As of 13 August 2026, AI can summarise your customer survey responses.

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

   How to actually do it

   1. Export the survey results from your survey platform as a spreadsheet or text file, keeping the question text, response text and any non-identifying segments.
   2. Remove names, email addresses, phone numbers, order numbers and other personal identifiers before opening a chatbot.
   3. Open ChatGPT, Claude or Gemini and paste the prompt, replacing the bracketed fields with the survey purpose, audience, collection period, questions and response data.
   4. Ask the chatbot to calculate counts only where the supplied data supports them, and to label interpretations, limitations and suggested actions separately from customer statements.
   5. Copy the summary into a working document and sample responses from every reported theme, comparing each quoted excerpt and count with the original export.
   6. Ask a colleague who understands the survey and customer context to check whether minority views, duplicate responses, sampling limits and important complaints have been represented.
   7. Send the checked summary to the relevant product, service or customer-care owner, keeping the original export alongside it as the evidence record.

   Prompt

   Summarise the customer survey data below for a UK business team.
   
   Survey purpose: [WHAT THE SURVEY WAS TRYING TO LEARN]
   Audience and collection period: [WHO RESPONDED AND WHEN]
   Questions: [PASTE THE QUESTIONS]
   Responses: [PASTE OR ATTACH THE RESPONSES]
   Useful segments, if available: [FOR EXAMPLE PRODUCT, REGION, CUSTOMER TYPE OR PLAN]
   
   Remove or ignore names, email addresses, phone numbers, order numbers and other identifying details. Do not infer personal characteristics that are not supplied. Do not invent responses, counts or explanations.
   
   Produce:
   1. A short executive summary with the main findings.
   2. The main positive themes and the main negative themes.
   3. For each theme, give its response count if it can be calculated, otherwise say that no reliable count is available.
   4. For each theme, include up to three short verbatim excerpts, clearly labelled as excerpts, and do not alter their meaning.
   5. Separate direct customer statements from your interpretation.
   6. Identify conflicting or minority views rather than smoothing them into the majority view.
   7. Highlight unanswered questions, possible sampling bias and duplicate or low-quality responses.
   8. Suggest practical follow-up questions or actions, clearly labelled as suggestions rather than findings.
   
   Use plain British English, headings and a table where useful. Finish with a section called "Checks for the team" listing the source responses or calculations we should inspect before acting on the summary.

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