When to use it
When you need to understand what "job" people hire your (or a future) product to do — before you build it. The role is a JTBD researcher. Output: an interview guide with questions about real past experience (not "would you buy this?"), rules that keep you from fooling yourself, and a template for analyzing the transcript.
The prompt (copy and paste)
You are a Jobs-to-be-Done researcher. Help me prepare a problem interview.
PRODUCT/IDEA: "<PASTE>". WHO I'M INTERVIEWING: <segment>. WHAT I WANT TO UNDERSTAND: "<hypothesis about the pain/job>".
Give me an interview guide:
1. The goal of the interview in one sentence and the 2-3 key hypotheses we're testing.
2. 10-12 OPEN questions about REAL past experience, in JTBD logic: context and trigger ("tell me about the last time you ran into…"), how they solved it, which alternatives they tried and why they dropped them, what was hard/annoying, what criteria they chose by, how it ended up.
3. Mark which questions are about past facts (good), and warn me which ones NOT to ask (about the future/hypotheticals: "would you buy this?", leading questions, questions that hint at the answer).
4. 3 follow-up "digging" questions ("why did that matter?", "what did you do next?").
5. An analysis template: where to record triggers, pains, current solutions, choice criteria, verbatim quotes.
Keep the tone of the questions neutral, with no pitching of the product.
Filled-in example
Idea: a nanny matching service. Who I'm interviewing: working parents of 1-7 year olds. What I want to understand: how they find a nanny today and what hurts about it.
What the AI should come back with: the goal and hypotheses; 10-12 questions — "Tell me about the last time you urgently needed someone to watch your child", "How exactly did you look, who did you turn to?", "What did you try before and why did you stop?", "What was the most annoying part?", "How did you decide this nanny could be trusted?"; notes like "this one is about the past — fine", a warning not to ask "would you use an app for this?"; digging questions like "why did that matter?"; an analysis table (trigger / pain / current solution / criterion / quote).
Variations
- Analyze a recording. Paste the transcript: "pull out triggers, pains, alternatives and strong quotes using my template".
- Roll-up across a series. "Here are 5 interviews — what repeats, what patterns show up, where is there consensus and where do people diverge?"
- B2B version. Built around the buying process: roles in the decision, criteria, who blocks it.
Pro tips
- The golden rule: ask about past behavior, not future intentions. "Would you buy this?" lies; "how did you handle it last time?" tells the truth. Insist that the AI only produce questions about real experience.
- Stay quiet and dig: the best insights come after "why did that matter?" plus a pause. The point of the guide is to keep you from selling and from feeding people the answer.
- The AI prepares the guide and analyzes transcripts, but the interviews are run by real people with real people. Record only with consent (this is personal data) and analyze what was actually said, not what you hoped to hear.