July 23, 2026
ServiceNow Says Customers Have Stopped Buying AI Pilots and Started Paying for Resolutions
ServiceNow closed the second quarter of its 2026 fiscal year with subscription revenues of $3.8 billion, up 23% year over year. President and CFO Gina Mastantuono told analysts that finance leaders have “watched budgets burn on pilots that demo beautifully and never touch the P&L,” and summarised the change in purchasing behaviour in a single line: “Customers aren’t paying us for tokens, they’re paying for resolutions.”
Chairman and CEO Bill McDermott said the company had crossed $1 billion in AI annual contract value and remained on course to exceed its $1.5 billion target by the end of 2026.
Voice Moves from Pilot to Production Volume
The disclosure with the most immediate relevance to contact centre operators concerned a single unnamed carrier.
“A large airline has gone all in running their customer service voice calls on ServiceNow’s voice AI CRM agents,” McDermott said. “This is live in production and handling 5 million annual voice calls in year one alone.” He added that customer satisfaction was “off the charts,” though no figures were attached to the claim.
Voice has been the slowest channel to absorb generative AI, largely because the data written back into systems of record during a call determines whether anything useful happens afterwards, a dependency examined in coverage of Vonage’s native integration with ServiceNow Voice earlier this year. A five-million-call deployment running end to end on a workflow platform rather than a dedicated CCaaS stack is an unusual proof point, and competitors including Salesforce, which recently brought AI agents, live voice and CRM data into a single native contact centre platform, will have to answer for it.
Level1 Automation and the Labour Arbitrage Argument
Amit Zavery, president, chief product officer and chief operating officer of ServiceNow, gave the most concrete numbers of the call on the company’s L1 service desk AI specialist, which has more than 40 customers in deployment. “We’re talking about closing down close to 80%-85% of the service requests without having any human interaction,” he said, adding: “For example, some of these requests might take two days for humans to get to a resolve. We’re doing that in 20 minutes.”
Zavery was unusually direct about the commercial logic underneath that. “It’s also labor arbitrage. We also get to monetize the labor cost, which employers have to pay the employees. Now they can do this in a much cheaper way, we can take away the labor cost as well.” Vendor messaging around agentic service has generally leaned on capacity release rather than headcount economics, including when ServiceNow first launched its Autonomous Workforce and AI customer service specialist.
The City of Raleigh was named as the first local government to run the L1 specialist in production without any in-house AI engineering team behind it, moving, in McDermott’s description, “to become fully autonomous, one ticket category at a time.”
CRM Reaches $2 billion in ACV
McDermott confirmed that the CRM business has passed $2 billion in annual contract value, with net new ACV growth accelerating both sequentially and year over year, sales CRM average deal size doubling, and the platform on track to process over two billion service cases this year.
Among the wins cited were a full front-office replacement of an incumbent CRM deployment closed by a partner in two months, a North American telecom infrastructure provider expecting to scale quote volume threefold with no added headcount and cut repricing from weeks to hours, and a regional financial institution consolidating loan origination workflows.
Governance as the Condition of Adoption
AI Control Tower now has more than 500 customers live within six months of launch, according to Zavery, who described kill-switch functionality for rogue agents and identity governance drawn from the Veza acquisition. Enterprise buyers are increasingly treating continuous oversight rather than deployment speed as the constraint on scaling agentic service, a theme running through Cognizant’s integration of its Neuro AI Trust platform with AI Control Tower in June and through the product announcements made when ServiceNow committed to autonomous enterprise AI at Knowledge 2026.
McDermott characterised buyer sentiment: “Most customers are now completely allergic to anything that looks like a project. They only want deterministic. ServiceNow only does deterministic.”
Pricing Tied to Outcomes Rather than Consumption
President and CFO Gina Mastantuono came back repeatedly to return on investment, telling analysts that CFOs have “watched budgets burn on pilots that demo beautifully and never touch the P&L.”
Her summary was the sharpest line of the call: “AI that only advises is a cost. AI that completes the work is a return,” also adding: “Customers aren’t paying us for tokens, they’re paying for resolutions.”
Price uplift on the new AI-native SKUs is running at 20% to 30%, with ProPlus above 30%. Fifty percent of net new business is already non-seat based, though the company continues to offer seat pricing because customers prefer the predictability.
Adoption Pace and What Comes Next
Customers with agentic AI in production have grown ninefold over nine months, deal volume among first-time agentic buyers rose over 45% year over year, and deals containing five or more AI products grew 5.5x. Mastantuono said the company is already tracking ahead of its target for AI to reach 30% of ACV by 2030.
McDermott also trailed an unannounced product: a conversational service desk experience with “no tickets, and AI-coded automation,” sold through a product-led motion aimed at what he called the Fortune 500,000, with several customers already in beta.
