JTBD looks easy on a slide and almost always falls apart in practice: teams write “job statements” that are really feature descriptions, skip the switch trigger entirely, and end up using JTBD as a brainstorm warmup instead of a decision filter. These 12 prompts force the rigor back in — switch-trigger interview design, four-forces analysis, defensible job statements, functional/emotional/social decomposition, and using JTBD to actually cut features from a backlog. Pair them with the feature prioritization prompts once you have a job locked in.
TL;DR
- Use a model with a long context window for transcript work — Claude Sonnet 4.6 and Gemini 3.1 Pro both ship 1M-token context as of June 2026, so a 60-minute interview transcript (roughly 9,000–11,000 words) fits in one paste with room for instructions. ChatGPT Plus tops out around 320 pages in-app; the full 1M context only ships on the $200 Pro tier.
- JTBD has three competing schools, and the prompts below respect that: Bob Moesta and Chris Spiek’s switch interview (timeline + four forces), Tony Ulwick’s Outcome-Driven Innovation (one functional job, many measurable outcome statements), and Alan Klement’s job stories (“When… I want… so I can…”, no persona).
- A job statement names the progress a customer is trying to make; a job story adds situation and emotional context. Neither is a feature.
- Run 10–15 switch interviews before you trust the pattern — Moesta’s rule of thumb is that this covers the vast majority of actionable patterns in a customer base.
- The single highest-leverage prompt here is #7 (feature filter): JTBD only pays for itself when it kills work, not when it generates more.
Pick the right model first
These prompts lean on long, messy interview text and ask the model to hold a whole timeline in working memory. Context window matters more than raw benchmark scores here.
| Model (June 2026) | Context | Best for in this workflow |
|---|---|---|
| Claude Sonnet 4.6 | 1M tokens | Transcript analysis (#12), long alternatives maps; strong at structured extraction |
| Gemini 3.1 Pro | 1M tokens | Same; cheapest API at $2/$12 per 1M tokens in/out |
| Claude Opus 4.7 | 1M tokens | Hardest judgment calls — which job statement is defensible (#2), buyer-vs-user conflicts (#9) |
| GPT-5.5 (ChatGPT Plus) | ~320 pages in-app | Drafting and ideation (#1, #6, #11); paste transcripts in chunks |
Run the transcript-heavy prompts (#3, #8, #12) on a 1M-context model so you never have to split an interview mid-thought. Save Opus 4.7 for the prompts where the model has to choose rather than extract.
Best for
- Pre-MVP customer discovery
- Re-positioning an existing product
- Sharpening landing-page copy
- Feature prioritization
- Sales-call structure
1. JTBD interview question bank
Generate 15 JTBD interview questions for a {product} customer who just signed up / churned / referred. Cover: trigger, evaluation, decision moment, expected outcome, the alternative they almost picked.2. Job-statement writer
Below: interview notes. Write 3 candidate job statements using "When {situation}, I want to {motivation}, so I can {expected outcome}". Pick the most defensible one. Justify why the other two are wrong (e.g. they describe a feature, not progress).
{paste notes}3. Switch-trigger analysis (four forces)
My customer just switched from {alternative} to my product. Walk me through the JTBD switch interview as a timeline: first thought (passive looking), trigger event (active looking), evaluation, decision (purchase), early usage. Then map the four forces of progress: push of the old situation, pull of my product, anxiety about switching, inertia/habit. Output the interview questions in timeline order.4. Alternatives map (not just competitors)
For job statement {paste}, list the real alternatives: direct competitors, indirect tools, manual workarounds, "doing nothing", and hiring help. For each: when they win, when they lose, and what force (push/pull/anxiety/inertia) keeps the customer there.5. Functional-vs-emotional-vs-social job map
Below is my JTBD. Break it into functional, emotional, and social dimensions. For each, give 2 specific cues from interviews that signal that dimension. Flag which dimension my current product ignores.
{paste}6. JTBD → positioning
For job {paste}, write a positioning statement that names the situation, the desired progress, and the unique way we deliver it. Then 3 variants. Ban generic "best in class" / "industry leading" framing.7. JTBD → feature filter
Below: 15 features in my backlog. Below that: the primary job to be done. Score each feature as "advances the JTBD" / "tangential" / "serves a different job entirely". Cut anything that scores below "advances", and explain in one line per feature why.
{paste}8. JTBD-misalignment detection
Below: my current landing page + a customer interview where they describe their job. Find every place my page solves a DIFFERENT job than the customer's. Suggest a specific rewrite for each mismatch.
{paste}9. JTBD for B2B (buyer-job vs user-job)
For my B2B product, separate the buyer-job (why a company purchases) from the user-job (why an individual uses it daily). Inputs: {paste}. Output: both jobs, where they conflict, and how to address both in messaging without confusing either audience.10. Job-segments map
I think my product serves 1 job, but the customer base is heterogeneous. Below: customer descriptions. Identify whether there are 2-3 distinct jobs being served, and recommend whether to focus on one or segment the messaging.
{paste}11. Job-story drafter
Convert this JTBD into 5 specific job stories. Format: "When {situation + emotional context}, I want to {motivation}, so I can {outcome + meta-outcome}". No personas. Each story should suggest exactly 1 feature.
{paste JTBD}12. JTBD interview transcript analyzer
Below: a 60-min JTBD interview transcript. Extract: trigger, alternatives considered, decision criteria, the moment of "I have to find something now", and the expected outcome. Note any of the four forces the customer voiced. Output as a structured 1-page summary.
{paste}Job statement vs job story (the distinction teams blur)
These are not interchangeable, and using the wrong one is the most common JTBD failure.
| Job statement (Ulwick-style) | Job story (Klement/Wodtke) | |
|---|---|---|
| Form | Verb + object + context: minimize the time to file expenses | When [situation], I want to [motivation], so I can [outcome] |
| Carries emotion? | No — kept one-dimensional and measurable | Yes — situation and emotional context are the point |
| Persona? | No | No (a deliberate break from user stories) |
| Best used for | Quantifying underserved outcomes across a survey | Designing a specific feature or screen |
Ulwick’s argument against cramming everything into one job story is that it becomes impossible to later measure exactly where a customer is underserved. Klement’s counter is that situation and motivation are what make a feature feel obvious. You want both in your toolkit; prompt #2 produces statements, prompt #11 produces stories.
Common mistakes
- Job statements that read like feature descriptions (“user can filter…”) instead of describing progress
- No alternatives map — pretending “doing nothing” and “manual workaround” aren’t real competitors
- Mixing buyer-job and user-job for B2B and producing messaging that satisfies neither
- Skipping the switch trigger — JTBD without the trigger event is fortune-cookie copy
- Using JTBD as a brainstorm warmup instead of a feature-cut filter
- One job statement for a heterogeneous base instead of segmenting into 2-3 jobs
- Trusting a pattern after 3 interviews — run 10–15 switch interviews before you act on it
FAQ
How many JTBD interviews do I actually need?
Bob Moesta’s working rule is 10–15 switch interviews to surface the vast majority of actionable patterns in a customer base. Fewer than that and you are pattern-matching on noise; far more and you hit diminishing returns. Recruit recent switchers (people who bought or churned in the last 60–90 days) so the timeline is still fresh in memory.
What is the difference between a job statement and a job story?
A job statement names the progress (“minimize the time to file an expense report”) and is kept one-dimensional so it can be measured. A job story adds the situation and the emotional context (“When I land at midnight and just want to sleep, I want to file my expenses in two taps, so I can stop carrying paper receipts”). Use statements to prioritize, stories to design. Prompts #2 and #11 cover each.
What are the “four forces” in prompt #3?
The four forces of progress, from Moesta and Spiek’s switch model, are: the push of the current situation (frustrations with the status quo), the pull of the new solution (the imagined better future), the anxiety about switching (fears and uncertainties), and the inertia/habit that keeps people where they are. A switch happens only when push + pull beat anxiety + inertia. Map all four or you will misread why someone bought.
Can I run these prompts on the free tier of an AI tool?
Yes for short prompts, but the transcript prompts (#3, #8, #12) need a long context window. As of June 2026, Claude’s free tier gives limited Sonnet 4.6 access and Google AI’s free tier gives limited Gemini access — both fine for testing. For a full 60-minute transcript in one paste, a 1M-context model (Sonnet 4.6, Gemini 3.1 Pro) on a paid tier is the reliable choice.
Is JTBD a replacement for user personas?
The job-story school (Klement, Wodtke) argues yes — personas are imaginary customers defined by demographics that don’t explain causality, whereas a job centers on situation and motivation. In practice many teams keep lightweight personas for recruiting and use JTBD for the actual product decisions. If you want both, generate jobs first with these prompts, then sanity-check against your user persona prompts.