Agentic change management: a risky change from request to close, with humans in control

Rajesh_Singh
Kilo Sage

I wanted to share a demo of Change Commander, an agentic change management app I built on native ServiceNow change management. The story in the video is a risky change at a fictional bank, Harbourline Bank.

When an engineer raises a change, the AI reads the screenshots, design documents, pull request and Jira story, and drafts the implementation, backout and test plans, with every line linked to its source. It checks the change schedule for conflicts, suggests a safe maintenance window, and runs the exact plan on a twin of production first, with test evidence attached automatically. Risk is scored factor by factor, then confirmed by a human risk assessor before it goes to the CAB on Microsoft Teams.

Once approved, a person handles the critical steps, and the release rolls out as a canary with health checks and automatic rollback. Going to 100% is always a human call. After go-live, the AI compares production with what the twin predicted and drafts the post-implementation review, so the change closes with a full audit trail.

I'd like to hear how far others are taking AI agents in change, and where you still want a person to make the decision.

I'd rather keep the discussion here in the community — questions and feedback are very welcome.

If you found this useful, please mark it Helpful and follow me for more ServiceNow content like this.

0 REPLIES 0