Automate the ticket. Keep the judgement.
A practical, responsible approach to AI in IT service management: what to automate, where humans must stay in the loop, and how to govern it.
Every ITSM platform now promises AI. Auto-classification, suggested resolutions, virtual agents, predictive incident detection. Some of it is genuinely transformative. Some of it will quietly make your service worse.
The difference is rarely the technology. It is whether you have decided, deliberately, what the AI is allowed to do on its own — and what it is not.
My rule of thumb is simple: automate the ticket, keep the judgement.
What to automate with confidence
AI is excellent at high-volume, repeatable work where a mistake is cheap and easy to spot:
- Categorising and routing tickets, so they reach the right queue first time.
- Suggesting knowledge articles to agents and end users. A well-maintained knowledge base was how one of my teams cut resolution time by 35% — AI makes that knowledge far easier to find.
- Summarising long ticket histories before a handover or escalation.
- Drafting status updates for a human to check and send.
- Spotting patterns across incidents that point to an underlying problem.
None of these remove a person from a decision that matters. They remove the typing, searching and sorting that stops people getting to those decisions.
Where humans stay in the loop
Judgement is different. Keep a named person accountable wherever an error would affect customers, cost, security or compliance:
- Declaring and prioritising major incidents. Severity is a business call, not a keyword match.
- Approving changes to production systems.
- Anything touching access, identity or security.
- Communicating with executives and customers during an incident.
- Closing problems and accepting risk.
This is not distrust of the technology. It is the same principle we apply to people: authority goes with accountability.
Govern AI like any other service
Responsible adoption does not need a new bureaucracy. It needs your existing governance applied properly:
- Define the use case and its decision rights. What exactly may the AI do without approval?
- Set quality measures. Routing accuracy, deflection that actually resolves the issue, customer satisfaction on AI-handled tickets.
- Keep a human override that is fast and obvious.
- Review regularly. Sample AI decisions, just as you would audit a new team member's work.
- Protect the data. Know what ticket content is sent where, and make sure it meets your security and privacy obligations.
Augmentation, not replacement
The strongest case for AI in ITSM is not headcount reduction. It is giving your best people back the time to do the work only people can do: leading incidents, managing vendors, fixing root causes, and talking to the business.
Used that way, AI raises the standard of service. Used as a shortcut around judgement, it simply makes mistakes faster.
Automate the ticket. Keep the judgement.
Want this working in your organisation?
Take the free 3-minute ITSM maturity check, or book a 15-minute call to talk it through.