AI Ticket Intelligence guide

Ticket Intelligence triages every incoming ticket automatically and gives technicians analysis, similar-ticket search and response drafting on the ticket itself.

Concepts

Concept	What it means
Auto-triage	Analysis that runs on ticket creation without anyone asking for it.
Analysis	The cached AI result on a ticket: category, priority + reasoning, summary, sentiment, urgency, business impact.
Freshness	Cached analysis is reused until the ticket is edited; then it re-runs.
Suggestion	A proposed category / priority / tag that a person accepts before it is applied.
Duplicate detection	Similarity search run while a ticket is being written, at a 0.75 threshold.
Deflection	Knowledge-base search offered to an end user as they describe a problem.

Roles and permissions

power user — not organization administrator — because the copilot renders on the org portal’s ticket page without its own gate, and not MSP technician, which ranks above organization administrator and would remove the org portal’s copilot entirely.

Walkthrough — triage happens on its own

A ticket is created in any portal.
The analysis is written to the ticket’s AI fields.
If the call is refused — for example the account is out of credits — the rejection is logged with its status.

There is no setting to turn this on; it is the default behaviour on creation. To re-run it deliberately, an automation rule can use the Run AI analysis action.

Walkthrough — working a ticket with the copilot

Open MSP → Tickets → <ticket> or Org → Tickets → <ticket>.
The AI copilot panel shows the analysis. If the ticket has not changed since it was last analysed, this is served from cache and costs nothing.
Open Similar tickets to see resolved tickets above 0.60 similarity, with their resolutions.
Choose Generate response, pick an audience (technical, non-technical, executive) and a tone (professional, empathetic, urgent, instructional).
Edit the draft, then send it as a normal comment.

Walkthrough — deflection while a user writes a ticket

The user starts Customer portal → Requests → New.
As they describe the problem, matching knowledge articles are offered.
Classification runs on the text they typed and pre-fills category and priority.

Configuration

Setting	Where	Effect if unset
Plan AI credit allowance	Plan tier	Free-tier floor applies
Queue AI confidence threshold	Queue settings	Defaults to 0.70. Below it, a classification is offered for review instead of auto-applied

Troubleshooting

Symptom	Cause
`404` on a ticket you can see in the UI	The ticket is outside the accounts you service.
`402 AI_CREDIT_LIMIT_EXCEEDED`	The paying account is out of credits for the period.
Analysis re-runs every time the panel opens	The ticket is being modified between reads — the cache is keyed on the ticket’s own last change.
`503` on analyse	The model call failed; the ticket keeps its previous analysis.

Limits and known behaviour

The analysis writes to the ticket’s AI fields, not its real fields. Applying a suggestion to the ticket’s own category or priority is an explicit acceptance.
The AI category vocabulary and the ticket form’s five-value category list are different sets.
New tickets are embedded by a background sweep, not on creation, so similarity results lag ticket creation.