auto-categorization of tickets with keyword rules
Auto-categorizes ITSM tickets using keyword rules. Covers rule design, confidence thresholds, and the human review loop. Use to speed triage. Not a replacement for agent judgment on ambiguous tickets.
TL;DR
Build keyword rules for your top categories (password, VPN, hardware, access), apply them at creation, and show the suggested category to the agent for one-click confirm or correct. Auto-apply only above a high confidence threshold; everything else suggests. Review corrections weekly to improve the rules.
The error
(Automation task; no error.)Steps
- Mine 500 recent tickets for the keywords that actually predict each top category. Expected: keyword lists grounded in your data, not guesses.
- Write the rules: ServiceNow business rules, Jira automation, or flow designer matching keywords in the short description. Expected: rules created with explicit priority order.
- Set two thresholds: auto-apply above 95 percent historical precision, suggest-and-confirm below that. Expected: configured. Auto-applying low-confidence rules creates miscategorized tickets nobody trusts.
- Show the agent the suggested category prominently with one-click correct. Expected: in the agent workspace. The correction loop is what improves the system.
- Review corrections weekly: promote corrected patterns into rules, retire rules that misfire. Expected: maintained. Rules decay as language changes.
When to use
- High ticket volume with clear language patterns
- Speeding up triage and reporting accuracy
When not to use
- Low volume or highly varied tickets
- Categories that need human judgment (security incidents)
Compatibility
- ServiceNow, Jira Service Management, or any ITSM with automation rules
Variants
ML categorization
Train on your corrected data after the rules are stable; ML handles the long tail.
Category-driven routing
Feed the category into assignment rules; categorization plus routing compounds the value.
Why it happens
Category drives routing, reporting, and knowledge suggestions. Manual categorization is inconsistent, and inconsistent categories make every downstream metric lie.
Edge cases
- Multilingual tickets need per-language keyword sets.
- Do not categorize security or HR-sensitive tickets automatically; route those by rule to the right queue uncategorized.
Provenance
Resolved from the public thread: https://vectle.com/posts/pst_IM3t02cRuJMDDHRoPs0wHw
Maintainer review
No maintainer verification is recorded for this version.
This records the version a maintainer checked. It does not assert that the version is the latest upstream release.