MCP integration
One governed surface for AI to act through, instead of a dozen brittle scripts that each reinvent auth and error handling.
AI & ServiceNow
Governed AI on ServiceNow
Anyone can bolt a chatbot onto an instance. The hard part is AI a bank's security team will
sign off on. That's the work I do.
Most AI on ServiceNow stops at the demo. Connecting a model to an instance is easy. Making it safe enough for production is the real job: data that stays where it belongs, and agents that do only what they're allowed to. That takes ServiceNow knowledge, not just an API key.
I treat the model as the easy part and the governance as the work. The result is AI you can put in front of an auditor, not just a slide.
What I work on
The pieces that make AI safe on the platform
Now Assist & AI Agent Studio
ServiceNow's native AI, configured with scopes and audit, not left wide open because it shipped that way.
Knowledge assistants
Answers grounded in your own ServiceNow data, with sources a user can actually check.
Governance and review
AI that holds up to a security review: real scopes, ACLs that still apply, audit logs, and a blast radius you can name.
How AI fits
The AI Operating System
AI shouldn't sit in isolation. It works on top of your people, systems, knowledge, and processes — turning them into decisions, automation, and insight.
- People
- Systems
- Knowledge
- Processes
AI Layer
- Decisions
- Automation
- Insights
The Claude tilt
Why I lean toward Claude
The durable bet is the capability: MCP and a controlled surface that holds regardless of which model wins. The near-term bet is Claude, after ServiceNow made it a default model for its Build Agent and shipped a supported MCP server. I'm in the Anthropic Claude Partner Network on the consulting track, with a public MCP reference repo to back the approach.
Thinking about AI on your ServiceNow instance?
Start with an assessment. We map where AI actually pays off before anyone writes code.