A service desk analyst helps people when their computer or apps go wrong. They decide what's wrong, how urgent it is, and who should fix it.
Service desk analysts answer IT questions, fix problems and pass harder ones on. Their day is full of quick decisions: how urgent is this, which category, can I fix it or should I escalate? Several of those suit AI suggestions. A few, like access and security calls, should stay firmly with people.
Closest O*NET match: 15-1232.00 Computer User Support Specialists. High-frequency decisions (categorise, prioritise, resolve or escalate, knowledge article choice) suit decision support with ticket-history grounding; access grants, security incident calls and VIP exceptions should remain human, with logged rationale.
The role
A service desk analyst is the first point of contact when people have problems with computers, software or accounts. The closest O*NET occupation is Computer User Support Specialists (15-1232.00). Its task list includes answering user enquiries about software or hardware, investigating problems by reading manuals, talking to users or running diagnostics, keeping records of problems and the action taken, and referring major problems to vendors or technicians.
That matches what most service desks do, with one addition O*NET doesn't stress: a large share of the work runs through a ticketing system, which records nearly every decision. That makes this role one of the easiest to study, as described in Ticket logs as a map of decisions.
The decisions
| Decision | Options | Frequency | Stakes | Reversible? |
|---|---|---|---|---|
| Category of a ticket | A list of categories | Very high | Low | Yes |
| Priority | Usually 4 or 5 levels | Very high | Medium: a missed outage costs many people time | Yes, if noticed |
| Resolve now or escalate | Fix, escalate, or ask for more information | High | Medium | Yes |
| Which knowledge article to use | Several articles, or none | High | Low | Yes |
| Whether to grant access someone asks for | Grant, refuse, or refer | Medium | High: wrong access can expose data | Partly |
| Whether an issue is a security incident | Yes, no, or unsure | Low | High | No, once missed |
The frequency and stakes labels are general; a real card would use the scoring guide.
Where AI helps
Categorising and suggesting a priority are good candidates for a decision model. They're frequent, mostly reversible, and the ticket history provides thousands of labelled examples. The model suggests; the analyst accepts or changes it with one click; the changes become training and test data.
Finding the right knowledge article suits a SharePoint agent or a Copilot Studio agent grounded in the knowledge base. Summarising a long ticket history when a case is passed on suits Microsoft 365 Copilot or a summary step in the ticketing tool.
Where people should stay in charge
Access requests and security judgements carry high stakes and can't always be undone. An AI tool can gather the facts, such as the requester's role and the approver on record. A person should make the call, and the reason should be logged.
I'd also keep "unsure" as a valid answer for security. An analyst who's unsure should be able to escalate without needing to justify it.
A worked example
These numbers are illustrative. A service desk with 12 analysts handles 6,000 tickets a month.
| Decision | Time now | With AI suggestion | Time saved a month |
|---|---|---|---|
| Category and priority | 40 seconds per ticket | 15 seconds to check a suggestion | About 42 hours |
| Knowledge article | 2 minutes on half of tickets | 1 minute | About 50 hours |
That's about 92 hours a month across the team, if the suggestions are good enough to accept most of the time. The real number depends on the acceptance rate, which is what I'd measure in a pilot.
What I'm still checking
I've assumed analysts check each suggestion. In practice, accepting becomes automatic over time, as described in Controls people actually keep. For priority, I'd want a weekly sample of accepted suggestions to catch drift.
Sources
- O*NET OnLine, 15-1232.00 Computer User Support Specialists, US Department of Labor, accessed 11 October 2026.
The decisions, time estimates and savings are illustrative.