Survey Comments to Themes With Counts

When to use it

The user has open-text answers from a survey: a customer feedback form, an employee engagement survey, event feedback, NPS "why did you give that score" verbatims, a cancellation reason box. They paste the column and want to know what people said, how many said it, and which answers need someone to act today.

Also for a repeat survey: the skill reuses the earlier codebook so the counts can be compared wave to wave.

Not for: call notes or interview transcripts (that is Voice of Customer From Call Notes, a separate Move), numeric survey results (that is a table, not coding), or writing the survey itself.

What it asks you

Ask in one message, numbered. Work from what comes back.

1. Paste the answers, one per line. If a score or a segment (team, region, plan) sits beside each answer, keep it on the same line, separated by a tab or a pipe.
2. How many answers are there, so the count can be checked against what was pasted?
3. What was the question people were answering? Paste it exactly.
4. Who answered: customers, staff, event attendees, people who cancelled? Roughly how many were invited, if known?
5. What will the report be used for, and who reads it: a leadership update, a product backlog, an HR action plan, a marketing page?
6. Is there an earlier codebook from a previous wave to reuse? Paste it if so.
7. Is there anything that must be flagged as urgent on sight: safety, legal, harassment, a named person, a customer threatening to leave, a bug costing money?

If the paste has no clear line breaks, split on the most likely separator, print the count found, and ask the user to confirm before coding. If the count the user gave and the count found differ, stop and resolve it. Every number in the report depends on it.

Never invent an answer, a count, a segment or a quote. Quotes are the respondent's words exactly as pasted, including spelling.

What comes back

Plain text, headings in capitals.

SURVEY THEMES: survey name, wave or date Summary (five lines) CODEBOOK (code, definition, inclusion rule) THEME TABLE (code, count as N of total, percentage if total is 30 or more, sentiment split if applicable, small flag) Reconciliation: answers coded N of N; code assignments X; answers with two or more codes Y. SEGMENTS (code by segment, only where the segment has 5 or more answers) QUOTES (one per theme, verbatim, line number; outlier if any) URGENT (line number, the answer, why it is urgent, who should see it today) OTHER AND NON-ANSWERS (counts, and the "other" answers listed if under ten) NOT RAISED (codes the user expected that scored zero) Method: one sentence. For your stated use: three lines on what to do with the top themes. Next: one line naming the single most useful next step.

For a repeat wave: the same codebook, this wave's counts beside last wave's, and one line per code that moved by three answers or more.

• 00_START_HERE.pdf: the two-minute set-up for each AI.
• 1 - Skill: open-text-survey-to-themes.zip, one upload to Claude, ChatGPT, Gemini, Grok or Microsoft 365 Copilot. It switches on by itself when this job comes up.
• 2 - Text file for any AI: the same Move as one .txt. Attach it to any chat, on any plan, and type "Follow this file".

It drafts, you read and send. It never contacts anyone for you and never invents a fact, a number or proof.

A Move is a single play. The whole function in a box is an agent pack.