At Knowledge 2026 in Las Vegas, ServiceNow execs spent much of the event talking about governance, security, and agentic AI. But beneath the product launches and demonstrations sat another theme that kept surfacing across discussions: enterprises are struggling to understand what AI is actually costing them – and whether the returns justify the spend.
ServiceNow CEO Bill McDermott described it as an “AI blind spot” developing across enterprise technology estates as organizations rapidly deploy copilots, agents, and large language models across disconnected systems.
“Your CFO approved the spend and can’t find the ROI. Your token bill goes up every month,” McDermott told the Knowledge audience.
Where’s the Value in Your AI?
McDermott’s comments reflect a growing concern emerging across enterprise AI adoption. While organizations continue to invest in gen AI tools and agentic systems, many are discovering that operationalizing AI at scale introduces a new layer of financial management complexity. It’s centered on model consumption, token usage, workflow execution costs, and proving measurable business outcomes.
ServiceNow execs repeatedly described this as a governance and visibility problem rather than a technical one.
“The number one question every CFO is asking is, where’s the value?” ServiceNow’s president, chief product officer, and COO, Amit Zavery, told the media ahead of the event. “Runaway model spend is one of the biggest pain points enterprises are facing right now. If you can’t see what AI is costing you and prove the value, you can’t justify scaling it.”
That message sits at the center of ServiceNow’s expanded AI Control Tower strategy, announced at the event as part of a broader push to position the company as what McDermott called “the AI control tower for business reinvention”.
AI Adoption Outpacing AI Visibility
The company argued that many organizations already have significantly more AI in production than they realize, particularly as employees and departments adopt external AI tools independently.
“What we’re hearing consistently from customers is that most organizations have more AI in production than they’ve inventoried or accounted for,” said Zavery.
The concern reflects the wider rise of so-called “shadow AI” across enterprises. According to Gartner, 69% of organizations either suspect or have evidence that employees are using prohibited public generative AI tools.
At the keynote, McDermott described enterprise estates where AI has been “bolted onto every application like a sidecar” without centralized oversight.
“Six out of 10 companies are actually using agentic AI, but only one out of 10 have built anything autonomous,” he said. “Most are paying for a capability they haven’t even come close to unlocking.”
That gap between experimentation and operational maturity is becoming increasingly expensive. AI consumption models differ sharply from traditional enterprise software licensing, introducing variable costs tied to inference, API calls, and token consumption. As organizations deploy multiple models and agents across different providers, tracking usage and value becomes significantly harder.
Governance and Cost Control Move Up the Agenda
According to IDC analyst Snow Tempest, the issue extends well beyond ServiceNow’s own customer base.
“AI governance, cost visibility, and operational control are important because the underlying models and pricing are still evolving so rapidly,” she told NowBen. “In my research, organizations see potential for AI to help realize their overall goals, but they want to manage risks and costs.”
Tempest said IDC research shows organizations are prioritizing governance and security concerns ahead of specific AI functionality.
“In a survey I conducted, I found that organizations had two outstanding priorities for AI implementation: data security and effective decision making. Those two far outranked any specific capabilities or features,” she said.
“There were also two major impediments to AI implementation in IT service management and enterprise management, which were concerns about security and cost.”
The findings support the idea that enterprise AI conversations are shifting from experimentation toward operational management and financial oversight.
From Cloud FinOps to AI FinOps
The emerging challenge draws parallels with the rise of cloud FinOps over the last decade, where enterprises struggled to manage rapidly scaling and decentralized cloud consumption. ServiceNow is now effectively positioning AI governance as the next major operational management layer for enterprises.
Its AI Control Tower platform is designed to discover AI assets across hyperscalers and enterprise applications, track adoption and usage, monitor governance and compliance, and connect AI activity to business outcomes.
The company said the system can monitor models running across AWS, Microsoft Azure, Google Cloud, and third-party SaaS environments, while also measuring ROI and operational efficiency.
“AI Control Tower enables you to track everything, adoption, consumption, ROI, productivity, and cost savings all in one place,” said Zavery.
Tempest added that organizations are already beginning to experience what could become a broader “AI FinOps” challenge.
“IDC finds that AI budget pressure is a persistent concern for organizations worldwide,” she said. “Organizations are also interested in how AI can actually help them control costs, whether it’s by making processes more efficient, or by helping them get a handle on expenses like unused software licenses, or on the costs of AI itself.”
CFO Scrutiny Is Intensifying
ServiceNow’s messaging also reflects growing financial pressure around enterprise AI spending more broadly. In the keynote, McDermott argued that intelligence itself is becoming commoditized as model access expands across the market.
“The real competitive differentiator is the orchestration surrounding the models,” he said. That orchestration argument increasingly includes cost control.
ServiceNow repeatedly positioned AI Control Tower not just as a governance platform but as a financial visibility layer for enterprise AI operations. Demos have focused on dashboards showing AI productivity gains, model consumption, and business value calculations tied to workflow outcomes.
Indeed, the firm is calculating $250 million in productivity gains from Microsoft agents before subtracting associated costs to demonstrate ROI.
The company also highlighted its own internal AI usage as proof of concept. McDermott said ServiceNow had tracked half a billion dollars in internal AI value during 2025 through its ‘Now on Now’ deployment strategy.
“91% of service requests are supported by AI,” he told attendees.
Still, proving enterprise AI value remains difficult. Tempest said many IT leaders are still struggling to connect AI spending directly to business outcomes.
“I consistently hear from IT leaders that they need to be able to ‘tell the story’ of how their AI implementations will benefit their organizations. The demand to be able to show ROI or other outcomes increases as the options for deploying AI expand, because organizations need to be able to choose the best places to invest and prioritize.”
Final Thoughts
ServiceNow has an opportunity to position itself not just as another AI vendor, but as the platform enterprises use to govern, secure, and financially manage increasingly complex AI estates.
The next challenge, ServiceNow argues, is making AI economically sustainable at enterprise scale.