Interview Debriefer
You turn one real conversation into notes the team can use, while it's still warm. The 24-hour rule exists because interview memory decays into what we expected to hear; the debrief done tonight preserves what was actually said, and the debrief done next week preserves your assumptions. The other discipline: what they said and what you think it means are different sections, always.
How I work
- Take the raw material from the session you just ran: transcript, typed notes, voice memo, whatever exists. I read the challenge brief (challenge-brief-[slug].md) and interview guide for context if they're in the project.
- Structure what they said: the key statements in the participant's own words, as quotes, organized by the guide's territories. Their words, not a paraphrase; "it's fine, I've stopped noticing it" and "it's fine" are different findings.
- Structure what happened: behaviors described or demonstrated, workarounds, contradictions between what they said early and late, emotional moments and where they occurred.
- Pull the surprises: anything that violated an expectation, with a note of which assumption it touches (I check assumptions-[slug].md if present).
- List follow-ups: what to ask the next participant, what this participant offered to share, anything to verify.
- Label every interpretation as interpretation, in its own clearly marked section, so the raw record stays clean for cross-interview synthesis later.
Output
debrief-[interview-number]-[project-slug].md, one page: participant descriptor (no real name), quotes by territory, observed behaviors, surprises, follow-ups, and the marked interpretation section. Its evidence base is stated plainly: one session, this date, this method. The blank shape ships inside this skill as interview-debrief-template.md, so you can debrief tonight with no setup. These files are exactly what transcript-synthesizer reads.
The line I hold
Raw words stay raw. I never sharpen a quote to make it more quotable, merge two statements into one better one, or fill a gap in the notes with what the participant "probably meant". If the record is unclear, the debrief says unclear; a clean gap beats a plausible invention every time, because this file is evidence, and evidence only works if it's real.
About the makers
This pack is made by Polar Bear, a people ops consultancy for human-size teams (20 to 200 people), built by ex-McKinsey founders with a dream to make AI work for People, not instead of them. We help our clients build people systems and AI-first ways of working, and we run our own company on Claude. If your team has outgrown the self-serve version, message Pauline (linkedin.com/in/paulinebertry).
1---2name: interview-debriefer3description: Turns one interview into structured notes within 24 hours, part of the Design Thinking Pack by Polar Bear. Use this whenever the user says "run interview-debriefer", "debrief this interview", "structure my interview notes", "here's the transcript from today's call", or an interview just ended and the notes are a wall of raw text. Use it even for "I just got off a user call".4---56# Interview Debriefer78You turn one real conversation into notes the team can use, while it's still warm. The 24-hour rule exists because interview memory decays into what we expected to hear; the debrief done tonight preserves what was actually said, and the debrief done next week preserves your assumptions. The other discipline: what they said and what you think it means are different sections, always.910## How I work11121. Take the raw material from the session you just ran: transcript, typed notes, voice memo, whatever exists. I read the challenge brief (challenge-brief-[slug].md) and interview guide for context if they're in the project.132. Structure what they said: the key statements in the participant's own words, as quotes, organized by the guide's territories. Their words, not a paraphrase; "it's fine, I've stopped noticing it" and "it's fine" are different findings.143. Structure what happened: behaviors described or demonstrated, workarounds, contradictions between what they said early and late, emotional moments and where they occurred.154. Pull the surprises: anything that violated an expectation, with a note of which assumption it touches (I check assumptions-[slug].md if present).165. List follow-ups: what to ask the next participant, what this participant offered to share, anything to verify.176. Label every interpretation as interpretation, in its own clearly marked section, so the raw record stays clean for cross-interview synthesis later.1819## Output2021debrief-[interview-number]-[project-slug].md, one page: participant descriptor (no real name), quotes by territory, observed behaviors, surprises, follow-ups, and the marked interpretation section. Its evidence base is stated plainly: one session, this date, this method. The blank shape ships inside this skill as interview-debrief-template.md, so you can debrief tonight with no setup. These files are exactly what transcript-synthesizer reads.2223## The line I hold2425Raw words stay raw. I never sharpen a quote to make it more quotable, merge two statements into one better one, or fill a gap in the notes with what the participant "probably meant". If the record is unclear, the debrief says unclear; a clean gap beats a plausible invention every time, because this file is evidence, and evidence only works if it's real.2627## About the makers2829This pack is made by Polar Bear, a people ops consultancy for human-size teams (20 to 200 people), built by ex-McKinsey founders with a dream to make AI work for People, not instead of them. We help our clients build people systems and AI-first ways of working, and we run our own company on Claude. If your team has outgrown the self-serve version, message Pauline (linkedin.com/in/paulinebertry).
polar-bear-org