People & HR

How to run an employee engagement survey with AI

People & HR3 AI tools7 steps6 friction points

Want an agent to handle this on your live data? See the Starch version →

Running an employee engagement survey means more than sending out a Google Form. It involves writing questions that actually surface what's bothering people, distributing the survey at the right cadence, aggregating responses, and turning raw data into a report leadership will act on. For most operators, this lands on the founder or a single HR generalist who also has twelve other things due that week.

The workflow feels like a natural fit for AI because so much of it is language work: drafting neutral, unbiased questions, writing the intro email to staff, summarizing open-ended responses, and framing findings for a leadership deck. These are tasks where a capable language model can genuinely compress hours into minutes — especially the question-writing and synthesis steps that would otherwise require careful iteration.

ChatGPT, Claude, and Gemini can all contribute meaningfully here. They're good at generating survey question sets by category (manager effectiveness, psychological safety, workload), rewriting questions to reduce leading bias, and summarizing qualitative themes from pasted responses. You won't get end-to-end automation, but you can cut the drafting and synthesis time significantly if you work through the workflow in structured steps.

Skip the manual work → See the Starch version

People & HR3 AI tools7 steps6 friction points

AI walkthrough

How to do it with AI today

A practical walkthrough using ChatGPT, Claude, and other off-the-shelf LLMs — what they're good at, what you'll have to do by hand.

Tools that work for this

ChatGPTClaudeGemini

Step-by-step

1 Open Claude or ChatGPT and prompt it to generate a 15-20 question employee engagement survey, specifying the categories you care about — for example, manager effectiveness, workload, belonging, and growth opportunity. Review the output and cut or rewrite questions that feel leading or vague.

2 Paste your draft questions back into the chat and ask the model to flag any that might make employees feel identified or uncomfortable responding honestly. Use the critique to revise sensitive items before sending.

3 Use ChatGPT or Claude to write the survey invitation email. Include context: why you're running it now, how responses will be used, and whether answers are anonymous. Ask the model to match your company's tone — paste in a sample of your past all-hands communications as a reference.

4 Distribute the survey using whatever tool you already use — Google Forms, Typeform, or a direct email. This step is entirely manual; the LLM is not connected to your distribution channel.

5 Once responses come in, export the open-ended text answers and paste them into Claude in batches. Prompt it to identify recurring themes, sort them by frequency of mention, and flag any responses that suggest urgent issues needing immediate follow-up.

6 Paste the quantitative score summaries (question averages, department breakdowns) into ChatGPT alongside the qualitative themes and ask it to draft an executive summary — a 1-page narrative of findings, key concerns, and suggested next steps.

7 Take that draft summary into whatever slide tool you use and build the leadership presentation manually. The model's output is a starting point; formatting and visual layout are on you.

Prompts you can copy

Generate a 15-question employee engagement survey covering: manager effectiveness, workload balance, psychological safety, career growth, and company communication. Avoid leading language. Format as numbered questions with a 1–5 scale where relevant.

Here are 20 draft survey questions. Review each for potential bias, identify any that might feel threatening to answer honestly in a small team, and suggest revised wording for those items.

Write a 150-word email inviting our 30-person team to complete our quarterly engagement survey. The survey is anonymous, takes 8 minutes, and closes Friday. Tone should be direct and low-pressure — not corporate.

Here are 47 open-ended survey responses to the question 'What's one thing you'd change about how we work?' Identify the top 5 recurring themes, note how many responses touch each theme, and flag any responses that suggest an urgent or serious concern.

Here's our engagement survey score summary by department and the top qualitative themes we identified. Draft a 300-word executive summary for our leadership team with key findings, 2-3 areas of concern, and 3 concrete suggested actions.

Reality check

Where this gets hard

The walkthrough above works — until your numbers change, the LLM hallucinates, or you have to re-paste everything next month.

No connection to your HR system or email platform — you manually export response data from Google Forms or Typeform and paste it in every single time.

Open-ended response analysis requires batching: paste 30 responses, summarize, paste the next 30. With a team over 50, this becomes a multi-session process and themes drift between batches.

Nothing persists between survey cycles — the question bank you refined last quarter, the themes you identified, the exec summary format you liked — all of it lives in a chat window you'll probably lose.

Distribution is entirely manual. The LLM can write the invitation email, but sending it to your actual team, tracking who responded, and sending reminders are all steps you do yourself in separate tools.

Output formatting is inconsistent between runs. The structured question categories you carefully prompted in March won't automatically carry over when you run the same workflow in June — you're re-prompting from scratch.

No place for the findings to live after the session. The summary sits in a chat or a copied doc, disconnected from prior surveys, making trend analysis across quarters a manual archaeology project.

Tired of the friction?

Starch runs the whole workflow on live data — no copy-paste, no hallucinated numbers, no re-prompting next month.

See the Starch version →

Starch alternative

The same workflow on Starch

Starch is an agentic operating system — you describe the engagement survey workflow you want, and an agent builds the persistent app that runs it continuously against your live people and communication data, instead of you re-prompting from scratch each quarter.

Connect Gmail or Outlook from Starch's integration catalog and Starch syncs your inbox; the Email Agent can draft and send the survey invitation and reminder emails directly, so distribution isn't a separate manual step.

Use the Knowledge Management app to store your live question bank, past survey findings, and recurring themes — so each new survey cycle starts from documented institutional knowledge, not a blank chat window.

Describe what you want in plain English: 'Build me an engagement survey tracker that stores each quarter's scores by department, surfaces themes from open-ended responses, and flags any category that dropped more than 5 points.' Starch builds the app.

Because Starch connects to Slack and syncs channel data, you can build an automation that posts the survey link to the right channels on a schedule and tracks response completion — no manual reminder emails.

When findings are ready, the Presentation Agent (currently in development — request beta access) will build the leadership deck from your summary data; in the meantime, Starch can build a dashboard that shows score trends across quarters so leadership has a live view without a new slide deck every cycle.

Everything persists. Your question sets, response themes, score history, and exec summaries live in Starch and stay connected — so the next survey cycle takes an hour, not a day.

Get closed-beta access →

Toolkit