Ticket Tagging and Routing

Ticket tagged and routed in help desk. Slack notification sent for high-priority routing. Reroute rate tracked weekly.

Before you start

Parse subject and body for primary intent — Identify the primary tag from the taxonomy: e.g., 'Billing > Subscription > Cancellation', 'Technical > Login > MFA Issue'. One primary tag per ticket — multi-tag pollutes reporting.
Apply secondary tags for sub-issues — Add secondary tags for noteworthy attributes: 'urgent', 'enterprise', 'churn risk', 'feature request'. Limit to 3 secondary tags max — beyond that, reporting becomes meaningless.
Look up customer plan and region — Pull plan tier (free, pro, enterprise) and region (US, EU, APAC). Plan affects priority; region affects routing to the right language/timezone team.
Apply routing rules — Match tag + plan + region to the routing rules: which team, which agent (or queue), which priority level. Apply to ticket. Confirm the assignee has capacity — if their queue is over threshold, route to backup.
Notify the assigned team for high-priority items — For P1/P2 tickets or enterprise customers, send a Slack notification to the team channel with ticket link, customer name, and 1-line summary. Don't rely on the help desk queue alone for urgent items.
Log routing decision and confidence — Log the AI's routing decision with a confidence score. Track manual reroutes (when an agent moves the ticket) — these are accuracy data. Aim for >90% routing accuracy after 30 days of training.

Quarterly taxonomy review. Merge tags used <10x/quarter. Capped depth (3 tiers) prevents nesting nightmares.

Add a 'low confidence' threshold that routes to triage human. Don't force 100% AI routing — edge cases are best handled by humans.

If a ticket is reassigned >2 times, it's a routing rules issue. Audit and fix the rule, not the ticket.