AI Ticket Resolution and the Verification Problem.
Ticket automation should not optimise closure rate at the expense of resolution. Eligibility, approved runbooks, outcome checks and visible escalation are the control system.
Closure is a claim, and claims need evidence
A closed ticket asserts that a problem is gone. Unless something checked that assertion, closure rates measure activity rather than service quality.
Eligibility beats blanket automation
Automating every category produces failures in the ones that mattered. Define which categories, assets and risk profiles are eligible, and leave the rest with people deliberately.
Treat a failed check as an assignment
When verification fails, the correct behaviour is to keep the ticket open and hand it to a named owner with the full action history — never to retry quietly.
Report on verified resolution, not deflection
The metric that changes behaviour is verified resolution rate. Publish it with its baseline, period and method, or it will not survive scrutiny.
Automate only eligible categories, verify before closure, route failed checks to a named owner, and report verified resolution with its method.