Turning open-text survey comments into themes leaders can act on
An engagement or pulse survey can return hundreds of free-text comments, and reading them into a coherent set of themes is slow, subjective work that often stalls before leadership ever sees the results. AI clusters the comments with representative quotes in minutes. The catch is that the comments are sensitive and frequently identifiable, so anonymization has to come first.
You are a people-analytics assistant. I will paste anonymized open-text responses from an employee survey. Identify the main themes without exaggerating or inventing sentiment. Survey question: {{survey_question}} Anonymized responses: {{responses}} Produce: 1. 5-8 themes, ranked by how often they appear. For each: a short label, an approximate share of comments, and 1-2 lightly paraphrased representative quotes. 2. Sentiment per theme (positive, mixed, or negative). 3. "Signal vs. noise": which themes are broad patterns versus one or two strong voices. Do not overweight a single vivid comment. 4. Suggested clarifying questions for a follow-up survey — not conclusions about individuals. Rules: - Base themes only on the responses provided. Do not infer who wrote a comment or attribute views to any person, team, or manager by name. - If a response names a specific person, replace the name with [REDACTED] and flag it. - Do not paraphrase in a way that changes meaning; preserve critical feedback honestly. - Report counts as approximate ("about a third"), never as false precision.
Fill in your details and the prompt updates live — then copy.
Theme 1 — Workload and staffing (about 40% of comments, negative): repeated mentions of covering vacant roles. Representative: "We've been down two people for months and it shows." Theme 2 — Manager support (about 25%, mixed): some praise their direct manager; others want more regular 1:1s. Signal vs. noise: Workload is a broad pattern; the two comments about the parking lot are isolated (noise), not a top theme. Flag: One response named a specific manager — replaced with [REDACTED]. Handle that comment through your ER process, not this summary.
The full workflow
- Export the free-text responses and strip names, IDs, and obvious identifiers before pasting
- Run the analysis in an approved enterprise tool, not a personal account
- Spot-read a sample of raw comments to confirm the themes and quotes are faithful
- Layer your own judgment on the themes, and route any comment describing harassment or safety to the proper process immediately
Watch out for
Survey comments feel anonymous but are often re-identifiable from the detail. Anonymize before pasting, use an enterprise tool with no-training and retention controls, and never load raw comments into a personal AI account — breaking survey confidentiality destroys trust and can violate the anonymity you promised.
AI can smooth over or overstate sentiment. Verify themes and quotes against the raw text before you report to leadership.
Comments describing harassment, discrimination, or a safety risk are not just data. Route them into your investigation or ER process; an AI summary is not a response.
Where this comes from
Every use case on this site is grounded in real reports from working hr managers — not invented by us.