Turn unstructured intake into a standardized title and description a dispatcher or
technician can act on immediately — restructured, not summarized away.
1. Read the raw source and extract: WHO (contact + <client>), WHAT (the symptom,
verbatim where useful), WHEN (onset, recurrence), IMPACT (users/systems affected,
urgency signals), and any error text.
2. Write the TITLE in the house convention. Default when the desk has not defined
one: "<Client> - <System/Device> - <Symptom>". Keep it under ~80 characters,
symptom-first, no "URGENT!!!" styling.
3. Write the DESCRIPTION as structured intake:
- Reported by / affected user(s)
- Symptom and exact error messages
- Onset and frequency
- Business impact
- Environment facts stated in the source (<device>, OS, application)
- Original source text preserved below a separator when it adds detail.
Do NOT put troubleshooting steps in the description — suggested or attempted fixes
belong in a separate internal note. The description records the problem, not the work.
4. Form-field parsing: if the source is a structured form or alert email, map each
labeled field to the intake sections; keep unmapped fields verbatim in a "Raw
fields" block rather than dropping them.
5. If a follow-up or callback date is requested, compute it weekend-adjusted: a
follow-up landing on Saturday or Sunday moves to the next business day (respect
stated desk holidays if known). State the adjusted date explicitly.
6. Restructure, do not summarize away specifics — exact error text, names, and
timestamps survive intact. Never place troubleshooting content, credentials, or
internal commentary in the client-visible description. If the desk runs a non-
English working language, produce the title/description in that language.
7. Only create or modify a ticket AFTER confirming the extracted details are right;
never guess the client/contact from name similarity — low confidence means ask,
do not assign. Then create the ticket (or update it when reformatting) with the
right board/status/priority; use the assistive AI for category/priority
suggestions when requested. Post attempted-fix info, if any, as an internal note.
