This article is human-conceptualized but written with the assistance of AI and may include some inaccuracies. Always validate independently.
In the video above, NVIDIA CEO Jensen Huang calls ServiceNow "the operating system of enterprise AI agents." But what does that actually mean? This blog is an analysis of that claim — we break down the architecture, the components, and the strategic moves that make this vision possible, and explore why Jensen believes every enterprise IT department will manage AI agents through ServiceNow.
In this short but loaded statement, Jensen Huang makes several key claims:
These are bold claims. Let's analyze each one and understand the architectural reasoning behind them.
Just as a computer's operating system manages hardware, software, security, and data flow so applications can run smoothly, Jensen is arguing that ServiceNow manages enterprise workflows, data, security, and human-to-machine interactions — making it the natural "OS" layer for AI agents in the enterprise.
At the Knowledge 2025 conference, Huang explicitly stated ServiceNow is "destined to be the best platform, the operating system of enterprise AI agents." His reasoning is straightforward: every company will need AI agents curated and managed by IT departments, and ServiceNow is already embedded in almost every major IT department globally.
He confirmed that NVIDIA itself runs AI agents on the ServiceNow platform, referring to it as the "operating system of NVIDIA." The partnership has yielded several technical collaborations:
Just as a computer's OS manages hardware, software, security, and data flow, ServiceNow manages enterprise workflows, data, security, and human-to-machine interactions. Here are the key components:
The underlying workflow engine that allows AI agents to trigger workflows across different departments (IT, HR, Customer Service, Security) using a single, unified codebase. AI agents need to do more than generate text; they need to take action.
AI agents are only as good as the data they can access. ServiceNow's Data Fabric connects unstructured and structured data across the enterprise without duplicating it, giving AI agents real-time, secure context for decision-making.
How ServiceNow talks to the rest of the company's software (Salesforce, Workday, SAP, Jira, etc.). AI agents sitting on ServiceNow use Integration Hub to reach into external systems to fetch data or execute tasks.
Routes prompts to the appropriate LLMs while stripping out sensitive data, enforcing role-based access control, and keeping an audit trail of every AI agent action.
The CMDB (Configuration Management Database) is a comprehensive database that tracks every piece of hardware, software, network component, and digital service in a company as Configuration Items (CIs), along with their relationships.
For IT to "curate, manage, and make AI agents better" (as Huang stated), those agents must be treated as CIs within the CMDB. Every AI agent becomes a managed enterprise asset.
Placing AI agents in the CMDB enables:
A modern CMDB should track anything that constitutes a Configuration Item, including:
Understanding the distinction:
ServiceNow historically operated proprietary "Advanced High Availability" (AHA) private cloud infrastructure. However, they are transitioning to a hybrid cloud approach, partnering with Microsoft Azure, AWS, and Google Cloud. Customers can now host ServiceNow instances on hyperscalers, and ServiceNow plans to sunset some proprietary data centers by 2028.
In late 2025, ServiceNow dropped over $11 billion acquiring security and AI companies to build the governance layer for AI agents:
| Acquisition | Value | Purpose |
|---|---|---|
| Armis | $7.75B | Device security for OT/IoT. Tracks and secures factory robots, HVAC systems, medical devices. Maps every physical device on the network. |
| Moveworks | $2.85B | Enterprise AI copilot. Conversational "front door" for employees to interact with systems. |
| Veza | ~$1B | AI-native identity security. Manages machine identities and maps exactly what data/systems an AI agent can access. |
By combining the CMDB, Armis (physical device mapping), and Veza (digital permission mapping) under the AI Control Tower, ServiceNow is selling the regulatory and security infrastructure for the entire AI economy.
The AI Control Tower is the enterprise orchestration and governance plane. It acts as the centralized workspace for managing all AI initiatives:
ServiceNow has moved beyond simple API integrations, now supporting two protocols for agent-to-agent communication:
| Protocol | What It Does | ServiceNow's Role |
|---|---|---|
| Model Context Protocol (MCP) | Agent-to-Tool Connection. Standardizes how an AI agent connects to data sources and tools. | MCP Server: ServiceNow acts as a repository of tools. External agents can "ask" ServiceNow to create incidents or check user status without custom API code. |
| Google A2A (Agent-to-Agent) | Agent-to-Agent Collaboration. Allows independent AI agents to negotiate and hand off tasks. | The Orchestrator: A ServiceNow agent can initiate a conversation with a Dynatrace or Google Cloud agent to investigate issues. |
Here is the vendor-agnostic reference architecture, followed by how real-world products map to each block:
| Layer | Functional Block | Role |
|---|---|---|
| Interaction | Incident Management | Entry point for human reports, system alerts, and ticket tracking |
| AI Governance | Security Controller | Verifies machine identity, enforces least-privilege access |
| Privacy & Compliance | Scans I/O to prevent PII or sensitive data exposure | |
| VUL Check | Pre-execution check to prevent security vulnerabilities | |
| Intelligence | AI Agents | Orchestrator Agent plans tasks; Specialized Agents execute |
| Execution | Observability | Feeds real-time telemetry (logs, metrics, traces) |
| CMDB | Maps assets, owners, dependencies, and current state | |
| Automation | Executes scripts, API calls, and runbooks |
| Block | Ecosystem Tools | ServiceNow Fit |
|---|---|---|
| Incident Management | Jira SM, PagerDuty, Zendesk | ServiceNow ITSM (system of record) |
| Security Controller | CyberArk, Okta | AI Control Tower + Veza |
| Privacy & Compliance | Azure AI Content Safety | AI Trust Layer |
| VUL Check | Tenable, CrowdStrike, Qualys | Vulnerability Response |
| AI Agents | MS Copilot, AWS Agent, Dynatrace | Now Assist Agents (Orchestrator via A2A) |
| Observability | Datadog, Splunk, Dynatrace | Cloud Observability + integrations |
| CMDB | Device42, Lansweeper, AWS Config | ServiceNow CMDB (crown jewel) |
| Automation | Ansible, Terraform, Azure DevOps | Integration Hub / Flow Designer |
While AI coding agents may accelerate development, replicating ServiceNow's capabilities is not just an engineering challenge. Its moat includes platform depth, integrations, workflow knowledge, customer trust, operational data, and years of enterprise deployment experience. These advantages create a significant barrier to entry even with advanced AI assistance.
Source: This blog is based on Jensen Huang's remarks at ServiceNow Knowledge 2025 and subsequent official announcements about the NVIDIA-ServiceNow partnership.