As of 13 August 2026, AI can detect emerging issues in your customer feedback.

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 1 hour until you can act on the result.

   How to actually do it

   1. Export the relevant feedback for the latest period and an earlier comparison period, including dates, feedback text, channel, score, product or service area and any non-identifying customer segment.
   2. Remove names, email addresses, telephone numbers, order numbers and other details that could identify a customer, then record how many rows were removed or excluded.
   3. Open a chatbot or analysis tool and paste the prompt, replacing the bracketed fields with your date ranges, record counts, sources and operational context.
   4. Paste the anonymised feedback in manageable batches if it is too large for one message, and tell the tool which batch and period each section represents.
   5. Check every proposed issue against the quoted comments, row counts, duplicate records and the original period labels, correcting any theme that does not match the source data.
   6. Ask a customer-service or product colleague to test the highest-priority findings against known incidents, contact volumes and recent changes before sending an action to the wider team.
   7. Record the confirmed issue, owner, evidence, next check and review date in your existing service or product issue tracker.

   Prompt

   Analyse the anonymised customer feedback below for emerging issues. The data covers [DATE RANGE] and contains [NUMBER OR DESCRIPTION OF RECORDS], with these fields: [LIST FIELDS]. The feedback comes from [CHANNELS OR SOURCES], and the relevant product, service or operational context is [CONTEXT].
   
   Do the following:
   1. Cleanly separate observations from interpretations, and do not invent missing data.
   2. Group comments into specific issue themes, allowing more than one theme where justified.
   3. Compare the latest period with the earlier comparison period, using the periods and counts provided rather than claiming statistical significance.
   4. Rank possible emerging issues by evidence, showing the number or proportion of relevant comments where the data supports it, the change between periods, and two or three representative quotes with personal details removed.
   5. Distinguish a genuinely increasing issue from a one-off complaint, a change caused by feedback volume, duplicated comments, or a change in how feedback was collected.
   6. State what evidence is missing and give each finding a confidence label of high, medium or low with a reason.
   7. Suggest a practical human check for each high or medium priority issue, followed by a proportionate next action. Do not diagnose customers, identify individuals, or recommend action based only on sentiment.
   
   Return: a short executive summary, a table of ranked issues, evidence and limitations, then the human checks and suggested next actions. Use plain UK English. Treat the result as an investigation shortlist, not proof that an issue exists.
   
   Feedback data:
   [PASTE ANONYMISED FEEDBACK HERE]

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