ServiceNow’s AI help desk bet paid off

ServiceNow just made a very uncomfortable point for anyone who still thinks AI always eats software revenue. The company says its internal help desk now lets AI handle more than 90% of targeted Level 1 IT tickets on its own, with 99% resolution rates in that narrow workflow. At the same time, quarterly subscription revenue hit $3.47 billion, up 21% year over year, and the stock jumped about 5% after the company introduced its Autonomous Workforce push.

That combination matters because it shows the old SaaS fear is too simple. If AI replaces human work inside the product, revenue does not have to fall. It depends on how the product is priced, and ServiceNow is now pricing around AI actions instead of only human seats.

What the 90% figure actually means

The headline number is real, but it is narrower than it first sounds. ServiceNow is talking about its own internal help desk, not every customer workflow across the platform. The bot is handling routine, high-volume requests where the rules are already well understood and the payoff from automation is easy to measure.

The company broke out the ticket mix, which makes the claim easier to judge. Password resets account for 33% of the targeted tickets, cloud authentication another 33%, collaboration tools 13%, laptop issues 8%, and software installation 6%. When the bot is unsure, it escalates. That detail matters because it shows the system is not pretending to solve every problem on its own.

• 90%+ autonomous handling on ServiceNow’s internal help desk
• 99% resolution rate on targeted routine tickets
• 33% password resets and 33% cloud authentication in the mix
• 13% collaboration tools, 8% laptop issues, 6% software installation

ServiceNow is also seeing similar results outside its own walls. One industrial conglomerate reportedly reached 90%+ autonomous handling with 99% routing accuracy, and a European drugstore chain cut support time from 9 minutes to 30 seconds. Those are the kinds of numbers that make IT leaders pay attention, because they are not about demo polish; they are about labor hours and ticket queues.

There is a practical lesson here for developers and IT teams. AI help desk tools work best when the workflow is repetitive, the system has a clean source of truth, and escalation paths are obvious. If those pieces are messy, the automation rate falls fast.

The pricing trick is the real story

The more interesting move is not the automation itself. It is how ServiceNow is charging for it. CEO Bill McDermott said half of all new business bookings now come from pricing models that are not classic per-seat licensing. That is a big statement, because it suggests the company is already shifting away from the old software math.

Per-seat pricing creates a problem when AI starts doing work that humans used to do. If a customer needs fewer agents, the vendor’s revenue gets squeezed. ServiceNow is trying to break that link by selling a mix of traditional user licenses and AI tokens, which it calls Assists. When a customer uses more AI actions than the base allocation allows, it buys more packs.

This is why the model matters so much. More automation does not reduce revenue; it can increase it. If the AI resolves more tickets, the customer consumes more Assists. The vendor gets paid for activity, not just headcount.

• Half of new bookings now use non-seat pricing models
• Pro Plus runs about 30% to 45% above standard user pricing
• Assist packs kick in after base AI allocation is used up
• More AI usage can mean more revenue, not less

That pricing structure changes the incentives in a way software buyers will notice quickly. ServiceNow now benefits when its AI performs well, which is very different from the old fear that automation would shrink the license base. The company is effectively monetizing throughput.

Why the numbers moved the stock

Investors liked the update because the growth story is still intact. ServiceNow said its Now Assist product passed $600 million in annual contract value by the end of 2025 and is on track to cross $1 billion by the end of 2026. That would make it the fastest product ramp in the company’s history, which is a strong signal even in a crowded enterprise software market.

The rest of the financial picture is strong too. Full-year 2025 subscription revenue landed around $13 billion, operating margins were 31%, and free cash flow reached $4.6 billion, up 34% year over year. Renewal rates sat at 98%. ServiceNow also said it has 2,109 customers spending more than $1 million in annual contract value, up from 1,080 in 2021.

Those numbers explain why the market did not punish the company for automating its own support team. The business is growing, the margin profile is healthy, and the AI product is not a side project anymore.

“AI doesn't replace enterprise orchestration. It depends on it.” — Bill McDermott, CEO of ServiceNow

That quote gets at the core of the strategy. ServiceNow is betting that AI agents still need workflow systems, approvals, records, and integrations to do useful work inside large companies. If that is true, then the orchestration layer becomes more valuable as AI adoption rises.

How this compares with other enterprise AI pricing

ServiceNow is not the only company trying to price AI differently, but it may be the clearest example of what happens when the model works. Salesforce Agentforce is testing multiple pricing approaches, Zendesk charges $1.50 per resolution in some cases, and Atlassian is bundling AI into existing plans.

The details differ, but the direction is the same: vendors want pricing that tracks value created by AI, not the number of humans sitting in front of a screen. That is a big shift for enterprise software, and it will affect how product teams think about usage meters, quotas, and packaging.

• Salesforce is testing multiple AI pricing models for Agentforce
• Zendesk has a per-resolution model at $1.50 in some cases
• Atlassian is bundling AI into current plans
• ServiceNow uses a hybrid of user licenses and AI Assists

For builders, the takeaway is straightforward. If your AI product only charges per seat, you are leaving money on the table when the software starts doing real work. If it charges per task, per action, or per outcome, automation can become a revenue engine instead of a margin problem.

What developers should watch next

ServiceNow’s move is a warning shot for the rest of enterprise SaaS. The companies that control the workflow layer, the system of record, and the action layer get paid when AI agents do more work. The companies that only sell interfaces will have a harder time defending their pricing once customers start asking why humans still need to be in the loop for basic tasks.

The next question is whether this pricing model spreads beyond IT service management. If AI can automate support, approvals, procurement, and internal operations, then the best vendors will charge for the work done, not the seats occupied. That is a cleaner fit for AI-era software, and ServiceNow is now showing the numbers to back it up.

My bet: over the next 12 months, more enterprise vendors will quietly move to hybrid pricing with usage-based AI credits, even if they keep the old seat language in the sales deck. The real question for buyers is simple: when your AI saves time, who gets paid for the savings?