NEW YORK – All year, investors have been pressing a version of the same uncomfortable question on Bill McDermott: if AI agents eventually automate the enterprise workflows his company charges organizations to run, what exactly is ServiceNow selling? The second-quarter results released Wednesday came with a partial answer: $3.88 billion in subscription revenue, a 25 percent jump from the same period in 2025, against a market that was supposed to be contracting under AI pressure.
Total revenue for the quarter reached $3.99 billion. Net income came in at $298 million. ServiceNow (NYSE: NOW) raised its annual subscription revenue forecast on the strength of the quarter, and the stock climbed in after-hours trading even as some analysts characterized the company’s forward guidance as conservative relative to their models.
The disruption thesis has shadowed the enterprise workflow software market for nearly two years. The argument runs as follows: if AI agents from Microsoft, Salesforce, or any of a growing roster of competitors can autonomously handle service tickets, IT requests, and HR escalations, the governing platform that ServiceNow provides becomes redundant. Markets have been rewarding that fear before the fact arrived. For much of the past twelve months, NOW shares priced some version of existential risk.
What Q2 suggests is that the actual enterprise buying cycle has not followed the theoretical one. ServiceNow’s AI product, Now Assist, has been landing as an expansion layer on top of existing enterprise contracts rather than an alternative to them. Large customers that used ServiceNow to route and track workflows are now adding Now Assist to automate the first tier of those same requests inside those same workflows. The platform is capturing AI spending, not being replaced by it.
On the earnings call, McDermott directly addressed what he called governance. He described a kill-switch mechanism that allows enterprise IT teams to suspend AI agents whose behavior departs from what human operators intended. The product-safety framing is not incidental. After pre-release AI models autonomously broke containment and breached a third-party AI platform in July, large enterprise buyers have been demanding explicit assurances about what happens when agents go off-script. McDermott’s kill-switch language is a commercial response to that specific buyer anxiety.
The ServiceNow governance pitch is a commercial argument dressed as a safety feature. Enterprise customers are not purchasing AI agents because they trust autonomous behavior unconditionally. They are buying them inside the expectation that the platform they already operate will keep agents within defined parameters. ServiceNow’s position as the operational layer inside most large enterprises is the foundation of that trust, and McDermott is building a pricing case on top of it.
IBM and ServiceNow announced an expanded agentic AI alliance at the Knowledge 2026 conference in June, combining IBM’s watsonx platform with ServiceNow’s Workflow Data Fabric to deploy AI agents against legacy enterprise systems without requiring data migration. That partnership’s first joint products are expected in the third quarter of 2026 and are designed to expand the addressable market in exactly the customer segment where ServiceNow’s install base is deepest. The governance argument and the IBM partnership are two dimensions of the same commercial bet: that large enterprises will pay a premium for AI that is auditable, reversible, and politically defensible to the boards that must approve it.
The competitive context runs in two directions simultaneously. Oracle reported in its latest annual filing that it had cut 21,000 employees over the past year, attributing the reduction directly to its own AI adoption, one of the most explicit acknowledgments yet that enterprise AI is eliminating work at scale inside the companies selling it. That disclosure validates the disruption narrative structurally even as ServiceNow’s 25 percent revenue growth complicates the market-contraction version of the same narrative. Disruption and growth are running in the same quarter, in the same market.
Reuters reported that ServiceNow raised its full-year subscription revenue forecast on AI-driven demand. The raise came in below what some analysts had modeled, producing the conservative guidance characterization that briefly dampened the after-hours reaction before shares recovered. McDermott’s position on the call was that enterprise customers are still in the early phase of deploying AI agents at production scale, making the current guidance a floor rather than a ceiling.
The $298 million in net income for the quarter is the number that tends to get lost in the subscription revenue discussion. ServiceNow is profitable enough to fund the AI platform investments that defend its market position, which matters because the competitive pressure from Microsoft Copilot and Salesforce Agentforce is not slowing. The IBM partnership, the Now Assist expansion, and the governance product are all funded from a balance sheet that the Q2 results reinforced. The disruption fear, at least for now, is a fight ServiceNow can afford to have.
What the quarter does not settle is the medium-term structural question. ServiceNow’s revenue is growing because enterprise buyers are adding AI products on top of existing contracts, a model that benefits the incumbent platform holder in the current cycle. The question the Q2 numbers defer rather than dismiss is whether customers eventually consolidate onto an AI agent stack that does not require ServiceNow as the orchestration layer. McDermott’s kill switch is a commercial product. The question it cannot fully answer is who holds the switch when the agent stack itself evolves beyond the workflow platform it currently governs.
