How AI reads new requests

Each new request from a person is read once by AI, usually within a few seconds. While it waits, the case shows as New.

1. Request form match: if the text asks for something one of your request forms offers, it becomes that request (keyword match first, then AI choosing among your forms). Reported faults are never turned into requests.
2. Kind of case: incident, request, change or suggested problem.
3. Summary: a short title in plain words. The requester's own text always stays in the conversation.
4. Priority: from the impact and urgency the requester describes, at most High (see Priority, impact and urgency).
5. Mood: neutral, frustrated or angry, from the requester's words.
6. Routing: your routing rules and the group's automatic assignment pick the group and owner.
7. Due date: from your service levels for that kind and priority.

When AI is unavailable

If the AI service is slow or down, or your organization reached its daily AI allowance, a simpler rule-based reading is used and the case still arrives, classified and routed. Nothing is lost; the case notes that a simpler answer was used.

The website widget's assistant uses the same allowance: each answer it writes is an AI call. When the allowance is used up it still lists matching articles and guides requests, without writing answers.

The daily AI allowance is a fair-use limit of 1,000 AI calls per service desk per day (reset at 00:00 UTC), far above normal use. It exists so that a mail loop or a misbehaving integration cannot use up the AI for everyone. If your organization needs more, contact ServAIceDesk support.