ServiceNow Knowledge 2026, held May 5–7 at The Venetian Resort and Wynn Las Vegas, signaled a departure from experimental automation toward a structured, agent-led operating model. Under the theme “You and AI Get to Work,” the event gathered thousands of practitioners and executives to examine how a unified platform handles the shift from human-assisted AI to autonomous AI specialists. While previous years focused on the potential of generative interfaces, the 2026 conference centered on the infrastructure and governance required to let those agents act on behalf of the enterprise.
Key Announcements
Expansion of AI Control Tower
ServiceNow introduced new capabilities to its AI Control Tower to provide a centralized hub for governing, securing, and measuring AI agents across any system. This update allows enterprises to discover every AI model in use, whether native to the platform or deployed via hyperscalers like AWS and Microsoft. It functions as a command center for IT leaders to monitor performance and enforce safety guardrails.
Launch of AI Specialists and Autonomous Workforce
The event marked the debut of AI Specialists, autonomous agents designed to execute end-to-end workflows without constant human prompts. These agents function as digital team members with defined roles, assigned to specific departments like IT support or HR. Unlike standard chatbots, these specialists are scoped to defined departmental roles rather than general-purpose Q&A.
Integration with Veza and NVIDIA
Deepened partnerships with Veza and NVIDIA aim to secure the infrastructure layer of AI deployments. The integration with Veza brings access graph technology and least-privilege enforcement to AI identities, treating agents with the same rigorous permission standards as human employees. Meanwhile, the collaboration with NVIDIA Enterprise AI Factory provides validated designs for agent observability, ensuring the hardware and software layers remain transparent.
Introduction of RaptorDB
RaptorDB is a new database engine built for the throughput and latency requirements of large-scale agentic workflows. This ensures that the underlying data architecture does not become a bottleneck as automation volume scales.
Strategic Insights
Security Has to Move With the Agents
Modern security requires a shift from static perimeter defenses to dynamic, mobile monitoring. Kevin Thompson, director of outbound product management, argued that securing AI agents like a medieval castle fails because agents operate outside traditional boundaries. A drone-like security model, where tools are highly mobile and constantly hovering over every interaction, provides the necessary oversight for autonomous actors.
Data Quality Dictates Agent Viability
Hallway conversations and technical sessions highlighted that fragmented data prevents agents from making accurate decisions. Companies like Shell demonstrated that returning to out-of-the-box configurations and eliminating technical debt is a prerequisite for achieving silent upgrades and functional AI specialists.
Agents Require Human Identity Standards
Enterprises are now treating AI agents as human identities for purposes of security and permissions. Holly Briedis, senior vice president for global industries, emphasized that agents must run within the same governance framework that protects human workers. Assigning agents specific roles rather than broad system access limits the “blast zone” if an agentic action fails or produces an unintended outcome.
Context Is the New Interface
Conversational platforms like Janus replace traditional portals with context-aware interactions. Users expect to interact with the platform through voice and natural language while staying within their existing workflows. This shift forces developers to prioritize user intent and data models over the construction of complex UI components.
The Undercurrent
Rapid AI deployment and rigid governance pulled against each other throughout the conference. While CEO Bill McDermott and other leaders championed the speed of “agentic business,” the recurring mention of “kill switches” and “failsafes” suggests genuine concern regarding the autonomy of these tools. The conference content frequently balanced the promise of 90% autonomous IT support with warnings about AI failures, such as the deleting of production databases when oversight is absent.
The strategic signal from Knowledge 2026 is that ServiceNow intends to move beyond being a system of record to becoming the “graph of graphs.” By integrating knowledge, action, asset, and decision graphs into a single platform, the company is positioning itself as the primary orchestrator for all corporate AI, regardless of where the underlying models reside. This moves the platform into direct competition with hyperscalers for the title of the enterprise’s “front door.”
Why It Matters
The shift to an agentic enterprise indicates that automation is no longer a bolt-on feature. Practitioners must prepare for a reality where AI agents possess identities, permissions, and the power to execute transactions. For tech leaders, this means a complete audit of data hygiene and identity governance. If the data is messy, the agents will be ineffective or worse, dangerous. The move toward “agent usage” instead of “per-seat” fees also suggests a fundamental change in how software value is measured and purchased.
What’s Next
Tech roadmaps must prioritize data consolidation to feed engines like RaptorDB. Organizations should begin defining specific “job descriptions” for AI specialists, moving away from broad, general-purpose LLM implementations.
The coming year will likely see a surge in the adoption of AI gateways and control towers as companies move from the “tinkering” phase seen in the Expo halls to full-scale production. Leaders who get governance right early will find themselves on shorter upgrade cycles, the metric ServiceNow itself flagged as a key indicator of agent readiness.
