PM Mapped
Working with AI PRODUCT THINKING · Jobs-to-be-Done

Jobs-to-be-Done with AI

JTBD lives or dies on synthesis — turning what people said into the real progress they're trying to make. AI accelerates the messy middle: it can surface candidate jobs from raw interview language and reword them for neutrality. But it pattern-matches on words; it can't tell which job actually drives behaviour. That's still your call against the evidence.

← Back to the Jobs-to-be-Done tool
1How to use AI for this

A step-by-step way to work

1

Feed it real language

Paste interview quotes or notes verbatim. JTBD is about the customer's words and struggles, so give it the raw material, not your paraphrase.

2

Ask for functional, emotional, and social jobs

Have it separate the three layers — most PMs over-index on the functional job and miss the emotional one that often drives the decision.

3

Demand solution-free phrasing

A job statement should describe the progress wanted, not your feature. Ask it to rewrite anything that names a solution.

4

Generate alternative framings

Ask for several phrasings of the same job; the right one often reframes the problem space.

5

Verify against behaviour

Take the candidate jobs back to your evidence. A job is only real if it explains what people actually do, not just what they say.

2Worked examples

Real prompts, and what good output looks like

Surfacing jobs from an interview quote
Your prompt

From a user interview: “I bought the standing desk because everyone on my team has one now, and honestly I didn't want to look like the only person still hunched over. I do feel less sluggish after lunch though.” What functional, emotional, and social jobs are in here?

What good output looks like

Functional: avoid the afternoon energy slump / stay alert after lunch.

Emotional: feel less sluggish, feel in control of my energy.

Social: fit in with the team; avoid standing out as behind-the-times.

Note: the stated reason (health) is functional, but the social job ('not look like the only one') may be the stronger driver of the actual purchase.

The insight: AI flags that the social job may outweigh the functional one. That's a hypothesis — valuable, but you verify it, you don't bank it.
Stripping a solution out of a job statement
Your prompt

Rewrite this as a solution-free job statement: “Users want a dark mode toggle.”

What good output looks like

Job: “When I work at night, I want to use the app without straining my eyes, so I can keep going comfortably.”

'Dark mode' is one solution to that job — but so are warmer default colours, a brightness setting, or a scheduled dimming. Framing it as the job keeps your options open.

Why it matters: 'users want dark mode' closes the design space; the job opens it. AI is good at this reframe once you ask for it.
3Copy-paste template

A prompt you can reuse

Fill in the highlighted parts and paste it into your AI tool of choice. Edit the output — it's a starting point, not a finished answer.

Reusable prompt
Help me extract Jobs-to-be-Done from customer research. I'll paste raw interview language; don't paraphrase it away.

Research notes / quotes:
[paste verbatim quotes or notes]

Please:
1. Surface the candidate jobs, separated into functional, emotional, and social.
2. Phrase each as the progress the customer wants — no solutions or features named.
3. Flag which job seems to most drive the actual behaviour, and note it as a hypothesis to verify.
4. Point out anywhere the stated reason and the likely real reason differ.
4Common pitfalls

What AI gets wrong here

Mistaking restated wishes for jobs

AI may echo 'users want X feature' back as a job. A feature request is a solution, not a job.

Do this instead: Insist on solution-free phrasing focused on the progress wanted, then check that the job would survive even if your feature didn't exist.

Over-weighting what people say

AI works from the words in front of it, but people misreport their own motives.

Do this instead: Treat AI-surfaced jobs as hypotheses and validate them against actual behaviour and data.

Flattening to only functional jobs

The emotional and social jobs are easy to miss and often the real drivers.

Do this instead: Explicitly ask for all three layers, and probe the emotional/social ones in follow-up research.
The judgment that stays yours

AI is a fast first pass on JTBD synthesis, but it can only see the words you give it — it can't tell which job genuinely moves people to act. That comes from triangulating what people say with what they do, and from your judgment about which struggle is worth building around. The candidate jobs are hypotheses; you decide which are real.