"Advisory AI has run its course. Enterprises want AI that can sense, decide and act safely." ServiceNow chief product officer Amit Zavery used Knowledge 2026 in Las Vegas to draw a sharp line for the market: assistants that merely sit beside employees and offer suggestions are no longer the endgame.
The company’s argument is direct. If AI can complete an entire business process, why should people still be asked to carry out the AI’s recommendation?
The numbers came first
ServiceNow put production metrics in front of enterprise buyers: its internal IT specialist resolves cases 99% faster than human handling, DocuSign is targeting fully automated resolution for 90% of IT tickets, Raleigh, North Carolina has reached a 98% deflection rate for employee requests, and ServiceNow’s CRM platform processes more than 100 million customer cases a month while configuring 7 million quotes a year.
The Raleigh example matters because it is not a large-enterprise-only story. A city government getting 98% of IT requests handled without human intervention points to a model that could materially reduce small public-sector IT costs.
What AI Specialists are meant to cover
The launch is not a single product. It is a roster of AI replacements for enterprise roles: IT operations agents that monitor, repair and close tickets; CRM agents that handle customer issues end to end; HR agents for employee requests and workflow completion; finance and legal agents for standardized audits and compliance checks; procurement agents for quotes and contract routing; and security or risk agents for asset discovery and identity governance.
IT specialists are scheduled for general release in June 2026, followed by security and risk specialists in September 2026. ServiceNow calls the management layer AI Control Tower and is bundling it by default across products and packages, moving what used to be an add-on into the infrastructure layer of its platform.
Microsoft and Nvidia are part of the platform play
The Microsoft Agent 365 integration allows ServiceNow specialists to run inside Outlook, Word and PowerPoint, while AI Control Tower extends governance across the Microsoft environment. In practice, an employee could receive an expense email in Outlook and have the AI complete the ServiceNow workflow without switching platforms.
ServiceNow also connected its AI Platform to Nvidia accelerated computing infrastructure to speed agent deployment and inference. Together, the integrations make the pitch broader than selling agents: ServiceNow is trying to position itself, with Microsoft and Nvidia, as an operating layer for enterprise AI.
Security is the trust layer
One of the most important figures from the event was that ServiceNow’s security and risk business has surpassed $1 billion in annual contract value. Customer examples included a global energy company deployed across 70 countries cutting threat containment time by 97%, a U.S. financial institution eliminating 96% of dormant non-human identities, and an aerospace manufacturer reducing control testing time by 75%.
That matters because the main barrier to deploying AI agents is not whether the technology can act, but whether enterprises trust it to touch production systems. ServiceNow has folded prior acquisitions Armis and Veza into Autonomous Security & Risk to unify asset and identity governance.
People still have to be retrained
The company also introduced SimStudio, a hands-on simulation environment for learning how to work with AI specialists, and AI Learning Guide, a personalized learning-path assistant. ServiceNow University is approaching 2 million learners, up 80% year over year.
The logic is clear: if AI specialists take over 98% of routine IT tickets, the remaining human work shifts toward using, tuning and governing AI. That creates a training requirement, and another business line.
What it means
ServiceNow is separating AI as helper from AI as worker. Over the past three years, most enterprise AI purchases have been copilots for employees: Copilot, Notion AI and Gemini for Workspace all keep the employee as the main actor. ServiceNow is betting that the process itself becomes the actor, with humans appearing mainly at exceptions and decision points.
If that model works, it looks less like office software and more like business-process outsourcing delivered by an AI team. That would put pressure on the most valuable parts of BPO providers such as Cognizant and Infosys. The real test will come after the June general availability of IT specialists, when customer retention data starts to show whether the model holds.
Sources: CocoLoop, ServiceNow just unveiled an AI workforce that can run your entire company (Fortune)