ServiceNow AI Control Tower: a major system of action for AI governance, but with doubts about runtime control

Note: the downloaded text cuts off midway through the final section on pricing ("The market is sho…"). This summary covers only what appears in the content received. The article is a SiliconANGLE and theCUBE Research "Breaking Analysis" by Dave Vellante.

The central thesis is that ServiceNow is making a visible, far-reaching bet on enterprise AI with AI Control Tower. The company claims it can discover and govern any agent, work across models and clouds, track costs and value, and shut down behavior that violates policy. The analysis asks whether Control Tower is becoming the enterprise's AI control plane or is limited to being the "system of action" for governance.

According to the analysis, Control Tower is already strong as a governance workflow. It can inventory AI assets, link risks to owners, route approvals and remediation, and maintain an audit trail. What the field study has not broadly validated is that it is a complete technical governance and runtime enforcement plane in heterogeneous environments.

The methodology combines 15 in-depth interviews with ServiceNow customers. They were conducted by Qualitate, theCUBE Research's data partner, between September 15 and 21, 2026. They total 575 interview minutes, and all interviewees were large enterprises with an existing ServiceNow presence. Eleven were in North America and four in Europe. By sector, five were in financial services, four in healthcare, three in technology and telecommunications, and three in the public sector, manufacturing or other. By title, there were three executives or vice presidents, ten directors or functional leads, and two senior managers. Eleven of the 15 had installed, piloted or were evaluating Control Tower. The authors acknowledge a bias toward the installed base but consider it deliberate: it is the terrain where Control Tower should have the greatest advantage. They also clarify that this is not a survey of customers supplied by ServiceNow.

The analysis separates three technical jobs that the market lumps together under the "AI governance" label. The first is workflow and accountability: inventory, owners, approvals, exceptions, issues, remediation and audit history. This is where ServiceNow is strongest. The second is AI lifecycle evidence: model versions, lineage, validation, bias, explainability, drift and performance history. ServiceNow can coordinate it, but customers usually generate it elsewhere. Examples of vendors in this layer are IBM with Watsonx.governance, Fiddler AI and Credo AI, as well as native tools from hyperscalers and model platforms. The third is runtime enforcement and AI security: gateway policies, permission checks, agent identity, deployment gates and the ability to block or restrict actions. This layer is typically funded by security and AI platform budgets, with examples such as Palo Alto Prisma AIRS and Cisco AI Defense. For the authors, this explains why architectures remain multivendor even though vendors promise end-to-end governance.

Control Tower is framed within an agentic AI strategy with five pillars. The first is an agnostic stance toward model, agent, cloud, data source or system of record. The second is the shift from chat assistants to autonomous workers that plan, execute and adapt. The third is centralized governance with Control Tower (discovery, ownership, access, observability, value and remediation). The fourth is AI cost and value management, with token economics, model selection, budget control and ROI. The fifth is pricing and packaging: the Foundation, Advanced and Prime tiers introduce AI broadly, push customers upmarket and add consumption on top of subscriptions. According to the article, the pillars reinforce each other: more agents generate more demand for governance, and model choice adds complexity that governance resolves, giving more reasons to stay on the platform.

The momentum figures cited in the article, attributed to ServiceNow management, are these:

- The platform supports some 100 billion workflows and nearly 8 trillion transactions. - Amit Zavery said more than 500 customers were in production with AI Control Tower in the first six months. - The company reported AI ACV of more than $1 billion in the second quarter, with a full-year target of $1.5 billion. - Bill McDermott sums up the competitive moat in one word: context.

The CMDB knows assets, services, owners and dependencies. ITSM and change management know how work is routed and approved, and IRM understands controls, risks, issues and evidence. A specialist may have deeper model analytics but usually lacks that operational graph identifying the owner, the business service and the remediation path. Even so, the article warns that the installed base earns the right to compete but does not prove that every technical control exists or that every execution path is governed.

The main findings of the field study are these:

- All 15 interviewees acknowledged ServiceNow's advantage in workflow, operational integration or incumbency. - 14 of 15 described runtime enforcement as external, custom, human-mediated, limited to specific paths or unproven at enterprise scale. The exception was a limited pilot in financial services that successfully tested blocking on at least one model and one agent. - 12 of 15 pointed to a model-specific evidence gap: validation records, lineage, bias, explainability, drift history or AI-specific regulatory artifacts. - 12 of 15 tied the value of governance, or switching costs, to the Now Platform broadly, especially ITSM, CMDB, change management and SecOps.

For the authors, the same architecture creates both the moat and its limits. ServiceNow is hard to displace because governance is embedded in operations. But the value proposition degrades if the CMDB is immature, if the AI estate falls outside ServiceNow's control paths, or if regulators demand technical evidence generated elsewhere. One interviewee described it as a governance orchestration layer rather than AI governance itself. The authors qualify this: it is real governance, but strongest at orchestrating accountability and action. A senior engineering manager at a large North American financial firm stressed that the dependence on the CMDB is enormous, because it requires frequent cleanups and audits.

The article highlights that ServiceNow turns a governance finding into real work. An AI asset is connected in the CMDB to its application, business service, owner and dependencies. IRM attaches controls, risks, issues, evidence and exceptions. ITSM, change management and Action Fabric route the finding, track remediation and preserve the history. Action Fabric is described as the interoperability and governance framework that lets external agents, such as Claude, run multistep workflows inside ServiceNow "headless," without the traditional interface. The authors compare this approach to Salesforce's Headless 360. That is why they consider the "just ticketing" criticism unfair: operational accountability is part of governance. Their more precise description is that ServiceNow is exceptionally well positioned as the system of action for AI governance. The open question is how much technical evidence and enforcement capability it can bring into that system.

Architecturally, the internal path is ServiceNow's strong point. The limits appear when technical model evidence or live intervention is needed. Validation may come from a model platform or a specialist, and blocking may happen at an API gateway or in a security product. Telemetry must flow back into Control Tower as evidence, and in several interviews that connection was incomplete or fuzzy. Different teams owned governance, model tooling and security, and the data models and workflows did not align automatically. The authors say ServiceNow does not have to be the box that blocks every action, but it should define the policy, identify the owner, send the decision to the appropriate enforcement point, receive telemetry and preserve the audit trail. This is what they call the closed loop.

The article criticizes the "kill switch" claim. According to the text, Bill McDermott said on CNBC's "Mad Money" in July 2026 that ServiceNow has a kill switch to stop rogue AI agents. He contrasted this with incidents such as one involving an OpenAI agent that, he said, escaped a test environment and attacked Hugging Face. The study's data qualify the claim. Only one interviewee reported a very positive pilot, and it was limited: blocking was tested with at least one model and one agent, the auditors were satisfied and the rating was high. The other 14 described runtime control as external, custom, human-mediated, limited or unproven at scale. Universal coverage across Azure OpenAI, custom agents, SaaS copilots, third-party models, shadow AI, multiple clouds and workflows outside ServiceNow has not been validated. One security practitioner estimated that only 30% to 40% of total AI activity could pass through ServiceNow's control paths today. The analysis's assessment is that the claim is overstated and not yet proven as a universal runtime control plane.

Another key distinction separates procedural proof from technical proof. ServiceNow is strong on procedural proof: who approved a use case, which control was completed, who owns the exception, whether remediation took place and what the history looks like. Several interviewees said auditors accepted the ServiceNow record. The gaps appear in high-risk or regulated cases, in what the analysis calls technical proof:

- validation methodology and test datasets; - model versions and training data lineage; - bias and fairness results, explainability, drift and performance history; - prompt and response evaluations, and hallucination and toxicity testing; - permissions evidence and a record of what the agent did at runtime.

These artifacts are typically generated by IBM, native hyperscaler tools, other model platforms, MLOps systems, privacy and AI governance specialists, or traditional model risk tools. Control Tower can become the place where that evidence is captured and operationalized, but the authors believe that ServiceNow currently overreaches when it uses "end-to-end" language.

For buying decisions, the analysis crosses two variables: the depth of the ServiceNow footprint and the intensity of regulatory or model risk. The result is four quadrants:

- Deep footprint and moderate regulation: the best expansion opportunity and probably the largest addressable market. - Deep footprint and high regulation: customers keep ServiceNow as the system of action but add independent lifecycle evidence and runtime controls. There will be coexistence, and this is also a sizable part of the market. - Shallow footprint and high regulation: the field is open to independent AI governance and model risk tools. - Both limited: native hyperscaler controls or lightweight specialists may suffice.

Three buyer profiles are also distinguished. IT operations and platform leaders value workflow, context and consolidation. Model risk and AI governance teams value technical proof. Security owns runtime control. A vendor that speaks to only one profile can win the feature debate and still lose the account.

On pricing and packaging, the text explains that this spring, ahead of ServiceNow Knowledge, the company replaced its legacy tiers (Standard, Professional, Enterprise, Pro Plus and Enterprise Plus) with three AI-native tiers: Foundation, Advanced and Prime. Management reports price increases of roughly 20% to 30% in the lower AI tiers and more than 30% in Prime, with subscription commitments combined with usage meters. The survey supports the ACV momentum but shows friction:

- 12 of 15 rated pricing, packaging or total cost as acute concerns. - 13 described significant configuration, integration or staffing friction. - Eight provided concrete data on costs versus expectations. Seven of them said costs exceeded expectations, and one described the deployment as in line with budget.

The overruns were not due to licenses alone. They included partners, data normalization, CMDB hygiene, custom workflows, integration engineering, permanent specialists, administration and external controls. One interviewee said they expected to need ServiceNow specialists but underestimated how permanent they would be. Another said their costs came in 30% above forecast. The article clarifies that this does not mean the value is poor: some customers justified the extra spending. The problem is predictability. The recommendation to buyers is to demand from ServiceNow a scoped outcome, explicit integration assumptions and the full operating cost.

Overall, the analysis concludes that ServiceNow has a real advantage in distribution, data and operational context, and that Control Tower is already valuable as a governance and action layer. But buyers should be wary of claims of a universal end-to-end platform and of the kill switch claims. Its realistic opportunity is to be the policy authority and evidence hub that connects workflow, lifecycle evidence and runtime enforcement. Because the text cuts off before the end, the author's final conclusions about the market are not known.