As of 13 August 2026, AI can categorise your customer feedback automatically.

What to actually do

1. Hand it to a person

   The route this page recommends

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

2. Use a tool built for this

   Second choice

3. Do it yourself

   The distant third

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

   How to actually do it

   1. Open the customer feedback export and remove names, email addresses, phone numbers, order numbers and any other identifying information before copying the text.
   2. Write a short definition and one or two examples for each category you want the model to use, including a needs-review category for unclear cases.
   3. Paste the prompt into ChatGPT, Claude or Gemini, then replace the bracketed sections with your category definitions and anonymised feedback.
   4. Ask the model to process a small sample first and compare its labels with your definitions and the original feedback.
   5. Adjust category definitions where the sample produces repeated ambiguities, then run the full anonymised dataset with the revised prompt.
   6. Give a colleague a random sample of the original feedback and the AI labels, and have them check the category, urgency and human-follow-up flag against your service rules.
   7. Send only the checked categories to your reporting or ticketing workflow, keeping complaints and other flagged cases for a person to handle.

   Prompt

   Categorise the anonymised customer feedback below for a UK business.
   
   Use only these categories: [PASTE CATEGORY LIST].
   For each item, return:
   - feedback_id
   - primary_category
   - optional_secondary_category
   - sentiment: positive, neutral or negative
   - urgency: low, medium or high
   - needs_human_follow_up: yes or no
   - a short reason using the customer's own meaning
   
   Rules:
   - Do not invent facts, customer details or causes.
   - Do not treat negative sentiment as proof that the business did something wrong.
   - Mark needs_human_follow_up as yes for complaints, threats of escalation, safeguarding concerns, requests involving personal data, refund disputes, or feedback that is too ambiguous to classify confidently.
   - Keep the customer's wording intact where a quote is necessary, but do not repeat names, email addresses, phone numbers, order numbers or other identifying information.
   - If an item does not fit the category list, label primary_category as needs-review and explain why.
   - If two categories are plausible, choose the best primary category and state the ambiguity in the reason.
   - After the table, give counts by primary category and list the items needing human follow-up.
   
   Category definitions and examples:
   [PASTE DEFINITIONS AND EXAMPLES]
   
   Anonymised feedback:
   [PASTE FEEDBACK, ONE ITEM PER LINE WITH AN ID]

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