That is the AI disruption question for enterprise software, and ServiceNow sits closer to its center than almost any company, because it is simultaneously one of the biggest potential beneficiaries and one of the more exposed incumbents.
First, the quarter itself
ServiceNow reported Q2 after the close on July 22. Analysts expected revenue around $3.92 billion, roughly 22% year-over-year growth, and the company had beaten estimates in each of the prior four quarters, so a headline beat was close to the base case.
Which is exactly why the headline was not the story. The prior quarter was the tell. On April 22, ServiceNow reported Q1 numbers that looked nearly flawless, subscription revenue and cRPO both beat, operating margin beat its own guide, and the stock fell 17.8%.
The crash came from three things buried in the guidance rather than the results. Full-year operating margin guidance was trimmed from 32% to 31.5% and free cash flow margin from 36% to 35%, both attributed to the newly closed Armis acquisition. And roughly 75 basis points of subscription growth was lost to delayed on-premise deals tied to Middle East conflict, a headwind management warned could persist. Beat the quarter, cut the runway. That pattern, not the print, is what to watch, because a company priced for perfection gets judged on the guidance, not the beat everyone already expects.
The pricing question underneath
Now the harder thing, the one no single quarter can settle.
The traditional enterprise software model, including ServiceNow's, is substantially seat-based: you pay per user who logs in and does work. It is a clean model because it scales with a customer's headcount, and headcount generally grew. The entire agentic AI pitch is that software agents now do tasks that human users used to do, which is genuinely valuable and also, for a seat-based vendor, structurally awkward. An agent that lets a customer handle the same IT ticket volume with fewer people is an agent that reduces the customer's seat count, and therefore the bill.
ServiceNow is aware of this, and its response is the thing actually worth evaluating. The company is shifting toward consumption and outcome-based pricing for AI, its Now Assist product, so that value delivered by agents is monetized through usage rather than seats. On the Q4 call, Now Assist net new ACV topped $600 million and more than doubled year over year, tracking toward a $1 billion-plus target for 2026, with token-based reloads driving additional revenue.
Token-based reloads is the phrase to notice. It means ServiceNow is trying to build a meter that runs on AI usage, so that if agents do more work, ServiceNow captures more revenue even as seat counts flatten or fall. That is the correct strategic response in principle. Whether it actually offsets seat erosion is an empirical question, and the honest answer is that nobody knows yet, including ServiceNow, because the transition is early and the two effects, seat compression and consumption growth, are running at the same time.
So the number worth hunting for is not total revenue. It is any disclosure on Now Assist ACV, net new ACV growth, and consumption trends, because those measure whether the new meter is spinning fast enough to replace the old one. A strong overall quarter with decelerating Now Assist growth would be a quiet warning. A slightly soft quarter with accelerating AI monetization would be the opposite.
The IBM signal, and why it matters here
There is a fresh piece of context that makes this more than a company-specific event.
Days before ServiceNow reported, IBM pre-released a large miss and said enterprise clients had redirected late-quarter budgets toward AI hardware and memory instead of software. That raised a question the whole software sector now has to answer: was that IBM-specific, or is there an industry-wide budget shift where money that used to buy enterprise software is being pulled toward AI infrastructure? ServiceNow's bookings are the next real data point on whether that fear shows up in actual demand or stays confined to IBM's particular mix.
The two disruption stories are related but distinct. One is budget reallocation, customers spending on GPUs instead of SaaS this quarter. The other is structural, AI agents reducing the seats SaaS is priced on. ServiceNow's report speaks to both, and cRPO, the contracted future revenue due within a year, is the cleaner read on demand than current revenue, because it reflects what customers are committing to going forward rather than what they signed months ago.
The near-term noise
Two smaller things cloud the read, and both deserve to be separated from the structural question.
The Armis acquisition is depressing margins by design, roughly 125 basis points to Q2 operating margin, with management saying integration headwinds normalize in 2027. That is a deliberate investment, not a demand problem, and conflating the two would misread the quarter. The relevant test is whether management holds to the 2027 normalization story.
And there was an actively exploited security vulnerability in ServiceNow's platform disclosed just ahead of earnings. For a company whose entire value proposition is being the trusted system of record for enterprise workflows, security incidents carry reputational weight beyond their technical severity.
How to read it
ServiceNow is one of the best-run companies in enterprise software, growing over 20% at real scale with high margins and near-total customer retention. None of the disruption questions dispute that. They ask something different: whether the model that produced those results survives the technology the company itself is selling.
The bull case is that ServiceNow becomes the control layer for enterprise AI agents, and consumption pricing lets it capture more value than seat pricing ever did, because agents do vastly more work than the humans they assist. The bear case is that seat erosion arrives faster than consumption revenue scales, and a 20%-plus grower quietly decelerates as customers realize they need fewer licenses.
The full report offers three readings on it: whether guidance gets cut again the way it did last quarter, whether Now Assist consumption is accelerating fast enough to matter, and whether cRPO shows the budget shift that spooked IBM bleeding into ServiceNow's demand. Those are the lines that carry information. The revenue beat, which was nearly a foregone conclusion, is not one of them.