By Jaspal Singh, Senior Technical Architect & ServiceNow MVP 

Las Vegas was buzzing. 25,000 people. Three days. Announcements that could genuinely reshape how enterprises run.

I watched all of it from Dubai.

No badge, no queue, no Venetian Hall — just a screen, a browser with too many tabs, and the kind of focused attention that only kicks in when you know what you’re looking at. Fourteen years working inside the ServiceNow ecosystem will do that to you.

This is my Knowledge26 debrief. Not the press release version — the one that tells you what actually mattered, and why.

The Opening Shot: McDermott and the Agentic Declaration

Bill McDermott set the tone immediately. Speaking to 25,000 attendees at the opening keynote, he opened with a number that reframes every enterprise AI conversation: a projected global shortage of 50 million workers by 2030. Not a gap. A structural deficit – arriving at the exact moment billions of AI agents are coming online to fill it.

“What we’re seeing right now is bigger than AI. Bigger than software. Bigger than any technology shift we’ve watched across this stage,” he said. He positioned ServiceNow as the answer to what he called the move from AI chaos to control– a platform that brings together AI, data, and workflows with the governance layer enterprises actually need to operate at scale.

He didn’t shy away from ambition either, declaring ServiceNow the “fastest growing enterprise software company at scale the world has ever known” and signalling plans to double the business again in the years ahead.

The headline theme carried through every announcement: intelligence without governance is liability, not leverage.

Jensen Huang on Stage and Project Arc

NVIDIA’s founder and CEO Jensen Huang joined McDermott in what was more than a partner cameo. Huang brought data: NVIDIA, itself a ServiceNow customer, had cut employee support interventions by two-thirds using the platform. His framing of ServiceNow’s market position, “the operating system of enterprise AI agents” carried real weight coming from someone building the hardware those agents run on.

The joint announcement was Project Arc: a long-running, self-evolving autonomous desktop agent for knowledge workers, developers, and IT teams. It runs inside NVIDIA’s OpenShell sandboxed runtime with policy-based management, connects natively to the ServiceNow AI Platform via Action Fabric, and is governed end-to-end by AI Control Tower. Every action sandboxed. Every agent auditable. Every step policy-bound.

The two companies also released NOWAI-Bench – an open benchmarking suite covering agentic evaluation across ITSM, Customer Service and HR workflows, plus a separate framework for evaluating voice agents in enterprise settings. That’s a meaningful signal: the industry is starting to need standardised ways to measure what autonomous agents actually do.

Amit Zavery’s Blueprint: The Technical Architecture of Agentic Business

Day two belonged to ServiceNow President, CPO, and COO Amit Zavery, who delivered the engineering substance behind the vision. “Advisory AI has run its course,” he said plainly. “Enterprises need AI that senses, decides, and securely acts in accordance with organisational guardrails.”

The centrepiece technical announcement was ServiceNow Action Fabric – an open integration layer that exposes ServiceNow’s entire system of action to any third-party AI agent via a generally available Model Context Protocol (MCP) Server. Whether the agent is built on Anthropic Claude, Microsoft Copilot, or a customer’s own stack, Action Fabric lets it securely trigger governed enterprise actions on ServiceNow. Every triggered action inherits the same audit trails, approval chains, and policy controls already embedded in the platform.

Anthropic is the first named design partner, enabling Claude to initiate ServiceNow workflows directly – a password reset, an onboarding sequence, a procurement request – all while remaining inside the governance boundary.

Zavery’s other memorable line: “Data alone doesn’t run a business. Action does.”

Otto: Solving Enterprise AI’s Completion Problem

ServiceNow Otto was the most consumer-visible announcement of the event – but the technical architecture behind it is what makes it interesting.

Otto merges Now Assist, Moveworks, and ServiceNow’s AI Experience into a single unified interface across four channels: conversational AI, enterprise search, AI voice agents (multilingual, no menu trees), and AI Data Explorer for natural language queries against enterprise data. It’s live today inside ServiceNow EmployeeWorks, with a full platform rollout planned across 2026.

The problem it’s designed to solve is specific: most enterprise AI tools ship inside their own application silo. They can generate insights but can’t complete work across departments or systems. Otto is built to close that gap – from intent to finished outcome, across workflows, approvals, and handoffs, without the user ever switching context.

In the keynote demo, Otto generated a workforce readiness brief, identified staffing gaps, recommended seasonal hiring, blocked interview calendars, and kicked off next steps – all within a single workflow. Not five tools. One front door.

Autonomous Workforce: AI Specialists That Actually Work

The Autonomous Workforce expansion at Knowledge26 is the clearest sign yet that ServiceNow is moving beyond proof-of-concept AI into production deployments at scale.

The L1 IT Service Desk AI Specialist now generally available resolves assigned IT cases 99% faster than human agents in ServiceNow’s own internal deployment. It’s being joined by a full roster of role-specific specialists:

• CRM AI Specialists (GA now): Covering sales qualification, quoting, order fulfilment, invoice disputes, service, and renewals
• IT AI Specialists (expected June 2026): IT operations, AIOps, SRE, asset lifecycle, portfolio planning
• HR, Finance, Legal, Procurement, Workplace Services, Supplier Management, Health & Safety (rolling out 2026)
• Security & Risk AI Specialists (preview June, GA September 2026)

These aren’t chatbots with expanded scope. They complete end-to-end processes alongside humans resolving cases, containing threats, managing incidents, handling high-volume requests with defined roles, permissions, and audit trails.

ServiceNow AI Specialists will also appear in the Microsoft Agent 365 Marketplace as digital employees with defined roles, able to draft Word documents, respond to Outlook emails, and act on assigned PowerPoint comments, subject to Microsoft 365 admin policy controls.

AI Control Tower: From Visibility to Enforcement

First introduced at Knowledge 2025, AI Control Tower has matured significantly and at Knowledge26, it crossed a meaningful threshold: it’s now included by default across every ServiceNow product and package, not sold as an add-on.

The new enforcement capabilities allow it to discover, monitor, and shut down rogue agents across AWS, Azure, and beyond not just within ServiceNow’s own ecosystem. New governance extensions span Microsoft Agent 365, Microsoft Copilot Studio, NVIDIA Enterprise AI Factory, and Azure-based agent environments. That’s a vendor-agnostic control layer with real teeth.

ServiceNow also launched a programme offering AI Control Tower free for one year, valued at $2 million. The Australia release is the latest platform version spotlighted at Knowledge26 and brings these capabilities into general availability in August 2026.

Autonomous Security & Risk: The Armis and Veza Payoff

ServiceNow’s acquisitions of Armis and Veza are coming into focus with the launch of Autonomous Security & Risk – a single platform for governing AI agent identities, permissions, and connected assets across the enterprise.

Veza’s Access Graph maps every access relationship in real time across IT and cloud systems. Armis provides visibility across IT, OT, IoT, and critical infrastructure assets. Together they give the platform a live view of who (or what) has access to what – a prerequisite for any organisation serious about deploying autonomous agents safely.

Security and risk AI specialists expected in preview June, GA September will operate inside this governance boundary, with the capability to autonomously contain threats and manage security incidents within defined policy limits.

For Developers: Build Agent Everywhere

The CreatorCon and developer keynote — where ServiceNow Developer Advocates brought the energy that the developer community looks forward to all year — also had a technically significant announcement that deserves more attention than it’s getting in mainstream coverage.

ServiceNow Build Agent is now generally available in ServiceNow Studio, and its core skills have been extended into Cursor, Windsurf, Claude Code, and GitHub Copilot. Developers can now build for the ServiceNow AI Platform from any environment they already work in, with full platform context – data models, ACLs, business rules carried through without configuration. Free access to App Engine Management Center keeps applications governed before deployment.

The practical implication: ServiceNow is dismantling the barrier between the broader developer ecosystem and its platform. You no longer need to learn a new environment to build on ServiceNow.

The Vibe Code Lounge at CreatorCon, with over 130 NVIDIA-sponsored stations, gave attendees a live demonstration: describe an app’s intent in natural language, and the platform builds it. Functional apps in minutes, on the expo floor. It’s a compelling statement about where low-code development is going.

The Australia release was also spotlighted for the developer community bringing agentic execution, orchestration, intelligence, and built-in governance into the next platform version, with general availability targeted for August 2026.

The Bigger Picture

Knowledge26 was a company demonstrating execution, not just strategy. The acquisitions are cohering. The product architecture is connecting. The governance layer is getting enforcement capabilities to match its visibility promise.

The financial ambition is equally clear: subscription revenues exceeding $30 billion by 2030, with AI products expected to represent over 30% of annual contract value.

The competitive framing was everywhere: AI alone is becoming commoditised. Orchestration, governance, and the ability to complete work across the enterprise that’s the differentiator. ServiceNow is making a direct play to own that layer.

I watched all of it from a screen, not a seat. But the picture that emerged was consistent and clear: the agentic enterprise isn’t a roadmap slide anymore. It’s being built right now, on the Now Platform.

All Knowledge26 announcements can be found at servicenow.com/events/ondemand/knowledge.