As of 13 August 2026, AI can categorise your customer support tickets.

What to actually do

1. Use a tool built for this

   The route this page recommends

2. Do it yourself

   Second choice

   A chat interface, chat-fluent skill, and roughly 30 minutes until you can act on the result.

   How to actually do it

   1. Open your support knowledge base or internal category list and write one clear definition for each category, including examples of what belongs there and what must be escalated.
   2. Gather a representative batch of already resolved tickets and record the correct category, destination team, and urgency for each one, removing unnecessary personal information.
   3. Paste the category definitions, routing rules, and labelled examples into the prompt, then paste the new tickets with an ID beside each ticket.
   4. Ask the chatbot to return the specified table and to mark ambiguous, urgent, complaint, security, and out-of-scope tickets for human review.
   5. Compare a sample of the classifications against the original ticket text and your current routing rules, correcting the taxonomy or examples where the same type of ticket receives inconsistent labels.
   6. Send only the checked classifications to your helpdesk workflow, and give a colleague the tickets marked for human review before any customer-facing action is taken.

   Prompt

   You are classifying customer support tickets for a UK business.
   
   Use only the taxonomy, routing rules, and examples supplied below. Do not invent categories, customer facts, product details, or policy decisions.
   
   Category definitions:
   [PASTE YOUR CATEGORY NAMES AND A ONE-SENTENCE DEFINITION FOR EACH]
   
   Routing rules:
   [PASTE RULES FOR URGENCY, TEAM, PRODUCT, REGION, OR ESCALATION]
   
   Labelled examples:
   [PASTE EXAMPLES OF TICKETS WITH THE CORRECT CATEGORY AND ROUTING]
   
   Tickets to classify:
   [PASTE THE TICKET ID AND TICKET TEXT]
   
   For each ticket, return one row with these fields:
   - ticket_id
   - category
   - subcategory, or null if none is defined
   - urgency: routine, priority, or urgent
   - destination_team, or null if the rules do not specify one
   - confidence: high, medium, or low
   - reason: one short sentence quoting or referring to the relevant ticket wording
   - needs_human_review: yes or no
   
   Use needs_human_review=yes when the ticket matches more than one category, lacks enough information, contains a complaint or threat of escalation, suggests a safety or security issue, or does not clearly fit the supplied rules. Do not treat low confidence as a reason to guess. Return the results as a table, followed by a separate list of tickets needing human review. Preserve every ticket_id and do not silently omit any ticket.

   Open it prefilled in ChatGPT or Claude, or copy it into Gemini, which takes no prefill link.

3. Hand it to a person

   The distant third

   A person who owns the outcome does this end to end, worth it when the failure is dear.