whether it reaches the correct outcome, follows company policy, and hands a case to an employee when required.

OpenAI intends each Presence deployment to deal with one kind of task, for example billing issues, insurance claims, or employee IT service requests, with agents getting only the knowledge and system access required for that task.

Presence is not a self-service product: Enterprises will have to sign up for the limited availability program, with integration performed by OpenAI or selected global systems integrators.

Companies exploring or testing Presence include Spanish bank BBVA, which is evaluating the service for everyday banking support in Mexico, and Japanese technology group SoftBank, which is using it in trials involving Japanese-language customer interactions. Australian insurer IAG is assessing whether the technology can help it respond to surges in customer demand during severe weather events.

Workforce impact

OpenAI’s announcement did not address the potential effect of Presence on employment. But its claimed automation rate raises questions about how the technology could affect staffing in customer service and other support functions.

Pareekh Jain, CEO of Pareekh Consulting, said CIOs should regard the 75% figure as evidence that the technology can work, rather than as a benchmark that every enterprise can expect to reach.

Jain said OpenAI’s deployment benefits from being built around the company’s own products and data. Large enterprises may achieve lower automation rates because they must contend with fragmented legacy systems, uneven knowledge bases and more complex compliance demands.

“Most organizations should expect lower initial automation levels that improve over time as the AI agent is refined,” Jain said.

The first workforce effect is more likely to be slower hiring than immediate layoffs, according to Tulika Sheel, senior vice president at Kadence International.

“The roles most exposed are likely to be repetitive, high-volume functions such as frontline customer support and routine back-office processing,” Sheel said. “However, I would expect the first impact to be on hiring and team growth rather than immediate large-scale job cuts. Over time, enterprises may redesign roles around AI-assisted workflows, with humans focusing more on complex cases, escalation, and relationship management.”

Jain said Tier-1 support agents handling predictable queries would face the most exposure. Broader reductions would become more likely only after companies reorganize their operations around the technology.

However, Lian Jye Su, chief analyst at Omdia, said Presence is unlikely to increase the threat of job displacement because companies have used similar customer-support automation from vendors such as Genesys, NiCE, Five9 and AWS for years.

Enterprises are more likely to use Presence alongside employees, with AI handling routine requests while people remain responsible for work requiring judgment and empathy, Su said.

Cost and operational risks

Analysts said CIOs should examine whether Presence can maintain resolution quality as usage grows, since fewer human handoffs could leave employees dealing with a more difficult mix of cases.

“The key question is not simply how many tasks AI can handle, but whether it can handle them reliably at scale,” Sheel said.

The financial case will depend partly on the cost of connecting Presence to existing systems and maintaining the controls needed to govern its use, according to Jain. “Often the biggest cost of enterprise AI is not tokens but integration and governance,” Jain added.

Companies will need to determine what systems and data the agents can access, monitor their performance, and audit the actions they take. Those investments could offset early savings.

Su said the complexity of enterprise IT will make it difficult for OpenAI to automate entire workflows on its own. Enterprises will still need to work with other technology providers and human employees, while CIOs will favor systems that can be audited and integrated with existing infrastructure.

Jain said the economics could improve if companies use the same integrations and governance controls across additional workflows.

This article first appeared on CIO.

Google’s AI offerings span its internal and cloud offerings. Its data centers are processing seven times more AI tokens compared to last year. To keep up, Google is upgrading its data-center hardware and software technologies at a faster clip. It plans to raise $80 billion to build new data centers. (See related story: Google transforms its data center architecture for agent era)

Network World spoke with Mark Lohmeyer, vice president and general manager of AI and computing at Google, about how the company’s infrastructure is keeping pace with AI demand.

Network World: What is the primary shift in infrastructure needs?

Mark Lohmeyer: We’ve seen the rise of agents and agentic use cases. Years ago, it was the chat phase: Ask a question, get an answer. Now we’re in the agentic era, where you express your intent, agents spin off multiple sub-agents, working in parallel, preserving state. This is a radical shift in what infrastructure needs to do; make them fast, cost effective, secure, reliable. We’re delivering infrastructure optimized for the age of agents.

NW: What’s the goal of the infrastructure buildout, and what should customers expect regarding costs?

ML: Ultimately, it’s about enabling customers with leading-edge capabilities and models at scale cost-effectively. With agents, inference transactions increase by 50x, 100x versus non-agentic workloads. We’re driving the cost per transaction down exponentially. In our latest platforms, we reduce the cost by almost 2x for the same work. Customers serve twice the number of users at the same cost, directly driving profitability.

NW: How are you addressing energy efficiency?

ML: Energy is a critical resource, and Google has optimized for years. We design data centers and compute [to drive] high PUE (power usage effectiveness). We introduced liquid cooling over five years ago, and these latest systems are all liquid cooled. For agentic workloads, CPUs come to the forefront… orchestrating agents, calling tools, doing evaluation loops in reinforcement learning. Our latest Axion-based CPU platform called N4A has energy efficiency and is significantly better than the prior generation and x86 comparables.

NW: How do you think about token efficiency as you build-out systems?

ML: Performance and efficiency gains are powered by co-design of the model and infrastructure. Gemini is trained on TPUs, primarily served on TPUs with high frontier model capability, in a token and cost-efficient way. This stems from co-design across the full stack.

NW: How do you project what infrastructure will be needed years in advance?

ML: Hardware cycles deliver a new next generation roughly every year, but design cycles are two years or more in advance. We work with DeepMind doing core research, to application teams taking models into production, to billions of users, to our team building infrastructure. We work upstream with DeepMind and application teams to understand what’s coming. Agents weren’t being broadly spoken of externally, but internally we had those insights around what they would need. That shows up in hardware design. We hit the timing right — these platforms are built for agents.

NW: What’s the eighth generation TPU platform?

ML: We deliver new platforms every year, and ones launched years ago are close to 100% utilized because demand for AI-optimized compute is high. The eighth-generation TPU platform is the first delivering two complete systems, from the chip all the way up to the network and storage and software, that are optimized.

TPU-8t is optimized for training, and TPU-8i is optimized for inference. For TPU-8i, we increased SRAM on the chip to 384MB — three times the prior generation — and increased the HBM by 50%.

NW: How are you approaching GPU and TPU compatibility?

ML: People in a single cluster do not commingle GPUs and TPUs. We offer both options based on specific workload needs. We’ve been investing on the TPU side in using software frameworks customers are comfortable with on GPUs and enabling those on TPUs. For example, PyTorch and vLLM. Customers could have a pool of GPUs and TPUs, running vLLM on top of that. Start with a workload on TPUs, but if the TPU pool is fully utilized, spill to GPUs or vice versa. This works because it’s all leveraging the same compatible software layer on top.

NW: How has the orchestration platform changed for agents?

ML: Kubernetes is becoming the orchestration platform of choice for AI. Google is transforming GKE [Google Kubernetes Engine] into an agent-native orchestration solution. When expressing intent to an agent and it spins up multiple sub-agents, compute needs to spin up rapidly — TPUs or GPUs — without long delays, then run and spin back down. We’re optimizing at every layer of the GKE stack: significantly improving node startup time and how rapidly we start and stop containers. Lovable demonstrates this with GKE, spinning up hundreds of sandboxes for live coding sessions on their platform in parallel, paying for infrastructure when needed.

NW: What is the role of the network and storage infrastructure?

ML: The network is critical for AI. This requires creating large-scale clusters of GPUs or TPUs and enabling them to talk to each other in a high-performance way. We created the Virgo network — a collapsed network architecture, non-blocking within a data center, where multiple pods or NVLink72 domains connect together.

In TPU8T, we can connect over a million TPUs together leveraging Virgo, creating large-scale, high-performance, reliable clusters that shrink innovation cycles. Storage is equally critical. In large-scale clusters, something is always failing. The ability to take snapshots and go back to a checkpoint is important.

We’ve introduced Managed Lustre 10T, with 10 terabytes per second of bandwidth, 18 petabytes of storage in single clusters. This is 10 times faster than last year and 20 times faster than competition. We have Rapid Bucket, low-latency storage backed by Google storage systems. Both are impactful in large-scale training environments.

NW: How does KV cache strategy differ between training and inference?

ML: For TPU-8i, we increased SRAM on the chip to 384 megabytes — three times the prior generation — and increased the HBM by 50%. Storing KV cache directly in chip memory allows responding to inference requests much more rapidly and cost-effectively than going to an external system. For inference workloads, storing as much KV cache as possible on-chip is critical.

We’re introducing a dedicated KV cache storage subsystem that works across GPUs and TPUs. As KV caches get larger, being able to fall back to this dedicated subsystem becomes critical. Loading model weights rapidly is important in dynamic inference environments where accelerators switch between models hour by hour.

Nelson Antoine | shutterstock.com

Der heimliche Cybercrime-Akt zweier KI-Modelle von OpenAI hat weltweit ein enormes Echo in Mainstream– und sozialen Medien hervorgerufen. Der Vorfall dürfte die Debatte über die allgemeine KI-Sicherheit und den verantwortungsvollen Umgang mit der Technologie neu befeuern.

Doch der Incident wirft auch spezifische Fragen auf. Etwa, wie genau die OpenAI-Modelle es geschafft haben, ihrer Sandbox zu entkommen und warum das beim ChatGPT-Erfinder zunächst niemandem aufgefallen ist. Oder, wie andere Unternehmen solche und ähnliche Vorkommnisse künftig verhindern können. Dazu haben wir die Einschätzung von Branchenexperten und Analysten eingeholt.

Zunächst werfen wir aber noch einen kurzen Blick darauf, was sich eigentlich abgespielt hat. Falls Sie bereits informiert sind, können Sie alternativ auch das nachfolgende Meme konsumieren, um sich den Vorfall noch einmal auf unkonventionellere Art und Weise vor Augen zu halten.

Der autonome Hugging-Face-Hack

Die KI-Plattform Hugging Face meldete Mitte Juli einen Cyberangriff auf ihre Produktions-Infrastruktur. Diese wurde offenbar vollständig autonom von einem agentenbasierten KI-System gefahren, das sich dabei sowohl Zugriff auf interne Datensätze als auch auf Zugangsdaten verschaffen konnte.  

Wie sich im Rahmen der folgenden Untersuchung herausstellte, stammte die autonom hackende KI von OpenAI. Eigentlich wollte der KI-Pionier lediglich zwei seiner neuen Modelle – GPT 5.6 Sol sowie ein nicht näher spezifiziertes Pre-Release-Modell – innerhalb einer Sandbox mit dem Benchmarking-Tool ExploitGym testen.

Die Aufgabe für die KI: Schwachstellen in tragfähige Exploits zu verwandeln. Die Testumgebung selbst war OpenAI zufolge mit beschränktem Netzwerkzugriff ausgestattet. Die KI sollte lediglich auf einen intern gehosteten Drittanbieter-Service zugreifen können, der als Proxy und Cache für eine Package-Registry fungierte. Die Guardrails auf Modellebene waren für den Test laut dem KI-Unternehmen reduziert worden.

Das war der KI aber offenbar nicht genug: Sie verkettete zunächst einige zuvor unbekannte Schwachstellen in der Testumgebung, um sich uneingeschränkten Internet-Zugriff zu verschaffen, wie OpenAI in einem Blogbeitrag offenlegt: „Mit diesem Zugriff führten unsere Modelle in unserer Forschungsumgebung eine Reihe von Privilege-Escalation- und Lateral-Movement-Aktionen aus, bis sie eine Node mit Internetzugang erreichten.“

Weil die OpenAI-Modelle bei Hugging Face offenbar Lösungen für ihre ExploitGym-Aufgaben vermuteten, suchten sie nach Möglichkeiten, Zugriff auf diese zu erhalten. Und fanden sie dann auch, wie OpenAI darlegt: „In einem Beispiel verkettete das Modell mehrere Angriffsvektoren, darunter gestohlene Zugangsdaten und Zero-Day-Schwachstellen, um auf den Servern von Hugging Face einen Remote-Execution-Pfad zu finden.“

Das Security-Team von OpenAI konnte diese Vorgänge im Nachgang identifizieren und schloss sich daraufhin mit dem Team von Hugging Face zusammen, welches die Attacke seinerseits bereits identifiziert und (mit einigen Anlaufschwierigkeiten) eingedämmt hatte.

„Wir betrachten dies als beispiellosen Cybervorfall mit hochentwickelten Fähigkeiten und reagieren entsprechend. Wir teilen zu diesem Zeitpunkt vorläufige Erkenntnisse, damit Sicherheitsverantwortliche nachvollziehen können, was passiert ist, und besser einschätzen können, wozu die Modelle inzwischen in der Lage sind“, schreibt OpenAI in seinem Blog – und verspricht, weitere Details zu veröffentlichen, sobald diese vorliegen.

KI-Ausbruch bei OpenAI – so reagieren Experten

Branchenexperten und Analysten bewerten den schlagzeilenträchtigen Incident um OpenAI und Hugging Face folgendermaßen:

• Alexei Balaganski, Lead Analyst bei KuppingerCole: „Dieser Vorfall sollte nicht als ‚Rogue AI‘-Geschichte betrachtet werden. Das Modell hat exakt das getan, wofür agentische Systeme gemacht sind: Es hat sich allen verfügbaren Tools und Wegen bedient, um das ihm gesetzte Ziel zu erreichen. Die Sicherheitsvorkehrungen, die es normalerweise in Zaum gehalten hätten, wurden von OpenAI selbst zu Testzwecken deaktiviert. Darin besteht die wahre Lektion.“
• Jonathan Care, Lead Analyst und AI Practice Lead bei KuppingerCole: „Es geht bei diesem Vorfall nicht darum, dass eine KI ausgebrochen ist und zum Angreifer wurde. Wir wussten, das würde passieren. Bemerkenswert ist allerdings, dass die Verteidiger – in diesem Fall das Team von Hugging Face – keine kommerziellen KI-Modelle nutzen konnten, um den Angriff zu analysieren. Denn deren Guardrails sorgen dafür, dass kein Exoploit-Code verarbeitet werden kann.“
• Gerald Beuchelt, CISO bei Acronis: „Der Vorfall verdeutlicht eine zentrale Herausforderung für Incident-Response-Teams: Angreifer sind nicht an Nutzungsrichtlinien gebunden. Verteidiger können hingegen an die Grenzen ihrer eigenen Tools stoßen, wenn diese genau jene Daten nicht verarbeiten, die für eine Untersuchung erforderlich sind. Im Ernstfall können daraus Verzögerungen mit unmittelbaren operativen Folgen entstehen.“
• Sabine Frömling, Experten-Autorin und Cybersecurity-Beraterin: „Der eigentliche Sicherheitsvorfall war nicht die KI – sondern die Sandbox, die aus Versehen eine Tür zum Internet hatte. Man hat ein Raubtier freigelassen und dem Zaun die Schuld gegeben.“
• Martin Zugec, Technical Solutions Director bei Bitdefender: „Was meiner Meinung nach für KI-generierte Malware galt, untermauert auch dieser Vorfall: Die Bedrohung ist real, KI ist aber keine Magie. Wer glaubt, es mit einer neuartigen Superwaffe zu tun zu haben, wartet auf eine neuartige Gegenmaßnahme. Wer jedoch erkennt, dass es sich um bereits bekannte, aber unerbittlich angewandte Angriffstechniken handelt, weiß bereits, was zu tun ist.“
• Richard Werner, Cybersecurity Platform Lead Europe bei TrendAI: „Das Narrativ von der ‚eigenmächtig handelnden KI‘ ist effizient darin, Verantwortung abzuwälzen. Das ist, als würden Sie eine autonome Waffe bauen, diese auf einem vermeintlich sicheren Testgelände erproben, sie außer Kontrolle geraten und jemanden treffen lassen – und der Welt anschließend erklären, die Waffe habe eigenständig gehandelt. Das ist zwar technisch korrekt. Dennoch bleibt es Ihre Waffe, Ihr Testgelände und Ihr Versagen.“

Was Unternehmen jetzt tun sollten

IT- und Sicherheitsentscheider können aus dem Hugging-Face-Hack mehrere Lektionen ziehen. Etwa, dass Sicherheitsvorkehrungen auf Modellebene nicht als primäre Security-Grenze für KI-Agenten geeignet sind, wie Biswajeet Mahapatra, Principal Analyst bei Forrester, festhält: „Prompt-Guardrails sind keine Sicherheits-, sondern Verhaltenskontrollmaßnahmen. Und diese können versagen, umgangen oder absichtlich deaktiviert werden.“

Der Forrester-Analyst rät Unternehmen deshalb dazu, KI-Agenten als hochriskante, nicht-menschliche Identitäten zu behandeln – und jeden einzelnen in einer isolierten Umgebung zu betreiben, in der Datenzugriff auf den jeweiligen Task beschränkt bleibt und die Zugangsdaten selbst möglichst schnell ablaufen: „Das sorgt für einen akzeptablen ‚Blast Radius‘: Wird ein Agent kompromittiert, kann er nur einen einzigen Workflow, Datensatz oder eine einzige Anwendung beeinträchtigen. Anstatt die gesamte Unternehmensinfrastruktur.“

Sanchit Vir Gogia, Chefanalyst bei Greyhound Research, warnt an dieser Stelle davor, (Drittanbieter-)Services unter den Tisch fallen zu lassen: „Dienste, die auf Package Registries, Update-Systeme oder andere externe Ressourcen zugreifen, können ebenfalls zu Einfallstoren werden, wenn sie nicht derselben, ausgiebigen Prüfung unterzogen werden wie der Agent selbst.“

Unabhängig davon sollten Unternehmen laut Gogia auch testen, ob ihre Containment-Grenzen auch funktionieren, anstatt sich allein auf Architekturdiagramme oder dokumentierte Richtlinien zu verlassen: „Im Rahmen dieser Tests sollte geprüft werden, ob Anmeldedaten erlangt, Trust-Grenzen überwunden und Systeme außerhalb der einem Agenten zugewiesenen Aufgabe erreicht werden können.“

KuppingerCole-Chefanalyst Care rät IT-Entscheidern und Unternehmen im Wesentlichen zu drei Maßnahmen, nämlich:

• ein fähiges Modell auf der eigenen Infrastruktur auszuführen, das unter der eigenen Kontrolle steht und mit Guardrails ausgestattet ist, die sowohl eine forensische als auch defensive Nutzung ermöglichen. Nur so ließen sich Angriffe dieser Art auch zuverlässig analysieren.
• jeden KI-Agent in der eigenen Umgebung als privilegierten Insider zu behandeln – statt als vertrauenswürdigen Benutzer: „Wenn die Modelle von OpenAI aus ihrer Sandbox ausgebrochen sind, sollten Sie davon ausgehen, dass Ihre Agenten dazu auch in der Lage sind.“
• den eigenen Incident-Response-Plan mit Blick auf Angriffe in maschineller Geschwindigkeit zu aktualisieren: „Hugging Face hatte einige Tage Zeit, um zu reagieren, Sie haben vielleicht nur Minuten.“   

Acronis-CISO Beuchelt rät Organisationen, die gehostete LLMs für Security-Untersuchungen einsetzen, dazu, deren Grenzen möglichst bereits im Vorfeld zu durchdringen und zu testen – sowie ein alternatives Modell auf der eigenen Infrastruktur bereitzuhalten: „So reduzieren Sie das Risiko, im entscheidenden Moment keinen Zugriff auf wichtige Analysefunktionen zu haben. Gleichzeitig bleiben sensible Incident-Daten und Zugangsinformationen innerhalb der eigenen Organisation.“

Udo Schneider, Governance, Risk & Compliance Lead Europe bei TrendAI weist darauf hin, dass die beiden naheliegendsten Lösungsansätze bei Angriffen wie dem der OpenAI-KI auf Hugging Face nur teilweise greifen. Human-in-the-Loop-Kontrollen funktionierten zwar, so der Experte, skalierten aber nicht für die langlaufenden, komplexen Workflows, denen Incidents dieser Art entspringen. Ebenso könnten engere Guardrails für Modelle oder Prompts zwar helfen, stellten jedoch keine Garantie dar: „Es handelt sich um probabilistische Systeme. Eine Guardrail ist insofern keine Mauer, sondern eher eine starke Wahrscheinlichkeitsannahme.“

Deshalb komme es laut Schneider vor allem auf die unspektakulären, nicht-KI-spezifischen Kontrollen an: „Zugriffsfilterung, Kontrolle darüber, was überhaupt als Input beim Modell ankommt, Sandboxes, die tatsächlich halten, und Berechtigungskonzepte nach dem Least-Privilege-Prinzip.“

In Panik zu verfallen, wäre nach Ansicht von Martin Zugec, Technical Solutions Director bei Bitdefender, in jedem Fall die falsche Reaktion:„Was gegen solche Angriffe wirkt, ist eine präventionsorientierte Security, die den Handlungsspielraum eines Angreifers von vorneherein einschränkt – und eine verhaltensbasierte Abwehr, die bösartige Muster kennzeichnet, unabhängig davon, mit welchen Tools diese generiert wurden.“

Dieser Artikel wurde mit Material unserer Schwesterpublikation CSOonline.com angereichert.

The traditional methods of measuring success no longer tell the full story. Here’s what should replace them — and why.

In brief:

• AI is transforming software as a service (SaaS), and the old ways of keeping score no longer apply.
• Smart companies are evolving new metrics that provide deeper insight into how AI-native software is performing in a new marketplace.
• These changes impact everything from pricing to valuations.

The transformation of the software-as-a-service (SaaS) industry toward AI-native operating companies is rapidly changing the unit of value across the industry.

The traditional metric of seats — which measured access — is rapidly giving way to credits designed to measure work performed. This evolution is upending the industry in multiple ways, impacting everything from pricing to enterprise valuations.

While many companies still cling to seat-based metrics to measure growth, efficiency and durability, the future is likely to be one in which companies utilize a credit-centric metrics framework, with seats and outcomes as the bookends of a spectrum.

Why do software companies need new metrics?

Why the rethink, and why now? There are five major forces that are driving this shift:

1. The unit of value is changing. Seats measured who could access software, and credits measure what the software actually does. But in an AI-native world, agents don’t have seats; they have workloads. Over the past 18 months, every major SaaS platform has moved to some forms of credit or consumption unit.
2. The cost of goods sold (COGS) is exploding. AI inference adds real per-unit costs that scale with usage. In an AI-native world, software companies can’t scale to infinite users at near‑zero marginal cost as before.
3. Buying is moving up the org chart. AI-native applications shift purchasing to higher-level operators — such as line-of-business leaders or chief operating officers — which expands the market from software budgets to labor budgets. And because AI agents replace services as well as software, the total market opportunity is 3x to 10x larger than traditional SaaS.
4. Time to value (TTV) is collapsing. With AI-native tools, customers start seeing meaningful results in weeks rather than quarters. Onboarding and setup are fast, workflows are pre-built, and there’s no need for extensive customer success or professional services — dramatically reducing implementation time and costs.
5. Retention is bifurcating. AI forces clarity in a way that traditional SaaS couldn’t. Products that can provide value become even “stickier” and retain customers. Those that don’t churn faster. In an AI-native marketplace, the middle disappears.

How this shift is impacting pricing

Given how AI-native software is transforming the market, the shift to more variable pricing options is inevitable.

Seats won’t go away completely. Subscription pricing based on the number of users is stable and predictable and will continue to work for some customers. Tokens — the use of pass-through pricing for underlying compute — will fit those customers where the AI feature is commoditized or the buyer wants transparency into costs.

Credits will likely become the dominant architecture because they provide a simple metric for both customers and providers. The vendor sets the conversation ratio between credits and underlying compute, shielding the customer from inference cost details. Credits are easy to understand and can be packaged into annual contracts for multiple features and products.

Finally, the industry will likely see some move toward outcome-based pricing for results such as resolved tickets, recovered revenue or qualified leads. This strategy will mostly be limited to verticals where it is easy to prove AI impacted the result.

Where a software vendor sits on this spectrum is a signal of differentiation and pricing power. Credits are where most defensible AI-native businesses are landing because they balance customer predictability with vendor margin control.

How AI upends classic SaaS metrics

When SaaS was in its infancy, companies settled on key metrics designed to answer a small set of core questions. Are we growing? Are customers using the product? Are we retaining and expanding accounts?

But as AI upends software itself, it is also requiring companies to adopt new metrics to track success. These new metrics fall into three primary buckets, rebuilt around the pricing spectrum described earlier and the trend toward credits as the primary frame:

Revenue composition

• Committed credit annual recurring revenue (ARR) vs. burndown ARR: Measuring the credits sold on annual commitment vs. those consumed and replenished. This is the single most important split for valuation. Committed credits behave like subscription and burndown behaves like usage.
• Credit utilization rate: The percentage of purchased credits consumed per period. This is a leading indicator of renewal sizing.
• Credit burn velocity: How fast is a customer consuming their credits, and is that consumption increasing or decreasing quarter over quarter? This metric predicts expansion or contraction before it shows up in ARR.
• Effective price per credit: The real revenue per credit after discounts, overage and rollover, which can detect revenue leakage and help companies set smarter guide rails.

Margin reality

• Credit margin: The gross profit the company earns per credit after subtracting inference costs. This is the core economic unit for AI-native, usage-based businesses — the replacement for gross margin per seat used in SaaS.
• Inference-adjusted gross margin: By carving out AI inference costs separately in the P&L statement, you can see true AI margins, avoid hiding deterioration inside blended SaaS margins, and clearly distinguish AI economics from legacy SaaS economics.
• Compute leverage ratio: This metric measures how efficiently the business converts compute spend into revenue. It shows whether your AI margins are improving as you scale.
• AI-adjusted “Rule of 40”: This updated metric recalibrates the traditional growth and profitability benchmark to account for AI’s lower gross margins and variable inference costs, giving a more accurate picture of business health for AI-native companies.

Behavioral and value signals

• Time-to-first outcome: Replaces traditional onboarding metrics. Tracks how fast a customer reaches their first measurable result.
• Adoption: AI-native adoption is measured by workflow penetration and active agent density, not seat count. As AI replaces human-driven usage, the unit of adoption shifts from people to automated workflows and agents.
• Net credit retention (NCR): Credit-volume retention across the customer base, tracked separately from net recurring revenue to avoid price-change impact.

Along with these new metrics, the industry’s transformation is prompting companies to retire or recalibrate old SaaS measures, including per-seat ARR as a primary key performance indicator (KPI), traditional magic number calibrated to subscription dynamics, unadjusted Rule of 40, customer success metrics tied to human touchpoints, and blended gross margin without AI COGS carve-outs.

What does this mean for enterprise value calculations?

As the internal metrics of success change, so do the ways the investment community measures growth and long-term viability.

Increasingly, a company’s valuation multiple depends on whether its revenue behaves like committed subscription ARR or volatile usage ARR, and the commit‑to‑burndown ratio is the metric investors use to decide where the company fits.

For example, a business with 80% committed credit ARR could trade closer to subscription comps and one with 80% burndown could trade closer to usage comps even though both have the same types of customers. Being able to proactively explain the commit‑to‑burndown mix can help companies avoid undervaluation.

In addition, utilization is expected to replace net promoter scores and seat usage as the primary predictor of churn or expansion. Low utilization guarantees downsizing at renewal, so companies must track utilization cohorts the same way SaaS tracks logo retention cohorts today.

We’re also seeing an inversion of the operating model, with R&D and COGS moving up the P&L and sales and marketing (S&M) and customer success (CS) moving down or sideways. The net operating leverage profile is structurally different from classical SaaS, and the cost-to-scale curve looks different too.

Finally, credit margin engineering is a hidden value-creation lever. The gap between price per credit and cost per credit is set by the software vendor and can be optimized. Most operators have barely started managing this rigorously, and the ones who do will pull away on margin.

What this means for leaders, boards and investors

The shift from classic SaaS metrics to new AI‑native measures isn’t cosmetic. It represents the seismic change the industry is experiencing as AI matures and transforms products and organizations.

While these metrics — and perhaps others yet to be determined — may evolve over time, there is no doubt they are already changing how AI companies allocate capital, price products, incent sales teams, evaluate performance and communicate with investors.

It’s important to remember that SaaS metrics were practical tools for a specific era of software. As that era draws to a close, winning companies will choose new metrics that shape behavior and drive smart decision-making.

The views reflected in this article are the views of the author and do not necessarily reflect the views of Ernst & Young LLP or other members of the global EY organization.

This article is published as part of the Foundry Expert Contributor Network.
Want to join?

Over the past few years, I have learned that basic prompts produce inconsistent, hallucination-prone results that no executive would trust in production. What turned the tide was my move to advanced prompting techniques. These weren’t theoretical experiments; they became a practical foundation for reliable, measurable outcomes. I want to share the techniques that consistently delivered the biggest gains in my projects, complete with real before-and-after examples, copy-paste templates, lessons from failures and guidance on when to evolve beyond prompting to agentic systems.

Why advanced prompting still matters in enterprise settings

Sophisticated prompting remains essential for control, reliability and compliance. If you “ask nicely” and hope for the best, you need deterministic behavior, auditable reasoning and minimal risk of hallucination. Here’s what worked for me.

1. Chain-of-Thought (CoT) and its variants: Unlocking step-by-step reasoning

The problem: Models would jump to conclusions on complex analysis tasks, especially involving data interpretation or multi-step logic.

What I did: I started explicitly instructing the model to “think step by step” and show its reasoning.

Before (basic prompt): “Analyze last quarter’s sales data and recommend three actions.”

After (CoT prompt):

“You’re a senior business analyst. Analyze the following sales data step by step: [data]. First, identify the key trends. Second, calculate the rates and anomalies. Third, link findings to business context. Finally, recommend the three prioritized actions with expected impact. Explain your reasoning at each step.”  

Results: Accuracy and depth improved dramatically.

Variants that worked well: Self-consistency. I ran the same CoT prompt multiple times and took the majority consensus. This reduced variability significantly.

Template you can use:

You are [expert role]. Solve this problem by thinking step by step.

[Task or question]

For each step:

1. State your observation or calculation.

2. Explain the implication.

3. Proceed only when confident.

Final answer in this format: [structured output]

2. Tree-of-Thoughts (ToT): Exploring multiple reasoning paths

For truly complex decisions such as resource allocation or risk assessment, linear CoT isn’t enough. Tree-of-Thoughts lets the model generate and evaluate multiple branches.

Example: I was helping a client evaluate three potential vendor platforms for an AI deployment. A standard prompt gave a superficial comparison. With ToT

Prompt Snippet:

Explore three different reasoning paths for selecting the best vendor platform:

Path 1: Focus on cost and scalability.

Path 2: Focus on security, compliance and integration.

Path 3: Focus on innovation and long-term roadmap.

For each path, evaluate pros/cons against our requirements [list].

Then, compare the paths and recommend the strongest overall option with justification.

Outcome: The model surfaced nuanced trade-offs (e.g., one vendor had superior security, but higher integration cost).

When to use: Strategic planning, troubleshooting or scenarios with high uncertainty and multiple viable approaches.

3. ReAct (Reason+ Act) and prompt chaining: Moving toward agentic behavior

One of the biggest leaps I have noticed comes from combining reasoning with tool use and chaining prompts.

ReAct example: (used in data analytics workflow)

You are an AI analyst with access to tools. For the query below:

1. Reason about what information you need.

2. Choose the appropriate tool or action.

3. Observe the result.

4. Repeat until you can answer confidently.

Query: [user request]

In practice, I chained this with retrieval tools. One automated quarterly compliance reporting; the system reasoned about required data, pulled relevant records, validated them, and generated the reports.

4. Meta-prompting and self-reflection: Letting the model improve itself

Use the model to refine its own prompt. This is a huge time-saver.

You are an expert prompt engineer. Improve the following prompt for clarity, structure and effectiveness with [target model]. Make it more precise while preserving intent.

Original prompt: [paste]

Provide the improved version and explain your changes.

Self-reflection loops (asking the model to critique its own output and revise) are a game-changer for content generation and code-review tasks.

5. Multimodal and structured output techniques

With vision-enabled models, I started combining text with images (e.g., uploading architecture diagrams or dashboards).

Tip from experience: Be extremely specific in describing what the models should focus on.

Best practices I learned the hard way

• Start simple, then layer complexity: Over-engineered prompts from Day One usually backfire.
• Model specific tuning: Some models respond better to XML delimiters; others to explicit reasoning.
• Evaluation and versioning: Treat prompts like code if you track versions and run automated evals.
• Security guardrails: Always include instructions against prompt injections and respect data boundaries.
• When to stop prompting: For repetitive, high-stakes workflows, move to full agents or an orchestration framework.

Final takeaways for technical leaders

Advanced prompt engineering has now become a core competency for anyone responsible for enterprise AI outcomes. Start by picking one technique and apply it rigorously to a real business problem. Document before/ after and you will notice why it’s worth mastering.

The field continues evolving towards more automated and agentic systems, but the ability to precisely direct AI reasoning remains foundational.

This article is published as part of the Foundry Expert Contributor Network.
Want to join?

When it comes to building an effective AI stack, context is king and power isn’t everything it’s cracked up to be.

“Smaller, smarter, safer — this is a bet our company has taken in how we deploy AI internally,” said Ricky Thakrar, head of sales and account management at Zoho, provider of a suite of popular cloud-based software solutions for sales, marketing, and finance.

“I’m on the business side, and so decisions made by our CIO and IT folks affect me directly, and my teams’ workflows and processes,” he added.

Speaking to a room of tech leaders at the CIO 100 Leadership Live New York event last week, Thakrar explained that every company wants the speed of AI-generated work wedded to the quality of human work, even though these two are diametrically opposed. No amount of model upgrades or spend will close that gap, so the only way forward is to architect your way out. Thakrar encapsulated this idea in a simple formula:

• Smaller: Stop deploying maximum firepower on every task. Many tasks don’t need it.
• Smarter: The system around the model decides more than the model does.
• Safer: Verify at the point a mistake gets locked in, not just downstream of it.

He noted that organizations that win with AI won’t be those deploying the biggest, most powerful models or the most sophisticated architecture, but the ones that figure out that the model is the easy part and the right architecture is harder. That means understanding the hardest element, and the biggest differentiator, is building a human system that learns and compounds alongside agentic systems.

To get it right, organizations need to prioritize the context layer. The size of frontier models like the GPT series, Claude, and Gemini mostly exist to compensate for missing context, Thakrar explained. Without enough context, models need to be able to reason harder and infer more about what a user actually means because it doesn’t know the user’s account, process, or history. A rich context layer makes it possible for enterprises to run workloads on much smaller, lower-power models.

“The intelligence moves from the model into the architecture around it,” he said.

A steep learning curve

One of Zoho’s earliest AI agents was a churn management agent to help the account management team detect churn in customer subscriptions. So when a subscription became inactive, the agent would collect context from notes, meeting recordings, and Zoho’s data enrichment tool, then create a summary of reasons the account might have churned, and schedule a call.

“What happened was I got this churn agent a couple months later, already embedded in our CRM, and within a week my team no longer trusted that agent,” Thakrar said. “The reason is we forgot to collect one very key point.”

In Zoho’s CRM, when a customer buys a bundle of products, that bundle is represented as a single line item. That means the status of any products the customer may have previously purchased individually changes to inactive as they’re moved to the bundle. That’s not churn, but it was interpreted it that way. Zoho fixed it in the second version of the agent.

Then a new problem arose. Many potential customers first purchase Zoho products as pilots or sandboxes. As those customers move from pilot to live instance, they close down the pilot versions. And again, the CRM would record that as subscriptions going inactive.

“The trust deteriorates again because everyone got excited for version 2,” Thakrar said.

Sometimes, a certain product might not be the best fit for a customer and Thakrar’s team will suggest the customer move to another product. That’s deliberate churn, not a churn risk.

“You may have a similar story like this where the agent sounds so good, it’s going to do something quick and add value, but it’s missing context from the account managers, and there are so many more pieces we’re still building out,” Thakrar said. “It’s been almost a year and the problem I have is my team still doesn’t trust it. They’ll see [a message from the agent] and go out and do all the research anyway to make sure it gave the correct answer.”

The team is more on top of potential churn, though, but the promised productivity gains have yet to materialize because the agent has to earn back lost trust due to a lack of context.

“My goal for this year is having an AI-assisted customer journey from sales to account management where the handoff is clean, the context flows, and every piece of information we gather about a customer is weighed, identified, and coached so the sales team can close more deals,” he said.

Zoho’s early experience with agents has led to the idea that constrained, context-rich, deterministic architectures consistently outperform expensive models bolted onto fragmented systems. It all comes down to three pillars: routing, harness, and specialization.

Routing

Routing is about sending workloads to the proper model for the job, which entails providing enough context to a given task that a small, cheap model can handle it without the need for spare reasoning capacity to fill gaps.

Frontier models are expensive and companies can burn through a year’s budget worth of tokens in months. But most tasks can be handled by much smaller, more constrained models at a fraction of the cost.

“You don’t always have to pay the frontier guys for every task,” he said. “We’ve observed with some clients that we could save them 95% with a 3 billion parameter model.”

Harness

An AI agent harness is the software infrastructure scaffolding around an LLM that differentiates an agent from a chatbot. It’s what enables an agent to act on tasks rather than simply respond to prompts. A model reasons through a problem and decides what to do about it. The harness connects the model to the tools, systems, memory, guardrails, and execution environments required to perform the actions determined by the model. The term is frequently used more or less interchangeably with orchestration layer.

“It’s the process around the model, which matters way more than the model itself,” Thakrar said.

In benchmark tests, a superior harness on a less powerful model produces better results than an inferior harness on a much bigger model.

For the best results, Thakrar said, it’s essential to understand the deterministic and non-deterministic elements of a given workload, and build that into the architecture. Machines can read, organize, and validate, and they excel at deterministic tasks. Humans, on the other hand, are exceptional at non-deterministic tasks like judging, synthesizing, and deciding.

Those non-deterministic tasks in a process are the ideal point for AI agents to incorporate a human in the loop, what Thakrar calls human harness. He pointed to a stakeholder mapping agent Zoho built for sales as an example, which takes the context of an initial meeting and third-party enriched data like a LinkedIn profile, weighs probabilities, and makes an educated guess about the stakeholder map.

“The initial goal was just to eliminate that task completely from the human workflow,” he said. “The stakeholder map is done, it’s in the folder, and you can look at it.”

But the agent would struggle to capture nuance. The meanings of titles in organizations always vary, and the politics and dynamics of any given meeting can be difficult for an AI agent to discern. Rather than keep feeding the agent data to try to make it intelligent enough to make those determinations, it was simpler and more efficient for the agent to create a proposed stakeholder map and hand it over to a human who could make changes and explain why those changes were necessary.

Ultimately, Thakrar said the agent still saved human team members time because the stakeholder map was usually pretty close, and the corrections also helped the model grow smarter by adding richer context.

Specialization

Specialization is transitioning a process from testing on a frontier model to production on a much narrower, smaller model. Once you’ve proven that an agent can do a job well, you want to stop paying master-craftsman rates to keep doing that one job well.

Specialization is all about capturing your subject matter experts’ best judgement and pattern recognition to build an open-weight, open source, trained, and fine-tuned model that can be deployed in your own data center.

“The true enterprise bet is to keep that orchestration layer, which is your IP and knowledge, in house,” Thakrar said. “You don’t want to host that on someone else’s model. The goal of everyone in enterprise should be to run, train, and host their own models.”

Leadership wants to scale AI. Budgets are tripling. Adoption is up.

Then the CFO asks the question every board now asks: which of these initiatives is actually profitable?

Most organizations cannot answer that question, not because they lack visibility into cost, but because the cost data they have was never designed to produce that answer.

Applying lessons learned from managing cloud spend won’t be a fix for the AI and ROI quandary. True, cloud taught a generation of CFOs that billing without business context is noise. So to get cloud ROI, they stitched two data sources together: cost data plus business data. AWS reveals which account, which region, which tag, which resource. Merge in customer and product mappings on top and the ROI of the cloud spend comes into focus.

But AI is harder. It requires three data sources: cost, business, and telemetry—the automatic collection of data from disparate sources that helps to clarify the whole picture of what happened and why. An executive or engineering lead can have AI invoices and customer revenue. But they have no way to connect them to business value. The token count on the OpenAI invoice does not specify which customer triggered which call, which feature it served, or whether the prompt produced a business outcome. That data does not exist in the provider’s billing.

AI providers won’t fix this problem

The situation is not likely to change anytime soon because AI providers are not in the business of attributing an enterprise’s costs to that enterprise’s customers. Instead, AI providers are in the business of selling tokens. The granularity they expose is the granularity their billing systems require, not the granularity a CFO requires.

Not convinced? Compare what AWS gives you to what an AI provider gives you.

AWS billing exposes resource IDs, account hierarchies, region, SKU, tag metadata, usage by the minute. Every dollar can be attributed to a workload, a team, a customer segment if it was tagged correctly. The data is rich enough that mature FinOps teams built unit economics on top of it years ago.

An AI provider invoice gives you tokens consumed by model, with optional grouping by API key. That is the resolution. No request-level attribution. No customer ID. No feature mapping. No prompt outcome. No retry identification. Multi-step agent workflows collapse into a token count. Imagine a large bank receives a multi-million dollar AI invoice each month. But it has no visibility into what parts of the business were responsible for what parts of the cost so cannot allocate them.

If an enterprise wants to know what AI cost drove which customer or feature, it has to capture that data itself, inside an application, before the call leaves it. 

Three required sources

Building AI ROI measurement requires three data sources, stitched together in a single model.

1. Cost data, normalized across providers. Every AI provider delivers cost differently. OpenAI invoices in one taxonomy, Anthropic in another, fine-tuning vendors and inference platforms each in their own. Cloud GPU costs sit in AWS or Azure billing. Vector database costs land in Pinecone or Snowflake invoices. None interoperate by default. Normalization is necessary but not sufficient. It will put all your AI costs in one schema. It does not tell you what they produced.
2. Application-layer telemetry. This is the source most organizations are missing, and the one that makes AI ROI structurally different from cloud ROI. It requires instrumenting AI calls inside your application across six categories: request-level tracing tied to a customer or session ID; feature attribution tied to the product surface that triggered the call; agent-step capture for multi-step workflows; retry and fallback identification so recovery costs don’t get attributed to primary calls; model selection logging that records which model was chosen and why; and outcome capture that ties each call to whether it produced business value. None of this data exists in the provider’s billing. All of it has to be captured at the moment the call is made and stored in a system that can be stitched to the cost data.
3. Business data. Revenue, customer segments, product hierarchies, and feature usage. The same business data already feeding your CRM and analytics stack, mapped to the customers and features the telemetry layer attributes calls to.

Stitched together, the three sources produce the unit economics every AI investment decision now requires: cost per customer interaction, margin per feature, profitability per agent workflow, ROI per model choice. None of these can be calculated from billing data alone. None can be calculated from telemetry alone. They require all three sources, modeled together in a way that maps cost to outcome.

Why agentic AI makes this urgent

Single-call inference is the easy case. One request, one cost, one customer, one outcome.

Agentic workflows are different. An agent decomposes a task into multiple steps. Each step calls a model. Some steps fall back to a different model when the first fails. Some steps retry on a poor result. Some steps invoke external tools that themselves cost money. A single user request can produce dozens of inference calls across multiple providers, with the cost compounding in ways the provider invoice cannot disaggregate.

If telemetry does not capture agent-step granularity, no one will know which steps are profitable. Aggregate costs will show up three weeks later in the invoice. By then, the workflow has been running at scale, customers are onboarded, and unprofitable paths have been retried thousands of times.

When agents make the calls, the volume of cost-generating events without business context attached grows by an order of magnitude. The window for instrumenting this before it becomes unmanageable is closing.

What changes when the three sources come together

Once the three sources are stitched together, the AI investment conversation changes.

Five different ways to build the same AI capability stop looking equivalent. They converge on adoption metrics and diverge by 10x on cost. The team picks the approach that delivers a similar business outcome at one-fifth the cost, because the team can finally see the difference. Product teams design features with margin awareness from the architecture phase, not from the post-launch budget review. Engineering teams choose model architectures with cost-per-outcome data alongside latency and quality. Leadership evaluates AI initiatives the way they evaluate any other capital allocation: on unit economics, not on the engagement chart. Aggregated invoices track the cost per customer interaction. Engagement metrics reveal margin per feature. Gut-instinct model selection is checked against real cost-per-outcome model selection results. 

Within seconds, everyone can see which AI features are profitable, which should scale, and which should be killed. This is the insight everyone is looking for and companies that achieve it will optimize the benefits of AI.

The build trap

AI costs are compounding now. The board is not waiting 18 months for an internal project to reach production.

The temptation to build it anyway has never been sharper. AI coding tools have changed what a small engineering team can ship in a quarter. The instrumentation layer looks tractable. The cost normalization looks like a weekend project. The semantic model feels like something a senior engineer could draft over a sprint.

It is a trap. Three reasons.

Volume is the first. A production AI footprint generates millions of telemetry events per hour, and that volume scales with agentic adoption. Real-time ingestion, correlation, and attribution at that scale is not the same problem as vibe coding a prototype in an afternoon. It is a permanent operational system that has to be right every minute of every day.

The vendor landscape is the second. Cost data arrives in delayed billing windows from providers with non-interoperable schemas. Schemas change without notice. New AI providers enter the landscape monthly, each with its own taxonomy and metering. The system is not built once. It is maintained against a moving target that moves faster than most internal release cycles.

The third is what the first two add up to: this is business-critical infrastructure. The CFO and the board are going to make capital allocation decisions on the data this system produces. When schema drift goes unnoticed for two weeks, when an agent telemetry stream stops correlating to a vendor that quietly changed its billing API, the cost of being wrong is not a sprint of cleanup. It is a quarter of misallocated capital.

The build-vs.-buy question for engineering leaders has changed. It’s not “can we build this?” The honest answer is yes. The real question is whether the marginal hour of your strongest engineers is best spent stitching cost data to telemetry to business outcomes, or building the AI products that produce the revenue the cost data is measuring.

The capability is reproducible in weeks. The choice is whether to spend the next 18 months building it, or the next 18 months acting on it.

—

New Tech Forum provides a venue for technology leaders—including vendors and other outside contributors—to explore and discuss emerging enterprise technology in unprecedented depth and breadth. The selection is subjective, based on our pick of the technologies we believe to be important and of greatest interest to InfoWorld readers. InfoWorld does not accept marketing collateral for publication and reserves the right to edit all contributed content. Send all inquiries to doug_dineley@foundryco.com.

Raushan_films | shutterstock.com

Ein Cyber Risk Assessment unterstützt dabei, potenzielle Bedrohungen und Schwachstellen für wichtige digitale und physische Unternehmens-Assets zu identifizieren, zu bewerten und zu priorisieren. Trotzdem stolpern in diesem Zusammenhang immer noch viele CISOs und Sicherheitsentscheider über Fallstricke, die sie daran hindern, ihre Risk-Assessment-Ziele vollumfänglich zu erreichen.

Welche das konkret sind und wie man sie gewissenhaft meidet, haben wir im Gespräch mit Security-Experten herausgefunden.

1. Einfach nur abhaken

Die wohl größte Falle im Zusammenhang mit Risk Assessments besteht darin, diese als Checkliste zu behandeln – statt als Entscheidungshilfe, die mit realem Business Impact oder Threat-Szenarien verknüpft ist. Shirsendu Mondal, Security-Forscher an der University of North Carolina, klärt auf: „Wenn sich Ihre Risikobewertung nur noch darum dreht, irgendwelche Kästchen abzuhaken, verlieren Sie die Fähigkeit, die tatsächlichen Risiken einer Umgebung zu Tage zu fördern. Das Ziel eines solchen Assessments sollte jedoch sein, aufzudecken, an welchen Stellen tatsächlich eine Gefährdungslage besteht.“ 

Der beste Weg, diese „Selbstzufriedenheits“-Falle zu umgehen, besteht laut dem Forscher darin, einen kontextorientierten Ansatz zu fahren: „Fragen Sie konkret danach, wo sich die betreffende Ressource befindet, wer darauf zugreifen kann, welche Daten sie berührt, wie wichtig sie für den Betrieb ist und was passiert, wenn sie ausfällt. Risiken sollten stets mit den geschäftlichen Auswirkungen korreliert werden – nicht bloß mit technischen Erkenntnissen.“

Eben, weil Risiken seiner Ansicht nach mehr sind als nur technische Probleme, empfiehlt Mondal Security-Entscheidern, andere Führungskräfte aus dem Unternehmen in das Security-Gefüge zu integrieren – etwa aus der IT und dem Betrieb.

2. Ergebnisse schönreden

Besonders in schwierigen Zeiten ist es das A und O, den Stakeholdern (und sich selbst) gegenüber ehrlich zu sein. Diese Auffassung vertritt auch Pablo Riboldi, CISO beim Softwareunternehmen BairesDev: „Wenn die Ergebnisse entmutigend sind, sollte man einfach zugeben, dass sich die Bedrohungslage deutlich schneller entwickelt hat, als über das bisherige Bewertungs-Framework abzusehen war.“

Anstatt einfach nur Schwachstellen-Listen zu übergeben, rät Riboldi dazu, konkrete Angriffsszenarien abzubilden: „Zum Beispiel, indem Sie die drei kritischsten Assets priorisieren und ein eingehendes Risk Assessment durchführen. So lässt sich auch ein unmittelbarer Mehrwert demonstrieren.“

3. Scope falsch einschätzen

Nicht wenige CISOs sichern Dokumentenkontrollen ab, haken Compliance-Checkboxen ab und erstellen ein Risikoregister, das den Eindruck vermittelt, dass alles in Ordnung ist. Der Schein trügt jedoch des Öfteren, wie Denis Calderone, CTO beim Sicherheitsdienstleister Suzu Labs, aus eigener Erfahrung weiß: „In solchen Fällen kommt es nicht selten vor, dass sich niemand die Mühe gemacht hat, zu testen, ob diese Kontrollen tatsächlich funktionieren. Oder, ob der Scope der Risikobewertung auch das abdeckt, worauf es wirklich ankommt.“

Der Technologieentscheider hat dazu auch ein Beispiel aus der Praxis auf Lager: „Wenn das Risk Assessment die Produktionsserver und das Unternehmensnetzwerk umfasst, der alte Dev-Rechner, ein Drittanbieter-Portal oder ein verwaister API-Endpunkt dabei aber außen vor bleiben, ist das ungünstig. Angreifer betrachten die gesamte Umgebung und finden genau den Einstiegspunkt, der zuvor als nicht bewertungswürdig erachtet wurde.“

Künstliche Intelligenz (KI) verschlimmere die Situation laut Calderone noch: Unternehmen setzten vielfach KI-Tools ein, verknüpften diese mit internen Systemen und gewährten ihnen Zugriff auf sensible Daten – ohne dass das in die Risikobewertung einfließe. Der Experte warnt: „Wenn Ihr Risk Assessment aufgesetzt wurde, bevor Ihr Unternehmen damit begonnen hat, KI in Workflows zu integrieren, ist es bereits veraltet.“

4. Annahmen nicht hinterfragen

Wenn sich die Zielsetzung einer Risikobewertung in Richtung „Hauptsache bestanden“ verschiebt, stellt das vielleicht Auditoren zufrieden. Die Unternehmensleitung könnte dadurch jedoch in die Irre geführt werden, wie Amit Basu, CIO und CISO beim Schifffahrtsunternehmen International Seaways, erklärt: „Führungskräfte und Vorstandsmitglieder sehen ein fertiges Risikoregister und gehen davon aus, dass das Unternehmen geschützt ist. Unterdessen bleiben echte Bedrohungen unberücksichtigt, weil sie nicht nahtlos in den Bewertungsrahmen passten. Dieser Fallstrick ist unsichtbar – er verbirgt sich hinter einem Dashboard.“

Nach Ansicht von Basu ist ein Risk Assessment nur so gut, wie die ihm zugrundeliegenden Annahmen: „Diese sollten Sie explizit dokumentieren und immer dann überprüfen, wenn sich das Business verändert, eine Bedrohungslage verschiebt oder ein Sicherheitsvorfall eine Lücke zu Tage fördert.“

Ein Risk Assessment, so der CISO, sei nicht als fertiges Produkt zu betrachten, sondern als lebendiger Beitrag zu einem fortlaufenden Dialog zwischen Security-Abteilung und Unternehmen.

5. Risiken nicht mit Impact verknüpfen

Probleme in den Hintergrund zu rücken oder herunterzuspielen, fällt deutlich leichter, wenn man den Zusammenhang zwischen Risiko und Business einfach ausblendet. Das erkennt auch Dan Moore, Senior Director of Strategy and Identity Standards beim CIAM-Spezialisten FusionAuth, an. Er warnt jedoch vor den Folgen dieses Gebarens: „So wird es sich diffizil gestalten, tatsächliche Risiken zu kommunizieren. Schlimmer noch: Es liefert den Mitgliedern des Security-Teams einen Vorwand, sich darüber zu beschweren, dass sie missverstanden oder nicht wertgeschätzt werden – und das beeinträchtigt die Effektivität des Teams.“

Der Manager erachtet es als wichtig, stattdessen konkret zu sein und zielgerichtet vorzugehen: „Verzichten Sie auf Angaben wie eine Patch-Compliance von 95 Prozent. Sprechen Sie stattdessen über das Risiko, das nicht gepatchte Systeme für das Unternehmen darstellen.“

Dabei seien manchen Systemen – etwa Legacy-Konstrukten, die nicht mit dem Internet verbunden sind – geringere Risiken inhärent als anderen, selbst wenn sie dieselben Patch-Probleme aufwiesen, meint Moore und empfiehlt, diese Tatsache anzuerkennen und die Reaktion entsprechend abzuwägen.  

6. Compliance mit Security verwechseln

„Compliance allein ist weder ein Garant für robuste Security, noch erfüllt sie die Mindestanforderungen für einen wirksamen Schutz“, hält Adriel Desautels, CEO der Security-Beratung Netragard, fest.

Unternehmen gerieten demnach besonders oft in diese Falle, wenn sie für Penetrationstests externe Firmen beauftragten, die sich auf Compliance konzentrieren und gleichzeitig „erstklassige Dienstleistungen“ versprechen. „In Wahrheit liefern diese oft automatisierte Scans, die als manuelle Tests getarnt sind“, meint Desautels.

Das Ergebnis sei ein falsches Sicherheitsgefühl, warnt der Manager: „Vergegenwärtigen Sie sich einfach, dass bei jedem größeren Sicherheitsvorfall der letzten zehn Jahre eine Organisation beteiligt war, die zum Zeitpunkt des Angriffs alle Compliance-Vorgaben erfüllt hatte.“

7. Risiken nicht vollständig verstehen

Unternehmen betrachten Risk Assessments oft als eine Art „Schwachstellenkatalogisierung“, bei der es darum geht, Lücken zu finden, Schweregrade zu erfassen und Audits zu bestehen. Letzteres heißt allerdings nicht, dass die Risiken auch verstanden wurden.

Geht es nach Safi Raza, Senior Director for Cybersecurity bei Fusion Risk Management, sollten sich CISOs darauf konzentrieren, technische Risikosignale mit betrieblichen Folgen zu verknüpfen: „Dazu muss man verstehen, welche Services betroffen sind, wie sich Störungen ausbreiten und was das für den Umsatz, die Kunden oder regulatorische Verpflichtungen bedeutet.“

Der Experte rät in diesem Zusammenhang dazu, zunächst von statischen Bewertungen zu einer kontinuierlichen, kontextbezogenen Risikotransparenz überzugehen, um sicherzustellen, dass Risiken nicht nur technisch verstanden werden.“ (fm)

Dieser Artikel ist im Original bei unserer Schwesterpublikation CSOonline.com erschienen.

Healthy software companies typically don’t suddenly eliminate one-fifth of their workforce, but monday.com is doing just that as it bets on flatter teams, AI agents, and customer implementation expertise as the winning combination in the AI era.

Monday.com co-founder and co-CEO Eran Zinman today announced the “very difficult decision” to reduce the AI work platform company’s global workforce by about 20%, or 620 people.

The move has nothing to do with increasing margins or replacing humans with AI, he insisted in his post on LinkedIn; rather, it’s a calculated decision to trim down and hone the company’s focus as AI becomes integral to day-to-day workflows.

“This is not a distress signal; it is a deliberate reset, disclosed with its price attached,” said Sanchit Vir Gogia, chief analyst at Greyhound Research. “The industry has quietly swapped the meaning of productivity, and this filing is the clearest exhibit yet.”

A ‘significant opportunity’ in technology

In a SEC filing this week, monday.com said its restructuring plan reflects the “ongoing transformation of its product, marketing, and go-to-market strategy.” The move is intended to support a “leaner, more focused operating model” as the company continues to invest in its AI-driven strategy.

Zinman noted in his post that the company has shifted to “doing the work with AI and not just managing it,” and is focused on building environments where “people and AI agents [work] together in one workspace.”

In recent months, monday.com has evolved its products, strategy, and the way it serves its customers, and Zinman contended that “the organization we built for our previous chapter is not the organization that fits the new AI era.” Monday.com needs to “execute more decisively,” take on new challenges, and quickly respond to market changes, he said.

“We have never seen such a significant opportunity in software, driven by such exciting technology,” Zinman noted. He emphasized that the reduction is not to replace people with AI, nor to improve margins; the “vast majority” of savings will be reinvested into talent, products, and AI.

The restructuring will result in a “flatter organization” with fewer management layers and smaller, more autonomous teams, and monday.com also has a new go-to-market model, Zinman explained. Customers expect “deeper implementation support” as they deploy AI, and the company will work more closely with customers, increase its on-site presence, create new roles, and “adapt many existing ones.” In its SEC filing, the company said it expects to continue hiring in “key strategic areas” throughout 2026.

Workers will be expected to work better, “not harder,” Zinman noted. He pointed to several past examples where work could have been done in a few days, but instead took many months with “multiple meetings and endless friction.”

“This wasn’t people’s fault and everyone was frustrated by this,” he said. “Our new org changes ownership to allow people to make decisions and move fast.”

A spokesperson for monday.com declined to comment further on the staff reductions.

Monday.com’s key market advantages

Monday.com certainly isn’t struggling; the company expects 19% to 20% year-over-year growth in 2026.

“Companies in that position do not restructure because they must,” Greyhound’s Gogia noted. “They restructure because they have decided to become something else.”

Melody Brue, VP and principal analyst at Moor Insights & Strategy, pointed out that organizational redesign is important for real AI transformation, but while it can signal confidence to the market, it can still be “devastating” to humans.

While the company looks as though it’s trying to do right, that ultimately remains to be seen, she said. “There are often hidden internal bruises that can surface long after layoffs.”

Monday.com’s advantage is in its “structured substrate,” Gogia noted; its boards, permissions and typed workflows give agents something firmer to act on than just documents and chat history. The company highlights its natively built agents that can be configured by any team member, as well as connectors with Claude, Microsoft Copilot, and ChatGPT, and dedicated routes for external agents to authenticate and operate.

“For some time, the sharper enterprise question has been shifting from who has an agent to who owns the governed runtime in which an agent can safely act,” he said. “Structured work is a serious claim on that runtime.”

But parts of monday.com’s agent estate remain in staged release, and its product is ultimately “mid-transition,” Gogia pointed out; its agent builder carried a beta label as recently as March,. Also, the company’s pricing model changed in May to a hybrid model charging for seats as well as mandatory AI credits. And, while its AI-powered no-code builder monday vibe passed $1 million in annual recurring revenue within two and a half months, monday.com has not released subsequent outcomes, usage volumes, or attach rates.

Further, there’s an element of “gravity” with its competitors, he observed. Asana is reorganizing teams around agents, Atlassian is wiring agents into the developer estate, and others are simply bundling them into their offerings: Microsoft is doing so across the productivity stack, and ServiceNow across enterprise operations, each with identity and procurement built in.

“Their pull is strongest exactly where monday.com wants to grow, in the largest accounts, where control-plane depth and administrative reach decide the deal,” said Gogia.

Actions for the near-term

Going forward, buyers should focus on operating risk, not headline risk, Moor’s Brue noted. In practice, that’s continuity of service, roadmap consistency, and strength of enterprise support. Productivity should be valued as better outcomes per unit of organizational effort, not mere activity.

“It should be a measure of how much smoother, faster, and more effective the operating model becomes when AI is built into the work,” said Brue.

Gogia noted that strain surfaces first in customer service, and monday.com’s attention is being redistributed. The company’s annual report disclosed that its focus is now concentrated on the largest accounts, with support for medium-sized clients moved to an AI-first and human-supported model.

During the first month of the transition, buyers should track named account continuity and escalation times, he advised. By the first quarter, keep an eye on whether credit governance and admin controls mature on schedule, and if the roadmap beyond the AI estate keeps pace. By the half-year mark, determine whether promised implementation depth is producing outcomes or “simply more billable engagement.”

Support tiers should be enumerated in writing before renewal, and buyers should contract for “side exits,” Gogia emphasized, with overage pricing fixed in advance, the right to pause consumption, and portability for workflows and agent configuration “if the relationship sours.” Finance should also insist on monthly consumption reporting by capability. Further, integration efforts, partner dependency, and change management should be considered first-class costs of the agent era, “not as afterthoughts to a license.”

“A license was a known cost,” said Gogia. “A meter is a behavior, and behavior is harder to forecast than headcount.”

This article originally appeared on CIO.com.

Healthy software companies typically don’t suddenly eliminate one-fifth of their workforce, but monday.com is doing just that as it bets on flatter teams, AI agents, and customer implementation expertise as the winning combination in the AI era.

Monday.com co-founder and co-CEO Eran Zinman today announced the “very difficult decision” to reduce the AI work platform company’s global workforce by about 20%, or 620 people.

The move has nothing to do with increasing margins or replacing humans with AI, he insisted in his post on LinkedIn; rather, it’s a calculated decision to trim down and hone the company’s focus as AI becomes integral to day-to-day workflows.

“This is not a distress signal; it is a deliberate reset, disclosed with its price attached,” said Sanchit Vir Gogia, chief analyst at Greyhound Research. “The industry has quietly swapped the meaning of productivity, and this filing is the clearest exhibit yet.”

A ‘significant opportunity’ in technology

In a SEC filing this week, monday.com said its restructuring plan reflects the “ongoing transformation of its product, marketing, and go-to-market strategy.” The move is intended to support a “leaner, more focused operating model” as the company continues to invest in its AI-driven strategy.

Zinman noted in his post that the company has shifted to “doing the work with AI and not just managing it,” and is focused on building environments where “people and AI agents [work] together in one workspace.”

In recent months, monday.com has evolved its products, strategy, and the way it serves its customers, and Zinman contended that “the organization we built for our previous chapter is not the organization that fits the new AI era.” Monday.com needs to “execute more decisively,” take on new challenges, and quickly respond to market changes, he said.

“We have never seen such a significant opportunity in software, driven by such exciting technology,” Zinman noted. He emphasized that the reduction is not to replace people with AI, nor to improve margins; the “vast majority” of savings will be reinvested into talent, products, and AI.

The restructuring will result in a “flatter organization” with fewer management layers and smaller, more autonomous teams, and monday.com also has a new go-to-market model, Zinman explained. Customers expect “deeper implementation support” as they deploy AI, and the company will work more closely with customers, increase its on-site presence, create new roles, and “adapt many existing ones.” In its SEC filing, the company said it expects to continue hiring in “key strategic areas” throughout 2026.

Workers will be expected to work better, “not harder,” Zinman noted. He pointed to several past examples where work could have been done in a few days, but instead took many months with “multiple meetings and endless friction.”

“This wasn’t people’s fault and everyone was frustrated by this,” he said. “Our new org changes ownership to allow people to make decisions and move fast.”

A spokesperson for monday.com declined to comment further on the staff reductions.

Monday’s key market advantages

Monday.com certainly isn’t struggling; the company expects 19% to 20% year-over-year growth in 2026.

“Companies in that position do not restructure because they must,” Greyhound’s Gogia noted. “They restructure because they have decided to become something else.”

Melody Brue, VP and principal analyst at Moor Insights & Strategy, pointed out that organizational redesign is important for real AI transformation, but while it can signal confidence to the market, it can still be “devastating” to humans.

While the company looks as though it’s trying to do right, that ultimately remains to be seen, she said. “There are often hidden internal bruises that can surface long after layoffs.”

Monday.com’s advantage is in its “structured substrate,” Gogia noted; its boards, permissions and typed workflows give agents something firmer to act on than just documents and chat history. The company highlights its natively built agents that can be configured by any team member, as well as connectors with Claude, Microsoft Copilot, and ChatGPT, and dedicated routes for external agents to authenticate and operate.

“For some time, the sharper enterprise question has been shifting from who has an agent to who owns the governed runtime in which an agent can safely act,” he said. “Structured work is a serious claim on that runtime.”

But parts of monday.com’s agent estate remain in staged release, and its product is ultimately “mid-transition,” Gogia pointed out; its agent builder carried a beta label as recently as March,. Also, the company’s pricing model changed in May to a hybrid model charging for seats as well as mandatory AI credits. And, while its AI-powered no-code builder monday vibe passed $1 million in annual recurring revenue within two and a half months, monday.com has not released subsequent outcomes, usage volumes, or attach rates.

Further, there’s an element of “gravity” with its competitors, he observed. Asana is reorganizing teams around agents, Atlassian is wiring agents into the developer estate, and others are simply bundling them into their offerings: Microsoft is doing so across the productivity stack, and ServiceNow across enterprise operations, each with identity and procurement built in.

“Their pull is strongest exactly where monday.com wants to grow, in the largest accounts, where control-plane depth and administrative reach decide the deal,” said Gogia.

Actions for the near-term

Going forward, buyers should focus on operating risk, not headline risk, Moor’s Brue noted. In practice, that’s continuity of service, roadmap consistency, and strength of enterprise support. Productivity should be valued as better outcomes per unit of organizational effort, not mere activity.

“It should be a measure of how much smoother, faster, and more effective the operating model becomes when AI is built into the work,” said Brue.

Gogia noted that strain surfaces first in customer service, and monday.com’s attention is being redistributed. The company’s annual report disclosed that its focus is now concentrated on the largest accounts, with support for medium-sized clients moved to an AI-first and human-supported model.

During the first month of the transition, buyers should track named account continuity and escalation times, he advised. By the first quarter, keep an eye on whether credit governance and admin controls mature on schedule, and if the roadmap beyond the AI estate keeps pace. By the half-year mark, determine whether promised implementation depth is producing outcomes or “simply more billable engagement.”

Support tiers should be enumerated in writing before renewal, and buyers should contract for “side exits,” Gogia emphasized, with overage pricing fixed in advance, the right to pause consumption, and portability for workflows and agent configuration “if the relationship sours.” Finance should also insist on monthly consumption reporting by capability. Further, integration efforts, partner dependency, and change management should be considered first-class costs of the agent era, “not as afterthoughts to a license.”

“A license was a known cost,” said Gogia. “A meter is a behavior, and behavior is harder to forecast than headcount.”

Oracle is expanding its Cloud@Customer on-premises portfolio with a managed database offering that it says will enable enterprises to run databases, applications, and AI agents in their own data centers, helping CIOs modernize mid-sized workloads while meeting data residency, regulatory, and low-latency requirements.

The hybrid cloud offering, Base Database Cloud@Customer, combines existing database and infrastructure services such as the Base Database Service and Data Infrastructure Cloud@Customer X11 platform. It is designed for enterprises that do not need the scale of Exadata Cloud@Customer but still want their infrastructure and AI capabilities on-premises, managed by Oracle, the company said.

The Cloud@Customer X11 platform itself consists of two Oracle X11 compute servers and shared all-flash storage, offering up to 60 usable processor cores and 660 GB of memory per server, 47.2 TB of storage, and 10/25 GbE networking.

For regulated industries or restricted connectivity

Analysts see the new offering filling a gap for enterprises that want the operational and economic benefits of the cloud but cannot send their data to a public cloud because of legal restrictions or technology limitations.

These enterprises, according to Ashish Chaturvedi, executive research leader at HFS Research, are likely to be in regulated industries such as financial services, healthcare, government, and defense that must comply with data residency requirements, or needing low-latency access from remote sites to operational databases.

The offering could also appeal to enterprises modernizing mid-sized workloads at remote locations or within individual business units that could never justify the investment in a full Exadata rack, said Amit Chandak, chief analytics officer at IT consulting firm Kanerika.

In all cases, Chaturvedi said, the appeal of the offering is its managed nature, which takes away the burden of looking after the underlying infrastructure.

Deployment and maintenance becomes easier too, said Michael Leone, principal analyst at Moor Strategy and Insights: “They get automation that mid-size teams rarely have the staff to build. Clustering, patching, standby databases, and backups arrive configured instead of hand-assembled because the offering is managed.”

The economics are equally compelling, Chaturvedi said. The pay-as-you-go pricing model, combined with online compute scaling, helps enterprises avoid overprovisioning and paying license fees for idle cores, which is a “classic waste” of fixed on-premises systems, he said.

Private AI behind the firewall

Beyond the operational and economic benefits, the architecture of the new offering enables databases, applications, VMs, and AI agents to be collocated on the same platform, removing what Chaturvedi called “the single biggest blocker” to AI adoption in regulated environments: the need to keep private data behind the firewall.

“For a CIO in a regulated sector who wants to deploy AI agents but can’t let regulated data touch an external model API, that’s a real unlock,” Chaturvedi said.

More so because most AI offerings, at least in their present form and state, cannot guarantee sensitive data protection, said Igor Ikonnikov, advisory fellow at Info-Tech Research Group.

Even if Base Database Cloud@Customer turns out more expensive than fully cloud-based options, “It’s still attractive as it eliminates reputational and economic risk caused by possible AI-induced data leakage,” Ikonnikov said.

The offering’s consolidation of databases, applications, and AI agents will also simplify deployment of AI-based workflows, said Forrester principal analyst Noel Yuhanna. “It reduces stack complexity and helps accelerate development cycles, deliver real-time data, and eliminate data movement challenges.”

Despite those advantages, Chandak cautioned that the offering is unlikely to see broad adoption outside Oracle’s existing customer base: “If a company isn’t already on Oracle, the pull is weak. You don’t buy into Oracle’s database just to get this.”

Enterprises seeking similar hybrid cloud capabilities have no shortage of alternatives: AWS, Microsoft, Google Cloud, IBM, Dell Technologies, and HPE all offer combinations of on-premises infrastructure, cloud management, and AI services.

However, those alternatives typically require customers to integrate multiple software and hardware components rather than consume them as a single managed offering.

Oracle’s differentiation, although narrow, is hard to match, Chaturvedi said: “The vertical integration of database, engineered hardware, cloud management, high-availability architecture, and now private AI, all engineered together and delivered as a managed on-prem subscription should be genuinely convenient and attractive.”

The offering is compatible with Oracle AI Database 26ai and Oracle Database 19c in Enterprise Edition and Standard Edition configurations. It also supports Oracle Real Application Clusters, Oracle Data Guard, and Zero Data Loss Recovery Appliance through Oracle-managed cloud automation for high availability and disaster recovery, the company said.

Base Database Cloud@Customer is now generally available, Oracle said. It did not provide pricing.

This article first appeared on CIO.

Oracle is expanding its Cloud@Customer on-premises portfolio with a managed database offering that it says will enable enterprises to run databases, applications, and AI agents in their own data centers, helping CIOs modernize mid-sized workloads while meeting data residency, regulatory, and low-latency requirements.

The hybrid cloud offering, Base Database Cloud@Customer, combines existing database and infrastructure services such as the Base Database Service and Data Infrastructure Cloud@Customer X11 platform. It is designed for enterprises that do not need the scale of Exadata Cloud@Customer but still want their infrastructure and AI capabilities on-premises, managed by Oracle, the company said.

The Cloud@Customer X11 platform itself consists of two Oracle X11 compute servers and shared all-flash storage, offering up to 60 usable processor cores and 660 GB of memory per server, 47.2 TB of storage, and 10/25 GbE networking.

For regulated industries or restricted connectivity

Analysts see the new offering filling a gap for enterprises that want the operational and economic benefits of the cloud but cannot send their data to a public cloud because of legal restrictions or technology limitations.

These enterprises, according to Ashish Chaturvedi, executive research leader at HFS Research, are likely to be in regulated industries such as financial services, healthcare, government, and defense that must comply with data residency requirements, or needing low-latency access from remote sites to operational databases.

The offering could also appeal to enterprises modernizing mid-sized workloads at remote locations or within individual business units that could never justify the investment in a full Exadata rack, said Amit Chandak, chief analytics officer at IT consulting firm Kanerika.

In all cases, Chaturvedi said, the appeal of the offering is its managed nature, which takes away the burden of looking after the underlying infrastructure.

Deployment and maintenance becomes easier too, said Michael Leone, principal analyst at Moor Strategy and Insights: “They get automation that mid-size teams rarely have the staff to build. Clustering, patching, standby databases, and backups arrive configured instead of hand-assembled because the offering is managed.”

The economics are equally compelling, Chaturvedi said. The pay-as-you-go pricing model, combined with online compute scaling, helps enterprises avoid overprovisioning and paying license fees for idle cores, which is a “classic waste” of fixed on-premises systems, he said.

Private AI behind the firewall

Beyond the operational and economic benefits, the architecture of the new offering enables databases, applications, VMs, and AI agents to be collocated on the same platform, removing what Chaturvedi called “the single biggest blocker” to AI adoption in regulated environments: the need to keep private data behind the firewall.

“For a CIO in a regulated sector who wants to deploy AI agents but can’t let regulated data touch an external model API, that’s a real unlock,” Chaturvedi said.

More so because most AI offerings, at least in their present form and state, cannot guarantee sensitive data protection, said Igor Ikonnikov, advisory fellow at Info-Tech Research Group.

Even if Base Database Cloud@Customer turns out more expensive than fully cloud-based options, “It’s still attractive as it eliminates reputational and economic risk caused by possible AI-induced data leakage,” Ikonnikov said.

The offering’s consolidation of databases, applications, and AI agents will also simplify deployment of AI-based workflows, said Forrester principal analyst Noel Yuhanna. “It reduces stack complexity and helps accelerate development cycles, deliver real-time data, and eliminate data movement challenges.”

Despite those advantages, Chandak cautioned that the offering is unlikely to see broad adoption outside Oracle’s existing customer base: “If a company isn’t already on Oracle, the pull is weak. You don’t buy into Oracle’s database just to get this.”

Enterprises seeking similar hybrid cloud capabilities have no shortage of alternatives: AWS, Microsoft, Google Cloud, IBM, Dell Technologies, and HPE all offer combinations of on-premises infrastructure, cloud management, and AI services.

However, those alternatives typically require customers to integrate multiple software and hardware components rather than consume them as a single managed offering.

Oracle’s differentiation, although narrow, is hard to match, Chaturvedi said: “The vertical integration of database, engineered hardware, cloud management, high-availability architecture, and now private AI, all engineered together and delivered as a managed on-prem subscription should be genuinely convenient and attractive.”

The offering is compatible with Oracle AI Database 26ai and Oracle Database 19c in Enterprise Edition and Standard Edition configurations. It also supports Oracle Real Application Clusters, Oracle Data Guard, and Zero Data Loss Recovery Appliance through Oracle-managed cloud automation for high availability and disaster recovery, the company said.

Base Database Cloud@Customer is now generally available, Oracle said. It did not provide pricing.

This article first appeared on CIO.

Oracle is expanding its Cloud@Customer on-premises portfolio with a managed database offering that it says will enable enterprises to run databases, applications, and AI agents in their own data centers, helping CIOs modernize mid-sized workloads while meeting data residency, regulatory, and low-latency requirements.

The hybrid cloud offering, Base Database Cloud@Customer, combines existing database and infrastructure services such as the Base Database Service and Data Infrastructure Cloud@Customer X11 platform. It is designed for enterprises that do not need the scale of Exadata Cloud@Customer but still want their infrastructure and AI capabilities on-premises, managed by Oracle, the company said.

The Cloud@Customer X11 platform itself consists of two Oracle X11 compute servers and shared all-flash storage, offering up to 60 usable processor cores and 660 GB of memory per server, 47.2 TB of storage, and 10/25 GbE networking.

For regulated industries or restricted connectivity

Analysts see the new offering filling a gap for enterprises that want the operational and economic benefits of the cloud but cannot send their data to a public cloud because of legal restrictions or technology limitations.

These enterprises, according to Ashish Chaturvedi, executive research leader at HFS Research, are likely to be in regulated industries such as financial services, healthcare, government, and defense that must comply with data residency requirements, or needing low-latency access from remote sites to operational databases.

The offering could also appeal to enterprises modernizing mid-sized workloads at remote locations or within individual business units that could never justify the investment in a full Exadata rack, said Amit Chandak, chief analytics officer at IT consulting firm Kanerika.

In all cases, Chaturvedi said, the appeal of the offering is its managed nature, which takes away the burden of looking after the underlying infrastructure.

Deployment and maintenance becomes easier too, said Michael Leone, principal analyst at Moor Strategy and Insights: “They get automation that mid-size teams rarely have the staff to build. Clustering, patching, standby databases, and backups arrive configured instead of hand-assembled because the offering is managed.”

The economics are equally compelling, Chaturvedi said. The pay-as-you-go pricing model, combined with online compute scaling, helps enterprises avoid overprovisioning and paying license fees for idle cores, which is a “classic waste” of fixed on-premises systems, he said.

Private AI behind the firewall

Beyond the operational and economic benefits, the architecture of the new offering enables databases, applications, VMs, and AI agents to be collocated on the same platform, removing what Chaturvedi called “the single biggest blocker” to AI adoption in regulated environments: the need to keep private data behind the firewall.

“For a CIO in a regulated sector who wants to deploy AI agents but can’t let regulated data touch an external model API, that’s a real unlock,” Chaturvedi said.

More so because most AI offerings, at least in their present form and state, cannot guarantee sensitive data protection, said Igor Ikonnikov, advisory fellow at Info-Tech Research Group.

Even if Base Database Cloud@Customer turns out more expensive than fully cloud-based options, “It’s still attractive as it eliminates reputational and economic risk caused by possible AI-induced data leakage,” Ikonnikov said.

The offering’s consolidation of databases, applications, and AI agents will also simplify deployment of AI-based workflows, said Forrester principal analyst Noel Yuhanna. “It reduces stack complexity and helps accelerate development cycles, deliver real-time data, and eliminate data movement challenges.”

Despite those advantages, Chandak cautioned that the offering is unlikely to see broad adoption outside Oracle’s existing customer base: “If a company isn’t already on Oracle, the pull is weak. You don’t buy into Oracle’s database just to get this.”

Enterprises seeking similar hybrid cloud capabilities have no shortage of alternatives: AWS, Microsoft, Google Cloud, IBM, Dell Technologies, and HPE all offer combinations of on-premises infrastructure, cloud management, and AI services.

However, those alternatives typically require customers to integrate multiple software and hardware components rather than consume them as a single managed offering.

Oracle’s differentiation, although narrow, is hard to match, Chaturvedi said: “The vertical integration of database, engineered hardware, cloud management, high-availability architecture, and now private AI, all engineered together and delivered as a managed on-prem subscription should be genuinely convenient and attractive.”

The offering is compatible with Oracle AI Database 26ai and Oracle Database 19c in Enterprise Edition and Standard Edition configurations. It also supports Oracle Real Application Clusters, Oracle Data Guard, and Zero Data Loss Recovery Appliance through Oracle-managed cloud automation for high availability and disaster recovery, the company said.

Base Database Cloud@Customer is now generally available, Oracle said. It did not provide pricing.

Cisco has revealed a family of open-weight AI models called Antares that, it said, can help security teams isolate potentially vulnerable parts of a software repository before deeper investigation begins.

Rather than detecting a specific CVE or generating a patch, these models search a codebase using only a Common Weakness Enumeration (CWE) description and return the files most likely to contain that class of vulnerability.

“Its purpose is to reduce a large codebase to a focused set of files that a security professional or a downstream security workflow should investigate,” Cisco’s AI researcher Supriti Vijay said via email. “The goal is not to replace a security engineer’s judgement or send them on a wild-goose chase, but to reduce fatigue and workload by helping them triage an issue earlier and focus their investigation on the most relevant parts of the codebase.”

The Antares family consists of models with 350 million, 1 billion, and 3 billion parameters trained specifically for repository-scale vulnerability localization.

The company said its largest model approaches the performance of GPT-5.5 on its internal vulnerability localization (Vloc) benchmark while remaining small enough for low-cost local deployment.

A search assistant, not a vulnerability detector

Cisco is careful to define what Antares is, and what it is not.

“Antares outputs a ranked list of source files likely to contain a relevant vulnerability, along with the terminal exploration trace that led to that result,” Cisco Foundation AI Chief Scientist Amin Karbasi wrote in a blog post, adding that the models are not meant to replace the broader application security toolchain: Human analysts or downstream security tools will still be needed to confirm exploitability, identify vulnerable lines of code, assess severity and generate fixes.

Antares differs from conventional static analysis platforms such as Semgrep or CodeQL, which primarily rely on predefined rules or queries. Cisco instead describes Antares as an evidence-driven exploration agent that adapts its search as it traverses the repository.

Cisco’s argument is that large repositories often contain thousands of files, making manual reviews exhaustive and unrealistic. By reducing the search space to a manageable shortlist, the company hopes to reduce investigation fatigue without replacing human judgement.

Claims of specialization over scale

Cisco is also making a statement about how cybersecurity models should evolve.

Instead of pursuing larger foundational models, Cisco argued that specialized, task-trained models can outperform much larger open-weight alternatives for vulnerability localization. In its evaluation Antares-3B, the largest model intended for single-GPU deployments, produced results comparable to GPT-5.5 while outperforming several substantially larger open models by Google, OpenAI and Meta.

The family also includes Antares-350M for resource-constrained environments and Antares-1B for laptops and workstations, which Cisco has made available as open-weight models on Hugging Face.

The command line interface (CLI) on the models supports targeted CWE investigations, repository-wide scans, SARIF output and local inference, which Cisco said enables organizations to keep proprietary code inside their own trust boundary.

However, because Antares identifies candidate files rather than confirmed vulnerabilities, organizations will still need to understand how often such repository-wide searches should be run, how much they improve existing triage workflows, and whether the reduction in investigation effort ultimately translates into measurable security or cost benefits.

Cisco has revealed a family of open-weight AI models called Antares that, it said, can help security teams isolate potentially vulnerable parts of a software repository before deeper investigation begins.

Rather than detecting a specific CVE or generating a patch, these models search a codebase using only a Common Weakness Enumeration (CWE) description and return the files most likely to contain that class of vulnerability.

“Its purpose is to reduce a large codebase to a focused set of files that a security professional or a downstream security workflow should investigate,” Cisco’s AI researcher Supriti Vijay said via email. “The goal is not to replace a security engineer’s judgement or send them on a wild-goose chase, but to reduce fatigue and workload by helping them triage an issue earlier and focus their investigation on the most relevant parts of the codebase.”

The Antares family consists of models with 350 million, 1 billion, and 3 billion parameters trained specifically for repository-scale vulnerability localization.

The company said its largest model approaches the performance of GPT-5.5 on its internal vulnerability localization (Vloc) benchmark while remaining small enough for low-cost local deployment.

A search assistant, not a vulnerability detector

Cisco is careful to define what Antares is, and what it is not.

“Antares outputs a ranked list of source files likely to contain a relevant vulnerability, along with the terminal exploration trace that led to that result,” Cisco Foundation AI Chief Scientist Amin Karbasi wrote in a blog post, adding that the models are not meant to replace the broader application security toolchain: Human analysts or downstream security tools will still be needed to confirm exploitability, identify vulnerable lines of code, assess severity and generate fixes.

Antares differs from conventional static analysis platforms such as Semgrep or CodeQL, which primarily rely on predefined rules or queries. Cisco instead describes Antares as an evidence-driven exploration agent that adapts its search as it traverses the repository.

Cisco’s argument is that large repositories often contain thousands of files, making manual reviews exhaustive and unrealistic. By reducing the search space to a manageable shortlist, the company hopes to reduce investigation fatigue without replacing human judgement.

Claims of specialization over scale

Cisco is also making a statement about how cybersecurity models should evolve.

Instead of pursuing larger foundational models, Cisco argued that specialized, task-trained models can outperform much larger open-weight alternatives for vulnerability localization. In its evaluation Antares-3B, the largest model intended for single-GPU deployments, produced results comparable to GPT-5.5 while outperforming several substantially larger open models by Google, OpenAI and Meta.

The family also includes Antares-350M for resource-constrained environments and Antares-1B for laptops and workstations, which Cisco has made available as open-weight models on Hugging Face.

The command line interface (CLI) on the models supports targeted CWE investigations, repository-wide scans, SARIF output and local inference, which Cisco said enables organizations to keep proprietary code inside their own trust boundary.

However, because Antares identifies candidate files rather than confirmed vulnerabilities, organizations will still need to understand how often such repository-wide searches should be run, how much they improve existing triage workflows, and whether the reduction in investigation effort ultimately translates into measurable security or cost benefits.

This article first appeared on InfoWorld.

GitLab has released GitLab 19.2, an update to the company’s devsecops platform that allows teams to fix vulnerable dependencies automatically, use Security Review Flow to catch logic flaws that scanners miss, and run AI agents straight from the terminal, the company said.

Highlights in GitLab 19.2 include the following:

• Dependency Scanning Auto-Remediation, in public beta, uses AI to fix build-breaking changes and iterates until your pipeline passes, with every change governed by your existing gates and audit trail.
• Security Review Flow, also in public beta, analyzes code changes as a security engineer would and catches authorization gaps, business-logic errors, and race conditions that static scanners structurally cannot see.
• GitLab Duo CLI, now generally available, gives developers access to agents and multi-step agentic flows for all software life cycle tasks without leaving the terminal. 
• Custom Flows, now generally available, let teams replace manual multi-step workflows with agentic automations for software development, triggered by GitLab events.

“Coding agents made it possible to generate far more code and moved the bottleneck downstream to reviews and security,” said Manav Khurana, chief product and marketing officer at GitLab, in a statement. “GitLab 19.2 puts agents to work on that bottleneck: fixing vulnerable dependencies, catching the flaws scanners miss, and automating the steps in between with a person still approving what ships.”

Across industries, a version of the same conversation is playing out in technology leadership meetings. Enterprises have deployed AI broadly, and foundation models keep getting more capable. Yet the outputs still feel generic, shaped by industry patterns rather than by the organization producing them.

McKinsey’s AI Trust Maturity Survey found that while overall AI maturity scores have improved, only about a third of organizations have reached a mature level of strategy and governance. Technical capability is advancing faster than organizational alignment. In my view, the gap is not a model problem. It is a context problem. Enterprises are feeding generic inputs into powerful models because sharing organizational context seamlessly with AI is neither easy nor intuitive today.

Building the analytical and creative capabilities to scale AI, something I explored in a recent piece on the left-brain and right-brain approach to enterprise AI, is necessary but not sufficient. Before either can function effectively, the enterprise needs something more fundamental. AI that actually understands the contextual fabric of the organization it is operating in. A frontier model has processed everything written about your sector, your competitors and your regulatory landscape. It cannot access the reasoning embedded in years of delivery decisions, the patterns encoded in how your teams scope and deliver work over time. That knowledge is organizational memory, and frontier models can’t get that easily. It exists inside every enterprise but has never been structured, connected or made available to any AI system. Without it, even the most capable model answers a generic version of your question.

The next competitive advantage in enterprise AI will not come from a better model. It will come from a better organizational context.

One global technology enterprise set out to solve this across its own operations, building a modular ecosystem of domain-specific agents grounded in its own data across contracting, talent and vendor management workflows. What emerged was not just operational efficiency but a shared intelligence layer connecting decisions across functions for the first time.

Competitive differentiation was never about the tools

Consider what actually separates high-performing enterprises from the rest. In a regulated industry like financial services or healthcare, organizations cannot meaningfully differentiate on product. A bank cannot offer substantially different products or services. A health system uses the same clinical protocols and the same electronic health record (EHR) platforms as its peers. What varies is everything underneath: the rigor of processes, the coherence of cross-functional decisions and the people who carry years of accumulated organizational judgment in how they make those decisions.

An organization with a proper context layer in place can say with precision that for this type of engagement, in this sector, with this risk profile, our institutional history tells us exactly where we stand. That level of specificity is what most enterprises have never made available to AI.

The enterprise AI brain that every organization has but has never assembled

Every enterprise already possesses what I think of as an enterprise AI brain. The problem is that it has never been assembled in one place. The data exists across contracts, project documentation, talent records, delivery metrics and the operational communications of daily execution — the informal reasoning that rarely makes it into formal systems.

None of the standard enterprise platforms were designed to connect this. A customer relationship management (CRM) system captures customer interactions. An enterprise resource planning (ERP) system captures transactions. A project management tool captures tasks and timelines. None of them captures the reasoning behind decisions and none of them surfaces a coherent picture of how the organization actually thinks and operates.

BCG’s study across hundreds of companies found that only 10% of AI value comes from the algorithms and another 20% from the technology that implements them, meaning the remaining 70% depends on people, processes and organizational change. The organizations extracting real value are those that have made their institutional knowledge available to AI in a structured, governed way.

Building a contextual AI fabric

A Contextual AI Fabric is the technical and organizational layer that makes the Enterprise AI Brain usable. It brings together unstructured data ingestion, semantic structuring, retrieval pipelines and governed model access to give AI systems the organizational context they need to produce outputs that are genuinely specific to your enterprise rather than generically accurate about your industry. It rests on three pillars. Core is the secure, governed and interoperable foundation that AI operations run on. Context is reliable, traceable access to the organization’s data, processes, knowledge and history. Coordination connects people, agents, applications and systems into process-driven workflows with clear controls and accountability, so the organization acts as one rather than a set of disconnected functions.

The data layer is where most organizations underestimate the work. Contracts, project reports, talent assessments and operational communications require extraction, chunking, embedding and indexing before a model can retrieve and reason over them meaningfully.

The semantic layer is what makes retrieval meaningful. Even well-ingested data fails if functions use different terminology for the same concepts. What legal calls a contract, delivery calls a scope. Without a shared ontology, AI systems remain precise about the wrong thing. And retrieval alone, however well-structured, only takes an organization so far. Retrieval surfaces the right information at the moment of a query, but it does not give a model genuine memory of the organization. The real source of unique, organization-level relevance comes from training domain-specific small language models on this context directly, models that carry organizational memory forward rather than fetching it fresh every time. That is what ultimately separates a Contextual AI Fabric from a well-organized database.

The governance layer is not an add-on. Access controls, data lineage, approval thresholds and human checkpoints need to be designed in before any agent goes into production. Security is not a layer you add afterward. It is the condition under which organizational AI is worth building. If the institutional intelligence that makes your enterprise distinct gets absorbed into a frontier model’s training data, it becomes everyone’s baseline. That is an architectural decision made, or avoided, at the point of deployment.

Proprietary by design

The institutional knowledge that makes up a contextual AI fabric — delivery history, commercial patterns, talent intelligence and operating culture — is proprietary in ways no external model can replicate. This is as much a security imperative as it is a competitive one. Organizational context, once exposed, cannot be unexposed.

Most enterprises are using AI to automate existing processes rather than questioning whether those processes should be redesigned entirely. The organizations extracting the most value are those willing to ask whether their current operating model, built before GenAI existed, is the one they would build today. That question is harder than any technology decision, and it is also the most consequential one.

From context to coordinated action

Context alone is not enough. When a delivery risk surfaces in project data, the talent function needs to respond. When a commercial signal changes in contract data, operations need to recalibrate. This kind of cross-functional coordination, driven by shared organizational intelligence rather than siloed data, is where the real value of enterprise AI shows up and where the absence of a shared context layer becomes most visible.

A global leader in digital payments and business services found its AI deployments across payroll, HR and risk compliance, each running in isolation, with no shared governance or common data foundation. Once the organization established a unified governance backbone connecting its operational data through a shared retrieval layer, business users could query across domains in plain language and new use cases across fraud analytics, forecasting and policy extraction became extensible without rebuilding infrastructure for each one. The shift was not in the models. It was in the shared foundation underneath them.

The leadership question behind the technology question

The enterprises pulling ahead in AI are not winning on model quality but on organizational memory. The ones that have done the hard work of structuring their institutional knowledge into a governed, secure Contextual AI Fabric are giving their AI something no competitor can replicate: the accumulated intelligence of how the business actually operates.

For CIOs, the question is no longer which model to deploy. It is whether the organization has built the foundation that would make any model worth deploying.

This article is published as part of the Foundry Expert Contributor Network.
Want to join?

Your CFO has a simple question. “We’re spending more on AI. What are we getting for it?” Most CIOs cannot answer it — not because AI isn’t creating value, but because the accounting systems we inherited were built before AI existed as a category of labor.

This June, the conversation shifted from token maxing to token cutting. The New York Times reported that Meta, Uber, Walmart and Amazon are capping employee AI usage. Uber blew through its 2026 AI budget in four months. Satya Nadella started framing it as human capital versus token capital.

All of that is true. None of it answers the CFO. Capping tokens is an input lever, not an output measure. And the human-versus-token framing names two sources of labor when the reality is four.

The enterprise now has 4 sources of labor

There are humans. There are humans assisted by AI. Humans are working alongside AI. And humans are managing AI. Sources two through four are all supervised machine labor at different intensities — none of them have a line item, a manager or an hourly rate. In our 2026 AI Labor Report, 78% of leaders view AI as both software and a labor force. The org chart has not caught up. Neither has the P&L.

Lexi Reese

Most enterprises are stuck at A-Level 1 with no accounting for any of it, while quietly sliding into A-Level 2. The job descriptions have not caught up. The budget has not caught up. You cannot upskill into a role that has not been named.

AI is the only category of work the modern enterprise has ever bought without a system of record for what it produced.

What you are actually running is supervised machine labor

The model does a first pass. A human makes it usable. One hundred percent of leaders we surveyed said AI work requires human review before it ships; 34% said substantial editing. That is a workforce with no manager, no hourly rate and no line on the income statement.

The accounting breaks in 3 places at once

Under GAAP: COGS if it helps produce the product, OpEx if it does work for you. The same workflow can hit all three buckets at once. A tier-one support resolution involves the human’s salary (OpEx), the AI’s tokens (COGS if support is a delivered service), and the supervisor’s review time (OpEx). Three buckets. One piece of work. No reconciliation. The token invoice arrives from Anthropic or OpenAI and gets coded to OpEx-software because that is what the bill looks like. Audit partners will be asking about this by next year.

When you call AI a tool, you book it like software. When you call it labor, you have to ask which kind and what it is producing.

The per-employee number is the wrong unit

Per-employee AI spend collapses a workforce into a per-head average. It hides the only number that matters: What AI is producing inside each workflow.

Lanai measured two teams inside the same finance organization. Same monthly prep and variance analysis. AI took the same amount of time to produce outputs of similar quality. The only variable was the model each team reached for by default — a choice nobody had made deliberately and nobody had seen until it was measured.

Lexi Reese

The gap existed for months before anyone saw it.

Faith-based budgeting — the organizational equivalent of putting money in the collection plate and hoping God handles the ROI — is what made it invisible.

Lexi Reese

AI labor orphaning

That is not a measurement problem. It is a category error. We call it AI Labor Orphaning. AI does the work. The output gets credited to the human who approved it. The token bill lands in OpEx-software. The supervision time absorbs into salaried hours nobody is auditing. Eighty-seven percent of leaders admitted AI output is sometimes or always credited entirely to the human employee. This is the last-click attribution problem of the AI era, running in reverse.

What fills the vacuum? Belief. Forty-three percent assume that if AI was involved, it contributed. Only twelve percent have a clear methodology. Seventy-nine percent are worried AI budgets will be cut because they cannot connect spend to results. The cuts are not coming because AI does not work. They are coming because nobody can prove that it did.

Capping tokens may look like responsible governance, but it is like turning off a staticky radio rather than tuning the dial. The companies cutting AI budgets in 2026 will discover in 2027 that they cut the workflows that worked alongside the ones that did not.

The real cost of AI is not the model. It is the redesign

Three layers. Most organizations only manage the first.

Lexi Reese

What to actually do

The 12% of organizations that can answer the CFO treat AI like every other category of labor — with a cost per AI Work Hour that is accounted for by a set of AI assistants, co-pilots and agents that are held accountable to performance standards. 

• Audit the four sources separately. Each A-Level has different token economics, SaaS implications and human redesign requirements.
• Find the embedded SaaS repricing before your next renewal. Pull your top 20 contracts. Ask whether AI features previously included are now priced incrementally.
• Redesign the human role at A-Level 2 before you scale it. You cannot upskill into a role that has not been named.
• Build a system of record before you build the next agent. Start with one department. Two weeks. You will find something that surprises you.
• Stop calling it a tool. Start calling it labor. The language determines the chart of accounts.

When your blended AI rate is $22 an hour, the conversation shifts from ‘we spent $340,000 on AI’ to ‘we acquired a skilled workforce at $22 an hour.’ That sentence is defensible. A vendor invoice is not.

The CIOs who will have a defensible AI story in 2027 are the ones who renamed the work in 2026. Not because technology changed. Because they finally built the accounting to see it.

Findings are drawn from the 2026 AI Labor Report, fielded by Wakefield Research with 200 senior technology leaders at US enterprises of 1,000-plus employees, March 20–April 8, 2026 (±6.9pp at 95% confidence).

This article is published as part of the Foundry Expert Contributor Network.
Want to join?

When Anthropic unveiled Project Glasswing and the Mythos model, much of the discussion focused on the capabilities themselves.

Security leaders debated what these systems could mean for vulnerability discovery, exploit development and the pace of offensive innovation. Researchers examined technical benchmarks. Industry observers questioned how quickly these capabilities might fall into attackers’ hands.

Those conversations are important. They also point to a larger question that predominates my discussions with CISOs: How much time do we have?

Over the past year, conversations about AI in cybersecurity have changed noticeably. Twelve months ago, security leaders wanted to understand whether AI could meaningfully improve security operations. They wanted to know whether it could accurately investigate alerts, reduce analyst workload and operate reliably in production environments.

Today, security leaders are asking about timelines, implementation, how quickly AI is changing the threat landscape and what that means for how security teams operate.

Anthropic’s Mythos and Glasswing, OpenAI’s Daybreak and advances in DeepSeek accelerate those conversations. Each development provides another glimpse into the pace at which AI capabilities are advancing.

AI now reasons through security problems that historically required highly specialized expertise. The implications span vulnerability discovery, attack-path analysis, reconnaissance, social engineering and security operations.

The shift reflects a broader reality: cybersecurity is entering a period where the pace of adaptation may matter as much as the quality of defenses themselves. AI is accelerating both offense and defense simultaneously. Organizations are quickly redesigning security operations around that reality.

One consequence is becoming increasingly visible. For years, cybersecurity teams invested enormous effort in discovering threats, identifying vulnerabilities, gathering telemetry and collecting intelligence. AI is accelerating many of those activities simultaneously. Visibility is improving. Discovery is accelerating. Investigations are becoming faster and more comprehensive.

The bottleneck is beginning to move. The challenge increasingly centers on how quickly organizations can act on what they know. The organizations that gain an advantage may not be the ones with the most information. They will be the ones who can operationalize that information the fastest.

The timeline is compressing

Cybersecurity has experienced many major technology transitions. Cloud computing changed infrastructure. Mobile devices expanded the attack surface. Digital transformation connected systems that were previously isolated.

AI introduces a different dynamic.

Most technology transitions unfolded over years. Organizations had time to evaluate, pilot, deploy and gradually adapt operating models.

The current AI cycle moves at a different pace.

Capabilities improve continuously. New models arrive every few months. New research emerges every few weeks. Security teams absorb developments at the same time attackers do.

Vulnerability discovery provides a useful example. Security teams have long operated around a familiar cycle of discovery, validation, remediation and protection. AI systems accelerate every stage of that process. Similar patterns exist in phishing, reconnaissance, social engineering and attack planning.

A vulnerability that once moved through that cycle over weeks increasingly now moves through those stages in days or, in some cases, hours.

Attackers are already operating at the speed of AI. Defenders are now focused on reaching the same level of operational speed.

This shift is changing the questions CISOs ask.

Early discussions focused on capability. Could AI investigate alerts accurately? Could it operate reliably in production environments? Could it be trusted with meaningful security work?

As organizations gained experience with AI, the discussion shifted toward implementation. Security teams began evaluating where AI could create operational leverage and how quickly they could deploy it into existing workflows.

Today, many CISOs are focused on timing.

The pace of advancement is influencing planning horizons, budget decisions and operating-model discussions. Security leaders are evaluating how quickly they can introduce AI into investigations, threat hunting, detection engineering and response workflows. Boards are asking questions. Executive teams are paying attention.

Security programs that once viewed AI as a future initiative increasingly view it as a current operational priority.

The industry is moving from evaluating AI as a technology to incorporating AI as a security capability.

The timeline compression creates pressure on the traditional security operations model. Investigation speed, response speed and defensive coverage increasingly determine whether organizations can keep pace with adversaries operating with AI assistance.

Security operations are entering a new phase

The impact of AI is becoming particularly visible inside the SOC.

Many security operations centers were built around a straightforward assumption: alerts flow to human analysts who conduct investigations. Operational capacity scales primarily through hiring.

The volume of security data, the number of alerts and the complexity of modern environments have steadily increased. Security teams have responded by building processes, adding tools and creating specialized analyst roles.

AI introduces a new source of operational capacity.

Investigations that require analysts to examine dozens or hundreds of artifacts across endpoint, identity, cloud, network and email systems can now be performed in minutes. Analysts gain access to investigative depth and consistency that would be difficult to achieve manually at scale.

Many security leaders now view this capability through the lens of operating model design. They are examining how investigations are performed, how work is distributed and where human expertise creates the greatest value.

The evolution of the analyst role

One of the most important developments emerging from early production deployments is the evolution of analyst responsibilities.

Security analysts remain central to security operations. Their expertise becomes even more valuable as AI systems take on larger portions of investigative work.

Threat hunting, detection engineering, response strategy, governance and oversight are receiving increased attention. Analysts spend more time shaping how investigations are conducted, evaluating outcomes and improving overall security operations.

Many organizations are already beginning this shift.

Teams are investing more heavily in proactive security activities. Detection engineering programs are expanding. Threat hunting is becoming more accessible. Analysts are spending more time improving systems and less time repeating investigative tasks.

These changes create what I think of as an analyst-amplified SOC: an environment where AI expands the reach of security professionals and enables deeper security work across the organization.

Trust is critical and it doesn’t have to compromise speed

Faced with a compressing timeline, the instinct is to treat speed and trust as a trade-off, i.e., move faster, verify less. That trade-off feels inevitable. It isn’t.

You don’t trust AI in the abstract. You trust that a system understands your tools, your telemetry and the edge cases that only exist in your network. The problem was never speed. It’s speed without context. The faster a context-blind system runs, the more decisions you’re left unable to verify.

The tension eases when the system is quick to deploy and tunes to your environment once it’s there, rather than treating every network the same. Speed stops being the thing you trade against trust. The more it learns about your environment, the sharper and more trustworthy it becomes, so the two compound rather than compete. Trust still develops through operational evidence such as measurable outcomes, visibility into decisions and consistent performance. But where that evidence accrues matters.

The organizations making the fastest real progress understand this. They don’t compromise quality and trust for speed. They invest in AI that earns trust inside their own environment, so they don’t have to choose.

Leadership during a period of rapid change

The conversations surrounding Mythos and Glasswing reflect a broader reality facing security leaders.

AI is becoming part of both offense and defense. Security teams are incorporating it into investigations, detection engineering, response workflows and threat hunting. Attackers are incorporating it into their own operations.

Security leaders have an opportunity to modernize operating models, expand defensive capacity and build organizational experience while these capabilities continue to evolve.

The organizations making progress today are investing in readiness. They are building experience, adapting workflows and preparing teams for a new model of security operations.

The next phase of cybersecurity will be defined by how effectively organizations combine human judgment with machine-scale execution.

The question facing security leaders is increasingly clear: How quickly can their organizations adapt to a continuously changing threat environment?

This article is published as part of the Foundry Expert Contributor Network.
Want to join?

dotshock | shutterstock.com

Es würde mich nicht wundern, wenn Ihnen dieses Szenario bekannt vorkommt. Denn Situationen wie diese sind keine Ausnahme. Laut der Trovarit-Studie „ERP in der Praxis 2024/25“, für die über 1.700 DACH-Unternehmen befragt wurden, nimmt eine typische ERP-Einführung im Mittelstand zwischen zehn und dreizehn Monaten in Anspruch. Parallel werden in mehr als der Hälfte der Projekte Zeit- oder Budgetziele nicht eingehalten. Die Implementierungskosten pro Nutzer liegen demnach bei durchschnittlich 5.917 Euro, wobei Schulung, Datenmigration und Anpassung 50 bis 70 Prozent der Gesamtkosten ausmachen – nicht die Lizenz.

Für einen Konzern mit 800 Mitarbeitern, SAP-Legacy und drei Produktionsstandorten sind das keine Schreckenszahlen. Für einen Immobilienmakler mit 40 Mitarbeitern, der seine Angebote automatisiert versenden will oder einen Steuerberater, der Mandantenpost automatisch klassifizieren möchte, durchaus. Diese Mittelständler brauchen kein ERP-Projekt, sondern im Wesentlichen drei funktionierende Workflows.

Das Low-Code-Versprechen – und die KMU-Praxis

Low-Code– und KI-Automatisierungsplattformen wie n8n oder Make.com versprechen, die Lücke zwischen ERP-Großprojekt und Stillstand aufzulösen. Dabei ist anzumerken, dass beide Plattformen keinen Ersatz für ein ERP darstellen: Sie substituieren keine Stücklisten-Verwaltung, keine mehrstufige Fertigungsplanung und keine IFRS-Konsolidierung. Dafür ist weiterhin ein ERP nötig.

Aber Plattformen wie die beiden genannten stellen Verbindungen zwischen Systemen her, die keine native Schnittstelle haben. Entscheidungen automatisieren, die nach klaren Regeln getroffen werden und Workflows bauen, die bisher über E-Mail, Excel-Export und Telefonanruf abgewickelt werden – das sind für den deutschen Mittelstand oft genau die Prozesse, die täglich die meiste Zeit kosten und zugleich selten im ERP verortet sind. Vielmehr sitzen diese in Outlook-Ordnern, in freigegebenen Tabellenblättern und in den Routinen von Mitarbeitern, die „das schon immer so machen”.

Das KfW-Mittelstandspanel vom Februar 2026 (PDF) stellt fest, dass 20 Prozent der mittelständischen Unternehmen in Deutschland KI einsetzen – fünfmal mehr als noch im Jahr 2018. Bei Unternehmen mit über 50 Mitarbeitern liegt die Quote bereits bei 36 Prozent. KI ist im Mittelstand also angekommen. Die Frage ist nur, in welcher Form. Hier hilft ein Blick auf die KPMG-Studie „Digitalisierung im Rechnungswesen 2025/2026“. Demnach:

• setzen 53 Prozent der Befragten KI in der Buchhaltung ein oder befinden sich gerade in der Einführungsphase.
• berichten 37 Prozent dadurch von sofortigen Zeiteinsparungen bei transaktionalen Prozessen.

Was diese Studien allerdings nicht zeigen, ist der Implementierungsaufwand. Und genau hier liegt der große Unterschied zwischen dem KI-Einsatz per Low-Code und einem ERP-Einführungsprojekt. Das verdeutlichen die folgenden drei Anwendungsfälle für KMU im Bereich KI und Low-Code. Diese sind dazu geeignet, innerhalb von Wochen (statt Monaten oder Jahren) Ergebnisse zu liefern:

• Rechnungseingang: PDF-Rechnungen oder XRechnung-Dokumente entgegennehmen, relevante Felder per KI extrahieren und gegen ERP-Stammdaten prüfen oder vorkontierte Buchungsvorschläge generieren – was früher ein Buchhalter täglich in Stunden manuell erledigte, kann heute als Hintergrundprozess laufen. Dabei werden Fehler weiterhin von Menschen überprüft, statt automatisch in die Buchung zu laufen. Die typische Implementierungszeit hierfür liegt (in überschaubaren Setups) bei zwei bis vier Wochen.
• Angebotsmanagement: Eine Kundenanfrage geht per E-Mail ein, KI extrahiert den Leistungsumfang, prüft gegen Preislisten, generiert einen Angebotsentwurf und leitet ihn zur Freigabe an einen Sachbearbeiter weiter. Der Kunde bekommt sein Angebot schneller, die Bearbeitung kostet weniger Zeit. Dabei ist kein ERP-Modul beteiligt – und es fällt keine siebenstellige Projektsumme an.
• Mandantenpost bei Steuerberatern: Eingehende Dokumente werden nach Typ und Dringlichkeit klassifiziert und Zuständigkeiten zugewiesen. Was in Kanzleien mit hohem Postaufkommen täglich zu Engpässen führt, lässt sich so als regelbasierter Workflow abbilden – ganz ohne großes DATEV-Einführungsprojekt.

Woran Low-Code und KI scheitern – und wie Sie das verhindern

Das größte operative Risiko bei Low-Code-Projekten ist Schatten-IT, beziehungsweise -KI. So hat der Digitalverband Bitkom in einer Befragung vom Oktober 2025 herausgefunden, dass 17 Prozent der Umfrageteilnehmer davon ausgehen, dass ihre Mitarbeiter KI-Tools über private Accounts dienstlich nutzen. Bei weiteren 17 Prozent gab es diesbezüglich vereinzelte Fälle – und bei acht Prozent ist dieses Gebaren weit verbreitet. Sind in einem solchen Fall Kundendaten im Spiel, handelt es sich um einen DSGVO-Verstoß ohne Auftragsverarbeitungsvertrag als Grundlage. Insofern braucht eine Low-Code-Strategie klare Nutzungsrichtlinien – ansonsten tauscht man ein ERP-Risiko gegen ein Compliance-Risiko, was eher kein Fortschritt wäre.

Darüber hinaus können Low-Code-Plattformen bei hochspezialisierten Logikstrukturen (mehrstufige Produktionsplanung, Konzernkonsolidierung, ISO-zertifizierte Auditpfade) an strukturelle Grenzen stoßen. Und: Der Vendor Lock-in ist real – proprietäre Plattformen lassen sich schlecht migrieren. Hier haben Open-Source- und Self-Hosting-Lösungen einen strukturellen Vorteil. Allerdings ist auch das kein Garant dafür, dass Abhängigkeiten künftig ausbleiben.

Vor diesem Hintergrund empfehle ich Mittelständlern, sich an den folgenden drei Schritten zu orientieren, um ihre Low-Code-KI-Ambitionen nachhaltig zu verwirklichen.

1. Pilotprozess nach harten Kriterien wählen: Der erste Kandidat sollte vier Kriterien gleichzeitig erfüllen: hohes Volumen, klare fachliche Regeln, keinen schreibenden Zugriff auf das ERP in der ersten Ausbaustufe, und einen Verantwortlichen, der den Prozess heute schon im Detail kennt. Die Bereiche Rechnungseingang und Angebotsvorbereitungerfüllen diese Voraussetzungen fast immer. Ein Prozess, der gleichzeitig produktionskritisch und schlecht dokumentiert ist, hingegen fast nie. Wer an dieser Stelle den falschen Piloten wählt, muss damit rechnen, dass die Akzeptanz bei einem zweiten Versuch deutlich geringer ist.
2. Datenschutz vor dem ersten Workflow klären: Ein Auftragsverarbeitungsvertrag mit dem Modellanbieter ist eine Sache, die vor dem Produktivstart stehen sollte, nicht in der Nachbereitung. Diese Prüfung muss auch mit Blick auf die Low-Code-Plattform selbst erfolgen, nicht nur das Sprachmodell dahinter. Wer hier sauber arbeitet, nimmt der eingangs beschriebenen Schatten-IT-Problematik die Grundlage: Wenn die offizielle Lösung schneller einsatzbereit ist als die inoffizielle, greifen Mitarbeiter auch seltener auf private KI-Accounts zurück.
3. Skalierungsgrenze vorab definieren: Legen Sie fest, ab welchem Komplexitätsgrad ein Prozess zurück auf die ERP-Roadmap wandert, bevor Sie ihn brauchen. Diese Grenze nachträglich einzuziehen, kostet – inmitten eines gewachsenen Workflows mit zwölf Verzweigungen – mehr Zeit als die ursprüngliche ERP-Entscheidung eingespart hat. Ein einfaches Signal funktioniert in der Praxis gut: Sobald ein Workflow mehr als drei verschachtelte Bedingungen benötigt, um eine einzelne Entscheidung abzubilden, sollte die Logik überprüft – nicht erweitert – werden.

Der eigentliche Nutzwert für KMU

Sechs Monate weniger Implementierungszeit bedeutet nicht nur sechs Monate früher produktiv sein zu können. Es bedeutet:

• Ergebnisse demonstrieren zu können, bevor das ERP-Budget freigegeben ist.
• Prozesse testen zu können, bevor sie skaliert werden.
• Scheitern zu können, weil es um einen Workflow für 4.000 Euro und nicht um ein Projekt für 200.000 Euro geht.

Ihr ERP-Projekt darf ruhig auf der Roadmap bleiben. Aber für die nächsten zwölf Monate gibt es meistens einen schnelleren Weg. (fm)

Dieser Beitrag wurde im Rahmen des deutschsprachigen Experten-Netzwerks von Foundry veröffentlicht. Lust mitzumachen? Jetzt bewerben!

Model Context Protocol (MCP) verbindet KI-Agenten mit Datenquellen und erfreut sich im Unternehmensumfeld wachsender Beliebtheit. Allerdings ist auch MCP nicht frei von Sicherheitslücken, wie entsprechende Entdeckungen, etwa beim SaaS-Anbieter Asana oder dem IT-Riesen Atlassian gezeigt haben. Inzwischen hat sich jedoch einiges in Sachen MCP-Sicherheit getan. Einerseits wurden mit Blick auf das Kernprotokoll etliche Fortschritte erzielt. Beispielsweise in Form von Support für OAuth sowie für Authentifizierungs-Server von Drittanbietern und Identity-Management-Systeme. Darüber hinaus wurde inzwischen auch eine offizielle MCP Registry geschaffen, die einen Überblick über sichere, öffentlich verfügbare MCP-Server bietet.

Dennoch bestehen weiterhin Sicherheitslücken, die sich für diverse Cyberschandtaten ausnutzen lassen – Prompt Injection, Tool Poisoning, Token-Diebstahl, Server-übergreifende Attacken oder manipulierte Messages sind nur einige von vielen Beispielen. Mit anderen Worten: Unternehmen, die sich beim Aufbau von Agentic-AI-Systemen einen Wettbewerbsvorteil verschaffen wollen, müssen erhebliche Anstrengungen unternehmen, um zu gewährleisten, dass sensible Daten nicht nach außen dringen. Glücklicherweise gibt es diverse Tools, die dabei Unterstützung versprechen.

In diesem Artikel lesen Sie:

• was Security-Tools für MCP leisten sollten, und
• welche Angebote in diesem Bereich interessant sind.

Das sollten MCP-Sicherheitslösungen können

Die Gefahr von Datenlecks, Prompt Injections und weiteren Sicherheitsbedrohungen besteht unabhängig davon, ob Unternehmen:

• ihre eigenen KI-Agenten mit MCP-Servern von Drittanbietern,
• ihre eigenen MCP-Server mit Drittanbieter-Agenten, oder
• ihre eigenen Server mit den eigenen Agenten verbinden.

Soll heißen: Unternehmen müssen in jedem Fall Autorisierungen und Berechtigungen überprüfen, detaillierte Zugriffskontrollen implementieren und alles protokollieren. Daraus ergeben sich auch die Anforderungen für MCP-Sicherheitslösungen. Diese sollten bieten:

• MCP-Servererkennung. Für Mitarbeiter eines Unternehmens ist es einfach, MCP-Server herunterzuladen und zu nutzen. Mit Scan-Services für MCP-Server können Unternehmen sämtliche Instanzen von Schatten-MCP-Servern in ihrer Umgebung finden.
• Laufzeitschutz. KI-Agenten kommunizieren mit MCP-Servern in natürlicher Sprache. MCP-Sicherheits-Tools sollten deshalb in der Lage sein, diese Kommunikation auf Sicherheitsprobleme wie Prompt Injections hin zu überwachen.
• Authentifizierungs- und Zugriffskontrollen. Das MCP-Protokoll unterstützt inzwischen OAuth, aber das ist nur ein erster Schritt. Für zusätzliche Sicherheit empfehlen sich Tools mit integrierten Kontroll-Frameworks für Zero Trust und Least Privilege.
• Logging und Observability. Tools und Plattformen sollten zudem die Möglichkeit bieten, MCP-Protokolle zu sammeln, Sicherheitsteams über Richtlinienverstöße zu informieren, Compliance-Daten zu erfassen oder Protokolle in die bestehende Sicherheitsinfrastruktur einzuspeisen.

MCP-Security-Angebote

Im Folgenden haben wir die Anbieter von MCP-Security-Tools in drei Kategorien aufgeteilt. Diese Aufstellung erhebt keinen Anspruch auf Vollständigkeit.

Hyperscaler

Für Unternehmen, die sich vollständig auf eine bestimmte Cloud-Plattform verlassen, bieten die MCP-Tools des jeweiligen Hyperscalers einen einfachen Einstieg.

• Amazon Web Services (AWS) hat Mitte 2025 seine eigene agentenbasierte KI-Plattform eingeführt. Amazon Bedrock AgentCore umfasst ein Gateway, das mehrere Protokolle unterstützt (darunter auch MCP), ein Identity-Management-System sowie Observability.
• Microsoft bietet einen grundlegenden Azure-MCP-Server an, inklusive Support für Azure Key Vault. Darüber hinaus unterstützen auch Azure AI Foundry Agent Service und Azure API Management das Model Context Protocol. Zudem bietet Microsoft mit dem Agent Framework auch ein Open-Source-Entwicklungskit, das sowohl MCP als auch Agent2Agent unterstützt und beispielsweise Schutz vor Prompt Injections verspricht.
• Google Cloud kündigte Anfang 2025 seine MCP Toolbox für Datenbanken an – inklusive integrierter Authentifizierung und Observability. Außerdem hat der Hyperscaler auch eine Referenzarchitektur veröffentlicht, um MCP-Server auf seiner Cloud-Plattform abzusichern.

Große Plattformanbieter

• Der IT-Dienstleister Cloudflare hat mit MCP Server Portals ein Tool veröffentlicht, mit dem Unternehmen MCP-Verbindungen zentralisiert absichern und überwachen können. Die Funktion ist Bestandteil der Cloudflare-One-Plattform.
• Palo Alto Networks hat mit Blick auf MCP-Sicherheit mehrere Eisen im Feuer. Mit Prisma AIRS hat das Unternehmen einen eigenen, intermediären MCP-Server veröffentlicht. Dieser sitzt zwischen den KI-Agenten und dem eigentlichen MCP-Server und erkennt schadhafte Inhalte und Daten. Das Tool MCP Security ist hingegen Bestandteil von Cortex Cloud WAAS und überprüft die MCP-Kommunikation an der Netzwerkgrenze auf bösartige Aktivitäten.
• SentinelOne gewährt mit seiner Singularity Platform ebenfalls Einblick in die MCP-Interaktionskette und bietet zum Beispiel Warnmeldungen und automatisierte Incident Response für MCP-Server auf lokaler oder Remote-Ebene.
• Die Plattform von Acuvity (seit Februar 2026 Teil von Proofpoint) verspricht, MCP-Server umfassend abzusichern. Dafür sorgt laut dem Anbieter eine Kombination aus Least-Privilege-Execution, unveränderlichen Laufzeiten, kontinuierlichen Schwachstellenscans, Authentifizierung und Bedrohungserkennung.
• Daneben hat auch Broadcom MCP-Sicherheitsfunktionen für VMware Cloud Foundation angekündigt, die künftig mehr Sicherheit für agentenbasierte Workflows gewährleisten sollen.

Startups

• Das API-Security-Startup Akto hat eine MCP-Security-Plattform im Angebot. Sie umfasst ein Discovery Tool, um MCP-Server in Unternehmensumgebungen zu identifizieren, Security-Testing-Werkzeuge sowie Monitoring- und Threat-Detection-Funktionen.
• Invariant Labs bietet mit MCP-Scan ein quelloffenes Tool, das die statische Analyse und Echtzeitüberwachung von MCP-Servern ermöglicht. Mit Guardrails hat das Startup auch ein kommerzielles Produkt im Angebot. Dabei handelt es sich um einen Proxy. Der zwischen KI-Agenten und MCP-Servern sitzt und vor Security-Risiken schützen soll. Das Tool befähigt Anwender außerdem dazu, Richtlinien aufzusetzen.
• Highflame (vormals Javelin) addressiert ebenfalls das Thema MCP-Sicherheit. Etwa mit Funktionen wie MCP-Server auf Risiken zu scannen oder Datenanfragen zu überprüfen.  
• Lasso Security stellt ein Open-Source-MCP-Gateway zur Verfügung, das die Konfiguration und das Lebenszyklusmanagement von MCP-Servern ermöglicht und Messages um sensible Informationen bereinigt.

(fm)

Dieser Artikel ist im Original bei unser Schwesterpublikation CSOonline.com erschienen.

Microsoft and Mistral are betting that the future of enterprise AI is in sovereign infrastructure and model choice, rather than with one locked-in system. 

The companies have announced a “significant expansion” of their strategic partnership, which includes a multibillion dollar commitment from Microsoft. Mistral will add to its GPU infrastructure in Europe and extend access to its frontier multilingual models, while Microsoft will expand its sovereign cloud capabilities. The companies will also align on a joint go-to-market plan and will pursue enterprise opportunities together across Europe and globally, as well as funding proofs of concept (PoCs), offering Azure credits, and leading workshops to drive AI innovation with customers.

The partnership between the tech giant and the three-year-old French startup might seem an odd combination at first glance, analysts note, as both develop enterprise AI models and offer access as-a-service. But it reflects changing AI market dynamics.

“It’s possible to be both a competitor and a partner at the same time,” noted technology analyst Carmi Levy. Large cloud providers are becoming AI marketplaces in their own right, he pointed out, and are drifting away from exclusively promoting their own models. Building Mistral support into their infrastructure avoids platform lock-in and removes a “key objection for customers looking for options.”

“As much as Microsoft would want everybody standardizing on Copilot and Phi, it recognizes the simple fact that customers increasingly want to choose their own models,” said Levy.

Expands model access, sovereign cloud capabilities

As part of the agreement, Mistral will expand its Europe-based capacity with thousands of Nvidia Vera Rubin GPUs.

Mistral CEO and co-founder Arthur Mensch described a “slight gap” in compute capacity in Europe, noting that this expansion will provide more compute capability and support Microsoft’s cloud and AI services, providing a “shared platform for training, inference and large-scale deployment.” The companies call it a critical step to allow Microsoft customers to benefit from Mistral’s “scientific and compute innovations.”

In addition, Mistral Medium 3.5 and OCR 4 models are now available in Microsoft Foundry, and Mistral Medium 3.5 can be used in Microsoft Copilot Studio.

The partnership also extends Microsoft’s sovereign cloud infrastructure as well as combining Mistral’s frontier models with Microsoft’s security, compliance, and cloud-to-edge platform. This gives enterprises, particularly those in regulated markets, the ability to deploy AI where they see fit, while maintaining control over their data and workloads, according to the companies.

Further, customers will be able to build AI using the same models, tools, APIs, and workflows they’re used to, across Microsoft Foundry, Foundry Local, and Azure Local, and opt for fully Azure-hosted cloud environments; cloud-connected, controlled Azure Local environments that only use cloud-based Azure when necessary; and fully-disconnected environments that can operate independently for more sensitive scenarios.

“Europe should have access to the world’s most capable AI without compromising control over their data, operations or digital future,” said Brad Smith, vice chair and president, Microsoft, noting that with this partnership, the company is honoring its European digital commitments and giving customers a foundation for AI so they can “operate on their own terms.” Customers with “heightened sovereignty needs” will be able to exercise more control with “resilience and assurance” and continued access to Mistral’s open-weight models.

Enterprise credibility

Gartner distinguished VP analyst Arun Chandrasekaran noted that there’s no doubt that this agreement strengthens Microsoft’s sovereignty messaging and its position in regulated industries, and the tech giant benefits by expanding its AI portfolio with a “credible European frontier model provider”

He pointed to key differences from the initial partnership struck by the two companies in 2024; whereas originally Microsoft was hosting Mistral’s models, it is now consuming capacity built by Mistral in Europe.

Ultimately, the deal emphasizes European data centers, customer-controlled deployments, Azure Local, and fully-disconnected environments, addressing many of the concerns that surrounded the original Azure cloud only relationship, Chandrasekaran explained.

For Mistral, the partnership provides “enterprise credibility, and repeatable infrastructure revenue” that can fund continued AI platform development, he said. The combination of Microsoft’s enterprise AI platform with Mistral’s models and European AI infrastructure will give joint customers more deployment flexibility and expand options around data residency, sovereign AI deployments, and disconnected/on-premises environments.

“It also gives customers more model choice, reducing dependence on a single AI provider,” said Chandrasekaran.

A complementary partnership

Mistral continues to innovate with its frontier AI models and its chat and coding agent, Vibe (formerly Le Chat), yet it doesn’t attract as much attention as Claude or ChatGPT.

One of the company’s key differentiators is its targeted business model. Levy pointed out that not every workload requires “full-flight GPT.” For customers trying to rein in costs and limit exposure with on-premises deployments, Mistral’s “more focused capabilities can represent a cost-effective alternative.”

Chandrasekaran pointed to Mistral’s combination of high-performance open-weight models, strong multilingual capabilities, and a “focus on efficient inference that lowers deployment costs.”

Unlike many frontier AI companies, it offers customers greater flexibility to self-host and customize models; this makes it particularly attractive for enterprises and governments with sovereignty or regulatory requirements, he said. Its European roots also position it as the leading alternative for organizations seeking cutting-edge AI outside the US and Chinese ecosystems.

Bill Wong, research fellow at Info-Tech Research Group, also pointed to Mistral’s high-quality models and “adeptness as a sovereign AI leader.” There is growing demand for AI companies that comply with regional laws and data residency, and Mistral is established as “one of the most prominent European players.”

“Such a strategic position makes it a great partner for Microsoft to further expand its AI offerings beyond just being a single-model provider,” he said. Customers get freedom of choice while complying with data sovereignty and regulatory limitations without having to execute a separate AI deployment, while Mistral, for its part, can go beyond Europe and gain more visibility with international businesses.

Mistral brings both “technological and political advantages,” Levy noted. The startup’s European roots give Microsoft more credibility “at a fraught time for geopolitical relationships.” Customers in Europe and beyond are concerned about US exposure, and Mistral can provide a safer choice.

Meanwhile, Microsoft can deploy European-developed AI models running on European infrastructure, thus maximizing regulatory compliance while offering next-level enterprise marketing scale that Mistral “simply couldn’t achieve on its own,” said Levy. Mistral-based workloads deployed on Azure will also benefit from Microsoft’s “comprehensive security certifications, governance frameworks, and monitoring.”

Bottom line: Both companies can maximize their unique roadmaps through the partnership, he said. “As the rules of the AI economy continue to evolve, expect more eyebrow-raising deals like this to be signed.”

Ever since Chinese AI startup DeepSeek launched three years ago, enterprise executives have been nervous about relying on Chinese AI models. 

But now that the latest Chinese AI offerings, Alibaba’s 2.4-trillion-parameter model Qwen3.8 Max and Moonshot’s 2.8-trillion-parameter model Kimi K3, are promising even more powerful performance, those IT executives are being forced to again ask if these models are worth using, even in a limited fashion.

Former Walmart head of risk Steven Eric Fisher, now an independent cybersecurity and risk advisor, thinks they should at least take another look.

“Enterprises should take these models seriously, but neither adopt nor reject them solely because they are Chinese,” he said. “They should be assessed like any other critical technology dependency: jurisdiction, ownership, training and software provenance, licensing, data handling, hosting, security, reliability, and the ability to independently test their behavior. Geopolitical exposure is a legitimate risk factor, but it should be incorporated into technical and supply-chain diligence rather than used as a substitute for it.”

Choose applications with care

He added, “Chinese models may be especially valuable for coding, multilingual processing, high-volume document analysis, research, synthetic-data generation, and privately operated security or forensic workflows, but they should be subject to task-specific testing rather than broad benchmark claims.”

Shashi Bellamkonda, principal research director at Info-Tech Research Group, agreed that the Chinese models can work well if they are only used in carefully chosen applications. 

“Although Moonshot’s K3 still trails Claude’s Fable 5 and GPT 5.6 Sol on performance and user experience, good companies that have governance and prompt guardrails will not face the instability and improvisation of [the Chinese] models,” he said. “These models will win in usage. US frontier models are leading as the best models, but Chinese models will be sufficient for high-volume, low-drama tasks that cost less for non-critical transactions.”

On the flipside, Bellamkonda suggested a variety of areas where enterprises should avoid Chinese AI models, including “customer-facing work without a human in the loop, regulated or sensitive data, and anything where a hallucinated answer creates legal or safety exposure. That is where the reliability gap and the political-radioactivity concern both bite, and where the closed American models still earn their premium.”

Bellamkonda said he didn’t see the differences in data reliability, mostly involving hallucination rates, as meaningful for enterprise AI strategy decisions.

“Every open-weight model in this class can get facts wrong or make things up. That is fixable with the right setup, so it is not a reason to avoid these models,” he said. “For high-volume tasks with clear limits, you feed the model your own trusted documents to answer from, and you keep a person checking the output. That combination is safe for production. The model on its own is not.”

Too early for enterprises to consider

However, not everyone agrees that the latest Chinese models have earned their place as enterprise AI decision options. 

Cybersecurity consultant Brian Levine, executive director of FormerGov, focused on Chinese technology concerns when he worked for the US Justice Department as its representative in the US law enforcement Joint Liaison Group (JLG) with China. 

“It is way too early for US enterprises to seriously consider these models,” he said. “Until proven otherwise, enterprises should assume that if they use these models, they may be granting China complete access to everything they do through the models, and potentially access to their networks and employees more broadly. At this point, any pros of using such models are strongly outweighed by the potential security, confidentiality, and reliability concerns.”

Tom Findling, CEO of Conifers.ai, was equally emphatic that enterprise CIOs need to steer clear of these newer Chinese models. 

“Using them inhouse? Absolutely not. You simply don’t know what is planted inside of it and you don’t know what training data is put into them,” Findling said. 

Mike Wilkes, enterprise CISO at Aikido Security, added that the very attractive pricing for these Chinese models may be appealing, but suggested that, despite the low cost, they’re ultimately too risky.

“Enterprises should take these models seriously, but not romantically. Parameter count is horsepower measured in a showroom, not braking distance in the rain,” he said. “The real tests are reliability on your data, the cost of a wrong answer, and whether the model behaves predictably under pressure.”

He noted that the benchmarks on the latest open-weights models are impressive, and very close to those of the frontier lab models, which makes the cost ”incredibly seductive, especially when a team does not want to risk their data being used to train those frontier models.”

But the Chinese models can still work in specific circumstances. “The strongest value will be in bounded, reversible and inspectable work: coding inside a sandbox, multilingual translation, document triage, data extraction and other high-volume tasks where outputs can be verified,” he said. “Cheap intelligence is valuable, but only when it is not mistaken for trustworthy judgment.”

Wilkes added that the regulatory issues surrounding Chinese models can be especially problematic. Texas, for example, has banned their usage.  

A rational choice for some workloads

However, Yuri Goryunov, CIO of consulting firm Acceligence, argued that CIOs should seriously consider these models. 

“Counterintuitively, the biggest benefit of Kimi and models like it is the lack of guardrails,” Goryunov said. “Think of it as stick shift cars in the era of automatics. If you want ease and comfort, stay with the frontiers because they have cruise control, shift the gears for you and they decide when. If you want performance and control, expand your horizons. But a stick shift assumes you know how to drive one: you bring your own governance, your own evals, your own safety layer. That’s a cost and specialized talent, which is super rare, and for the right organization it’s also the whole point.”

Goryunov’s bottom line: “For internal, high-volume, well-harnessed workloads, [the Chinese models] have moved from ‘watch list’ to ‘rational choice.’”

This article originally appeared on CIO.com.

Ever since Chinese AI startup DeepSeek launched three years ago, enterprise executives have been nervous about relying on Chinese AI models. 

But now that the latest Chinese AI offerings, Alibaba’s 2.4-trillion-parameter model Qwen3.8 Max and Moonshot’s 2.8-trillion-parameter model Kimi K3, are promising even more powerful performance, those IT executives are being forced to again ask if these models are worth using, even in a limited fashion.

Former Walmart head of risk Steven Eric Fisher, now an independent cybersecurity and risk advisor, thinks they should at least take another look.

“Enterprises should take these models seriously, but neither adopt nor reject them solely because they are Chinese,” he said. “They should be assessed like any other critical technology dependency: jurisdiction, ownership, training and software provenance, licensing, data handling, hosting, security, reliability, and the ability to independently test their behavior. Geopolitical exposure is a legitimate risk factor, but it should be incorporated into technical and supply-chain diligence rather than used as a substitute for it.”

Choose applications with care

He added, “Chinese models may be especially valuable for coding, multilingual processing, high-volume document analysis, research, synthetic-data generation, and privately operated security or forensic workflows, but they should be subject to task-specific testing rather than broad benchmark claims.”

Shashi Bellamkonda, principal research director at Info-Tech Research Group, agreed that the Chinese models can work well if they are only used in carefully chosen applications. 

“Although Moonshot’s K3 still trails Claude’s Fable 5 and GPT 5.6 Sol on performance and user experience, good companies that have governance and prompt guardrails will not face the instability and improvisation of [the Chinese] models,” he said. “These models will win in usage. US frontier models are leading as the best models, but Chinese models will be sufficient for high-volume, low-drama tasks that cost less for non-critical transactions.”

On the flipside, Bellamkonda suggested a variety of areas where enterprises should avoid Chinese AI models, including “customer-facing work without a human in the loop, regulated or sensitive data, and anything where a hallucinated answer creates legal or safety exposure. That is where the reliability gap and the political-radioactivity concern both bite, and where the closed American models still earn their premium.”

Bellamkonda said he didn’t see the differences in data reliability, mostly involving hallucination rates, as meaningful for enterprise AI strategy decisions.

“Every open-weight model in this class can get facts wrong or make things up. That is fixable with the right setup, so it is not a reason to avoid these models,” he said. “For high-volume tasks with clear limits, you feed the model your own trusted documents to answer from, and you keep a person checking the output. That combination is safe for production. The model on its own is not.”

Too early for enterprises to consider

However, not everyone agrees that the latest Chinese models have earned their place as enterprise AI decision options. 

Cybersecurity consultant Brian Levine, executive director of FormerGov, focused on Chinese technology concerns when he worked for the US Justice Department as its representative in the US law enforcement Joint Liaison Group (JLG) with China. 

“It is way too early for US enterprises to seriously consider these models,” he said. “Until proven otherwise, enterprises should assume that if they use these models, they may be granting China complete access to everything they do through the models, and potentially access to their networks and employees more broadly. At this point, any pros of using such models are strongly outweighed by the potential security, confidentiality, and reliability concerns.”

Tom Findling, CEO of Conifers.ai, was equally emphatic that enterprise CIOs need to steer clear of these newer Chinese models. 

“Using them inhouse? Absolutely not. You simply don’t know what is planted inside of it and you don’t know what training data is put into them,” Findling said. 

Mike Wilkes, enterprise CISO at Aikido Security, added that the very attractive pricing for these Chinese models may be appealing, but suggested that, despite the low cost, they’re ultimately too risky.

“Enterprises should take these models seriously, but not romantically. Parameter count is horsepower measured in a showroom, not braking distance in the rain,” he said. “The real tests are reliability on your data, the cost of a wrong answer, and whether the model behaves predictably under pressure.”

He noted that the benchmarks on the latest open-weights models are impressive, and very close to those of the frontier lab models, which makes the cost ”incredibly seductive, especially when a team does not want to risk their data being used to train those frontier models.”

But the Chinese models can still work in specific circumstances. “The strongest value will be in bounded, reversible and inspectable work: coding inside a sandbox, multilingual translation, document triage, data extraction and other high-volume tasks where outputs can be verified,” he said. “Cheap intelligence is valuable, but only when it is not mistaken for trustworthy judgment.”

Wilkes added that the regulatory issues surrounding Chinese models can be especially problematic. Texas, for example, has banned their usage.  

A rational choice for some workloads

However, Yuri Goryunov, CIO of consulting firm Acceligence, argued that CIOs should seriously consider these models. 

“Counterintuitively, the biggest benefit of Kimi and models like it is the lack of guardrails,” Goryunov said. “Think of it as stick shift cars in the era of automatics. If you want ease and comfort, stay with the frontiers because they have cruise control, shift the gears for you and they decide when. If you want performance and control, expand your horizons. But a stick shift assumes you know how to drive one: you bring your own governance, your own evals, your own safety layer. That’s a cost and specialized talent, which is super rare, and for the right organization it’s also the whole point.”

Goryunov’s bottom line: “For internal, high-volume, well-harnessed workloads, [the Chinese models] have moved from ‘watch list’ to ‘rational choice.’”

AMD has expanded its AI infrastructure portfolio with the launch of Helios, an open, rackscale AI infrastructure designed for frontier AI and sovereign computing. Helios is built around AMD’s next-generation Instinct GPUs, EPYC Venice processors, Pensando networking and the ROCm software stack.

“Helios is AMD’s first complete AI rack system with GPUs, CPUs, and networking built together, instead of selling separate chips. It is well suited for training large AI models, memory heavy models, long context processing and high volume inference, and AMD’s biggest shot yet at challenging Nvidia’s dominance,” said Pareekh Jain, CEO at EIIRTrend & Pareekh Consulting.

AMD has also secured an early hyperscale deployment for Helios with Microsoft agreeing to deploy it to power its frontier model AI inference, its AI customers, and support Azure AI services.

The architecture behind Helios

The launch of Helios marks AMD’s latest attempt to strengthen its position in a market where Nvidia continues to dominate AI infrastructure. Unlike previous AMD AI offerings centred on individual accelerators, Helios is designed as a complete rack-scale system integrating compute, networking and software.

According to Jain, Helios goes up against Nvidia’s Vera Rubin rack. “Nvidia is faster on raw inference speed and has a faster internal connection between chips whereas AMD wins on memory size and offers better value for the price and power used. It’s standout feature is memory, where each rack packs about 50% more total memory than Nvidia’s competing system, which helps run very large AI models. It also uses open, industry-standard connections instead of Nvidia’s private technology, giving buyers more flexibility,” he said.

The AMD Helios rackscale design includes 72 AMD Instinct MI455X GPUs with AMD EPYC Venice CPUs and AMD Pensando Vulcano networking using UALink, optimized for compute, data movement, and system efficiency. The platform also supports both OCP and MX data types, delivering up to 2.9 EFLOPS of FP4 and 1.4 EFLOPS of FP8 compute for AI training and inference.

It also integrates 31TB of HBM4 memory with 19.6TB/s of memory bandwidth, while a liquid-cooling design uses quick-disconnect connections to efficiently dissipate heat. It is designed on open standards including OCP Open Rack Wide (ORW), Ultra Accelerator Link (UALink), and Ultra Ethernet Consortium (UEC) and can be scaled efficiently across datacenters while optimizing power, cooling, and serviceability for modern AI infrastructure, said the company.

On the security front, Helios incorporates a hardware root of trust and continuous attestation at every layer. It supports hardware-enforced isolation, encrypted memory and interconnects to help protect AI models, data and workloads in multi-tenant environments.

The software challenge

While the launch of Helios might help AMD close the hardware gap with Nvidia’s rack-scale systems, it will be the software compatibility that will be the real driver of enterprise adoption.

For this, AMD is expanding its ROCm AI software platform too, which supports frameworks including PyTorch, TensorFlow, and JAX, for enabling high-throughput inference and efficient distributed training while preserving familiar developer workflows.

Jain stated While hardware parity or superiority in memory bandwidth is achievable, software maturity remains the key differentiator for Nvidia. The Nvidia’s CUDA software has a 15-20 year head start, and almost every AI tool, tutorial, and codebase defaults to it.

He added software has been AMD’s weak spot. AMD has improved  ROCm a lot but it still lags behind on the newest, most specialized optimizations, and setup is more complicated. For everyday AI work, ROCm is usable but for cutting-edge performance, CUDA still leads.

Evaluating the trade-offs

For CIOs evaluating AI infrastructure, Helios launch brings in another option to a market that has largely revolved around Nvidia’s dominance. But when considering Helios, CIOs will have to evaluate factors such as performance, software readiness, deployment models, procurement timelines and total cost of ownership before committing to a platform.

While AMD has not publicly announced a specific price tag for the Helios, Jain believes it to be noticeably cheaper to buy and run with lower chip prices and lower power use per GPU.

“It gives companies a real second option besides Nvidia, easing supply shortages and giving leverage in negotiations. The catch is software, where teams need to check whether their AI tools run well on AMD’s stack, since some advanced tools are still CUDA only,” Jain said.

For CIOs planning to deploy both, Jain warns the two systems can’t be plugged together into one combined machine as they use different, incompatible connection technology. But companies can and do run both side by side in the same data center, just as separate systems handling different jobs.

Something interesting happened during an internal evaluation of AI meeting transcription tools at Tungsten Automation. The products worked. They weren’t bad. But sitting across from the pricing, we kept asking the same question: what exactly are we paying for?

We already had a secure enterprise AI environment. Building a meeting summary workflow took days, not months. We customized the outputs, injected our own internal context, and controlled security our way instead of working around someone else’s roadmap. We built it. It works better. We own it.

That’s not a knock on those vendors. It’s a signal of something more fundamental happening across enterprise software.

The moat was never the product

For two decades, SaaS rode a favorable asymmetry: building internal tools was hard, integrations were messy, and even modest automation required developers and long timelines. Buying was faster and cheaper than building. That asymmetry fueled the explosion of SaaS into every corner of the enterprise stack.

AI is collapsing that asymmetry. Large language models and agentic workflows can orchestrate APIs, move data between systems, generate interfaces, and automate business logic with a fraction of the engineering effort required even two years ago. The integration friction that once protected entire product categories is evaporating.

The vendors most exposed are not the deeply embedded enterprise platforms. They’re the lightweight workflow layers, the products that essentially put a polished interface on top of accessible data and relatively straightforward processes. Reporting dashboards. Meeting tools. Narrow productivity applications. These products created value by simplifying implementation. That rationale is getting harder to sustain when implementation is no longer the real barrier.

Here’s the part most analyses miss: it’s not just that AI makes development faster. It’s that agents change the integration model entirely. For 30 years, enterprise software was built for humans navigating UIs. Agentic systems don’t use UIs. They call APIs, read from multiple sources simultaneously, and move data freely across systems. The switching costs that once made incumbent software sticky are collapsing, because an agent doesn’t care which UI it used last quarter.

The SaaS that survives

The question isn’t whether SaaS survives. It’s which SaaS survives.

The companies with durable positions are not the ones with the cleanest interface. They’re the ones that transfer operational risk customers genuinely cannot absorb themselves. Compliance. Regulatory certification. Accumulated domain expertise. Liability.

Think about compliant invoicing across 140 countries. That’s not a workflow someone builds in a sprint. The certifications alone take years. A single regulatory change in one jurisdiction can break an AP process for a global enterprise overnight. Customers don’t pay for that capability because it’s technically complex. They pay because they cannot afford to own the risk of getting it wrong.

That’s the distinction that matters: AI lowers the cost of building software. It does not lower the cost of absorbing risk. The vendors who understand this are building durable businesses. The ones who don’t are quietly subsidizing their customers’ internal build programs.

Software sells features. Platforms sell accountability.

The prototype trap

The danger for enterprise buyers right now is overcorrection. Every successful prototype looks like a cost-saving opportunity. Very few survive the jump to production.

Building a workflow with generative AI is becoming straightforward. Maintaining it is not. Models evolve. Outputs drift. Governance requirements tighten. What worked cleanly in a controlled environment behaves differently at scale, and the failure mode is worse than traditional software. Rule-based automation, when it fails, fails obviously. Agents fail silently, confidently, at scale, often with a completely reasonable-sounding explanation.

Engineering teams that take on AI-powered systems need to solve for observability, model drift, access controls, audit trails, and long-term maintenance ownership. In regulated industries, they need to demonstrate exactly how the system reached every decision. That’s not a weekend project. That’s an ongoing operational commitment that compounds over time as models change and regulatory requirements evolve.

Before a team decides to replace an external platform with internal AI tooling, the honest question isn’t, “Can we build this?” The real question is, “Are we prepared to own this in production, for years, as the underlying models change beneath us?” Sometimes the answer is yes. Often the answer is no, and the true cost only becomes visible after the vendor contract is canceled.

Build vs. partner: a sharper frame

The build vs. buy framing has always been too binary. The right question is build vs. partner.

Partner for the capabilities where risk transfer, regulatory complexity, and domain expertise create genuine value your team cannot replicate. Build for the capabilities that actually differentiate your business from your competitors. Don’t burn your best engineers rebuilding compliant invoice processing or production-grade document extraction. Those aren’t competitive advantages. They’re table stakes, and someone else has already paid the cost, across decades, to make them reliable.

The organizations getting this right are honest about where they create unique value. They focus development there, and partner for everything else. The ones getting it wrong are vibe-coding solutions to non-differentiating problems while their actual competitive moat goes unattended.

The true value of software

We’re not watching the death of SaaS. We’re watching the end of the friction-based value proposition: the idea that software is worth renewing because integration used to be painful. That rationale is largely gone.

What survives is software that does something customers cannot reasonably replicate internally: absorb risk, maintain regulatory compliance, deliver operational reliability at scale, and bring genuine domain expertise into a production-grade system that someone else already stress-tested for years.

The vendors who recognize this are already repositioning around accountability, governance, and outcomes. The ones who haven’t will find the next renewal conversation noticeably harder.

Software sells features. Platforms sell accountability. That distinction is about to separate a lot of winners from a lot of cautionary tales.

—

New Tech Forum provides a venue for technology leaders—including vendors and other outside contributors—to explore and discuss emerging enterprise technology in unprecedented depth and breadth. The selection is subjective, based on our pick of the technologies we believe to be important and of greatest interest to InfoWorld readers. InfoWorld does not accept marketing collateral for publication and reserves the right to edit all contributed content. Send all inquiries to doug_dineley@foundryco.com.

Site reliability engineers (SREs) have the tough assignment of resolving thorny performance and reliability issues. But their primary mission is to provide devops teams with operational insights and to suggest implementation improvements on business system performance, security, and overall robustness.

Google introduced its SRE playbook in 2003, but it took some time for the role’s definition, tools, and techniques to become mainstream. Startups were the first to adopt observability for cloud-native applications and create dedicated SRE positions. As tools matured and SRE responsibilities became more clearly defined, larger enterprises assigned SREs to work as a bridge between devops and IT ops teams to improve resilience across a wider range of applications, APIs, and data pipelines.

SRE is a career path for multidisciplinary engineers with strong investigative instincts, sharp data analytics skills, and the temperament to perform under pressure. It has become a critical responsibility as tech became mission-critical for enterprises, and it is a growing role in the genAI era as more businesses deploy AI agents.

But the critical need for resiliency and greater technological complexity brings new challenges for SREs. According to the 2026 State of Production Reliability and AI Adoption report, 44% of respondents experienced an outage linked to ignored or suppressed alerts in the past year, and 35% report their engineers occasionally ignore or dismiss alerts due to alert fatigue. More than 70% of alerts received are not actionable, according to 57% of organizations.

So, is AI making the SRE’s role easier and helping businesses run more reliable technology operations? On the other hand, AI is also driving complexity, as companies deploy genAI tools and AI agents across more business functions and seek to automate more decision-making across operations.

AIops and agentic ops aid SREs

Over the past decade, SRE responsibilities have become somewhat easier through improvements in monitoring platforms, observability practices, tools for centralizing operational data, and AI applied in IT operations (AIops). But during the heat of resolving an outage or performance issue, it’s not easy to correctly identify what system triggered the issue versus other downstream systems impacted by it.

According to the Komodore 2025 Enterprise Kubernetes Report, 79% of production incidents originate from recent system changes, including deployments and changes to compute environments. But the other 21% of incidents stem from issues outside of the business’s control, including network failures, third-party changes, and cloud provider failures.

“SREs using AI capabilities succeed or fail in the moment an incident unfolds, when engineers are deciding what to investigate next,” says Itiel Shwartz, CTO at Komodor. “If the system streamlines root cause detection, connects signals to recent changes, and explains its reasoning in a way engineers recognize, it earns trust. If it adds uncertainty or demands extra validation, it gets sidelined, regardless of how bespoke the model behind it may be. What’s less obvious is what it takes to make AI for SREs work in production, and how different that reality is from prototypes, demos, or early internal builds.”

AIops is not a new capability, especially in using machine learning to correlate logs, metrics, and traces across monitoring and alerting systems. IT service management and SREs have been using AIops to reduce the mean time to resolve incidents and to perform accurate root cause analysis (RCA) efficiently. Agentic ops is the next wave of genAI operational capabilities, including tools for monitoring AI agents, managing their access rights, and detecting AI model accuracy drift.

 “AI is useful during major incidents because it can pull together a lot of context into a few clear sentences, which is exactly what an SRE needs in the moment,” suggests Shani Shoham, chief revenue officer at OpenObserve. “The complexity of architecture and the different tooling make it easier for AI than for a human, but autonomous resolution is still a way off.”

AI’s impact on people and burnout

The business pressure to keep systems up, secure, and performing well is a 24/7 stressful responsibility. According to The SRE Report 2025 from Catchpoint, 36% of SREs often or always experience elevated stress during an incident, and 28% said the stress persists even after the incident is resolved. AI capabilities may prove to be a game-changer in helping SREs avoid burnout and reduce stress.

“AI can improve RCA by taking in a much larger incident context than any engineer can hold at 3am, reasoning across traces, logs, metrics, deploys, config changes, alerts, ownership, and recent production behavior,” says Noam Levy, founding engineer and field CTO at Groundcover. “Beyond attempting a full RCA, its immediate value is distilling the signals that actually matter, reconstructing a clear timeline of cause and effect, and helping engineers separate correlation from likely causality. Once a fix is deployed, agents can also verify remediation by comparing pre- and post-fix behavior, but this depends on broad access to rich, correlated production signals and a cost model that does not discourage adoption or experimentation.”

Not only are incidents resolved faster and with less stress, but AI can also free up SRE time to focus on proactive work and create a career path for junior developers into SRE roles. Quais Taraki, CTO at EDB Postgres AI, adds, “AI reduces toil by automating repetitive tasks while accelerating incident resolution through copilots that correlate signals across distributed systems, allowing SREs to focus more on resilience strategies like chaos engineering and failure analysis.”

AI can have long-lasting operational impacts, especially for organizations looking to deploy more mission-critical technology and AI capabilities. Two longer-term benefits of AI for SREs are reducing the number of bridge calls needed for incident response and the number of engineers required in “war rooms” to coordinate root cause analyses.

“When something goes wrong, AI that guides SREs can do the full analysis, get to the root cause, and perform the remediation,” says Spiros Xanthos, founder and CEO of Resolve AI. “AI also helps avoid many escalations, and when escalations are needed, it targets the right people from the network, infrastructure, and the application teams. AI for SREs centralizes operational intelligence, exposes tribal knowledge, and can guide more junior developers.” 

AI agent reliability

While AI capabilities have been a net positive in helping SREs improve system reliability, the growth of AI code generators, vibe coding, and spec-driven development is adding to their workloads. According to one study, 41% of all global code is now AI-generated, and Gartner predicts that 40% of new enterprise production software will be created using vibe coding techniques by 2028.

But coding velocity is creating new issues for SREs as AI pull requests have 1.4 times more critical issues and 1.7 times more major issues, according to CodeRabbit. “AI-assisted development has created an unprecedented velocity of code reaching production, expanding surface area, edge cases, and failure rates faster than traditional SRE practices can absorb,” says Vinod Jayaraman, cofounder and CTO at NeuBird AI. “The speed of shipping has far outpaced the speed of understanding what breaks in production. To close this loop, SREs need enterprise agents that can capture precise diagnostic context, including correlated traces, service dependencies, and anomaly timelines, and structure it as actionable input for the engineers and AI coding tools responsible for the fix.”

The growing number of AI agents deployed to production creates new challenges. AI agents are not just code; they have multiple failure points. They are built using language models, connect to proprietary sources for context, and integrate with Model Context Protocol servers to support more complex workflows. Changes are ongoing and not deployment events, so the SRE’s job of identifying the source of performance and accuracy drifts isn’t trivial. 

“Traditional SRE was built for systems that fail in reproducible ways, but agents fail differently and drift when a model provider pushes an update, and behavior shifts silently with no baseline for comparison,” says Mohammed Aboul-Magd, vice president of product at SandboxAQ. “Most organizations can’t even answer the basics: how many agents are running, what they have access to, and whether they’re still doing what they were built to do.”

“Every time a senior engineer leaves, they take years of learned failure patterns with them, and the next outage starts from square one,” adds Ronak Desai, cofounder and CEO at Ciroos. “Using AI for compounding operational memory changes that, and every incident your system resolves, the AI learns it.”

SREs should take a leadership role in emerging best practices, including defining their standards for AI agent non-functional acceptance criteria, observability practices, and release-readiness criteria. SREs should update their service-level objectives (SLOs) and define error budgets for AI agents in production.

Ryan Downing, vice president and CIO of enterprise business solutions at Principal Financial Group, says, “Standard SLOs and error budgets give teams the guardrails, and AI helps interpret the telemetry against those targets, reducing noise so engineers can get to the real issue faster and automate parts of remediation before customers are impacted.”

AI raises the SRE’s business impact

The more dramatic shift in site reliability engineering is an evolution of its business scope. IT leaders focus on uptime, performance, and issue resolution, as well as understanding their impacts. Business leaders will look to IT and SREs to identify, determine root cause, and remediate a broader class of issues, including rogue AI agents and the impacts of rapidly deploying new agentic capabilities. 

“AI agents are handing SREs categories of problems they’ve never had to solve before, specifically failures defined in business terms, not technical ones,” says Blake Sherwood, distinguished technologist for AI and platform strategy at Smarsh. “Traditional reliability engineering is built around latency, errors, and crashes, but agents now fail due to skipped compliance steps or outcomes that looked fine technically but were wrong contextually. Most SRE teams aren’t wired for that yet.”

The question is whether SREs with AI-augmented tools can keep up with the velocity, complexity, and business urgency of deploying new AI business capabilities.

Every enterprise AI conversation seems to begin with the same question: Which model should we use?

I understand why. Models are visible. They have names, benchmarks, release notes, pricing pages and impressive demos. They are easy to compare in a leadership meeting. One model promises better reasoning. Another offers a larger context window. Another appears faster, cheaper or more specialized.

But after years of working around enterprise platforms, integration layers, cloud migration, middleware, production operations and mission-critical systems, I see the AI conversation differently.

The model matters. But it is not where most enterprises will struggle next.

The next AI bottleneck is the infrastructure behind the model.

I do not mean only GPUs, cloud capacity or data storage. I mean the full enterprise operating layer that allows AI to work safely in the real world: data pipelines, identity, APIs, messaging, observability, security controls, deployment automation, cost governance, auditability, support ownership and recovery design.

That layer is what determines whether AI remains an exciting experiment or becomes a trusted business capability.

Pilots hide the hard part

Most organizations can build an impressive AI pilot. A small team can connect a model to a dataset, create a workflow and show a use case that works well in a controlled setting.

The harder part starts when that pilot moves into a real production process.

That is when practical questions show up. Who owns the data quality? What systems can the AI access? How do we trace which prompt, policy or retrieval flow produced a specific answer? What happens when an API slows down, a queue backs up or a downstream system is unavailable?

To me, these are not model problems. They are infrastructure problems.

This is where many enterprises are now headed. The first phase of AI was experimentation. The next phase is operationalization, and that is where the real gap becomes clear.