How To Talk To Users | Startup School
The canonical 17 minutes on user interviews; everything the AI helps with builds on this.
Watch on YouTube youtube.com →Do your first 10 to 15 interviews yourself using Mom Test rules (past behaviour, not opinions about your idea), then let AI take over the mechanical work: drafting non-leading question guides, transcribing, and clustering themes across transcripts. Claude and ChatGPT are excellent at turning a pile of raw transcripts into themes, quotes, and contradictions in minutes, and AI-moderated platforms can then run hundreds of consistent interviews once you know what to ask. Treat AI's themes as a first draft to check against the recordings, not as the conclusion.
A quick orientation. The real value is below: resources worth your time, from people who've actually done it.
The canonical 17 minutes on user interviews; everything the AI helps with builds on this.
Watch on YouTube youtube.com →The same essential material in audio form for commute listening.
Listen on Spotify open.spotify.com →The Mom Test author himself draws the line between what AI should and should not do in discovery.
Watch on YouTube youtube.com →Watch transcripts become themes and visuals, the core analysis move this question is about.
Watch on YouTube youtube.com →End-to-end demo of the transcribe-then-analyze loop founders can copy.
Watch on YouTube youtube.com →A ready question bank that keeps your AI-drafted guides non-leading.
Open getperspective.ai →Updates Mom Test practice for the AI-moderated interview era.
Open koji.so →The hybrid model: humans build intuition first, AI scales the interviewing after.
Open prelaunch.com →An honest technical account of what LLM qualitative analysis gets right and wrong.
Open skimle.com →A reusable Claude workflow that converts messy transcripts into themes and actions in 90 seconds.
Open sidsaladi.substack.com →A copy-paste transcript analysis prompt to start with before building anything fancier.
Open genaiuniversity.com →Shows Claude Code processing many interviews in parallel, the power-user version of this workflow.
Open marketbetter.ai →A professional research firm's assessment of Claude for qualitative analysis, limits included.
Open cascadeinsights.com →A working researcher's real AI workflow rather than a tool vendor's pitch.
Open userinterviews.com →Primary research on AI running interviews at scale and feeding results back to humans.
Open anthropic.com →A leading AI-moderated interview platform for when you need dozens of interviews this week.
Open outset.ai →How AI moderation removes the depth-versus-scale trade-off in discovery.
Open listenlabs.ai →Paste your interview questions and it flags the leading ones before you embarrass yourself.
Open momtest.ai →A full founder reference for discovery interviews with AI assistance woven in.
Open userintuition.ai →The written companion to the video, easy to excerpt into your AI question-drafting prompts.
Open ycombinator.com →