Verified against ChatGPT · 2026-08-12
Design a team health check that surfaces the thing people won't say out loud in the regular team meeting
Builds a lightweight, recurring team health check tuned to specific dynamics this team is actually navigating, structured to reduce the social risk of an honest answer so it catches problems before they show up in exit interviews.
The prompt
Ready to copy — highlighted parts are example details you can swap.
Design a recurring team health check for the team described below — a short, regular pulse check on team dynamics, not a full annual engagement survey — tuned to what's actually going on with this team right now. TEAM CONTEXT 6-person design team, has been stable for a year but recently added two new hires in the last quarter. SPECIFIC DYNAMIC YOU'RE WORRIED ABOUT Suspect the two newest hires feel like they can't push back on design decisions in critique sessions because the two senior designers dominate the conversation. HOW OFTEN AND HOW THIS GETS RUN Monthly, anonymous form, results shared back in the following team meeting. WHAT'S BEEN TRIED BEFORE THAT DIDN'T SURFACE ANYTHING USEFUL We ran a generic 'rate your happiness 1-10' survey twice last year and got flat 7s and 8s from everyone with no comments, so we stopped. Design 5-7 questions, mixing a few numeric or scale ratings (so trends are trackable over time) with at least two open-ended questions specifically worded to lower the social risk of an honest negative answer — for example, asking "what's one thing that would make this team work better" rather than "is anything wrong," since the first is framed as constructive input anyone can answer honestly while the second requires someone to affirmatively name a problem, which people are far less willing to do in a small team where anonymity is thin. If a specific dynamic is named as a concern (uneven workload, one person dominating discussion, quiet disengagement from a specific subgroup), include at least one question aimed at surfacing that exact pattern without naming a person or being so specific that a single respondent's answer is instantly identifiable. If a past attempt is described as having failed to surface anything useful, diagnose in one sentence why it likely failed (too infrequent, too generic, answers were visible to the whole team, no visible follow-through last time) and design this version specifically to avoid that failure. WHAT TO DO WITH THE RESULTS Include a short closing note on how results should be shared back to the team — always in aggregate, never traceable to an individual respondent on a small team, and always with at least one visible action taken as a result, since a team health check that produces no visible follow-through trains people to stop answering honestly the next time. OUTPUT FORMAT The question list (numbered, noting scale vs. open-ended), a one-line note on why each open-ended question is worded the way it is, and the closing note on sharing results and follow-through.
Customize
Optional — swap in your own details for the highlighted parts above.
Why this works
The near-universal failure of generic "rate your happiness 1-10" pulse checks, exactly like the past attempt described in this prompt, has a specific mechanistic cause: a bare numeric scale with no concrete framing gives respondents nothing to anchor an honest answer to, so people default to a socially safe middle-to-positive number rather than doing the harder cognitive work of identifying and naming a real problem, which produces flat, uninformative results that look fine on paper regardless of the team's actual state. Wording open-ended questions as constructive asks ("what would make this better") rather than problem-identification asks ("is anything wrong") lowers the specific social cost of an honest answer, because suggesting an improvement is a normal, low-risk act of engagement while naming a problem implicitly requires blaming a cause, which respondents in a small team — where anonymity is thin because there are only a handful of possible authors for any given comment — are reasonably reluctant to do directly. Targeting a named suspected dynamic (senior designers dominating critique) with a carefully generic-enough question threads a real design constraint: specific enough to actually surface the pattern if it's real, but not so specific that a single respondent's answer would be instantly attributable, which matters because if respondents believe their answer is identifiable, the entire mechanism of honest anonymous feedback collapses regardless of how the question is worded. The requirement to close the loop with visible action addresses the single most common reason pulse checks degrade over repeated cycles: people calibrate their willingness to answer honestly based on whether past honesty produced any visible change, so a health check that surfaces real feedback but is followed by no visible action trains the team to stop bothering with honest answers well before the underlying problem is fixed.
What you get back
Q3 (open-ended): 'What's one thing that would make design critique sessions work better for you?' — worded as a constructive ask rather than 'is critique working,' which requires naming a problem directly. Closing note: results shared in aggregate only at the next team meeting, with at least one specific change (e.g., a rotating critique-lead role) named as a direct response to what came up.
Verified against
ChatGPT GPT-5.1 · 2026-08-12
Changelog
- 2026-08-12 — Initial publish, verified against ChatGPT GPT-5.1.
Need this built into your business?
If a prompt isn't enough — what Scult builds, built and maintained for you — that's Scult's day job.
EXPLORE WHAT SCULT BUILDS
