As of 13 August 2026, AI can identify your NPS promoters.

What to actually do

1. Do it yourself

   The route this page recommends

   A chat interface, chat-fluent skill, and roughly 15 minutes until you can act on the result.

   How to actually do it

   1. Export the survey results from your NPS platform as a CSV or spreadsheet, including a stable respondent or customer ID, the NPS score and any response text you want analysed.
   2. Remove columns that are not needed for this task, and check your organisation's approved AI and data-handling rules before uploading customer names or email addresses.
   3. Open an approved chatbot, paste the prompt, and attach or paste the survey export.
   4. Ask the model to identify scores of 9 or 10 only, preserve the original IDs and produce the separate exceptions table requested in the prompt.
   5. Filter the original export yourself for scores of 9 and 10, then compare the IDs, scores and row count with the model's promoter table.
   6. Resolve every duplicate, missing, invalid or unmatched row in the exceptions table, then send the checked list through your normal CRM or customer-contact process rather than asking the model to contact customers.

   Prompt

   You are analysing an NPS survey export for a UK business. Identify promoters using the standard NPS definition: respondents with a score of 9 or 10 are promoters. Do not classify anyone as a promoter from positive wording alone, and do not infer a score that is missing or unclear.
   
   I will provide a CSV or pasted table with some or all of these columns: respondent ID, customer name, email, NPS score, free-text response, survey date and account or segment. Use only the data supplied. Preserve the respondent ID exactly, retain the original score and flag duplicate IDs, missing scores, invalid scores and rows where the customer identity cannot be matched confidently.
   
   Return:
   1. A table of promoters containing respondent ID, customer name or account if supplied, email only if supplied, original NPS score, survey date if supplied, and the response text if supplied.
   2. A count of promoters.
   3. A separate table of excluded, ambiguous or invalid rows with the reason for exclusion.
   4. A short summary of recurring themes in promoter comments, quoting no more than three short examples and keeping them attributed only to respondent ID unless I explicitly provide permission to use names.
   5. A list of checks I must complete before contacting anyone.
   
   Do not invent missing fields, merge people with similar names, expose personal data unnecessarily, or recommend contacting a person. If the export does not contain an NPS score column, stop and tell me what is missing instead of guessing.

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

2. Use a tool built for this

   Second choice

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.