Jobs-to-be-Done in the AI Era
There’s a moment in every product manager’s week, usually late and usually caffeinated, when you stare at a blank PRD and wonder: “What job am I really trying to get done here?” (pun intended). The Jobs-to-be-Done (JTBD) framework, that trusty lens for seeing past features and into the heart of user needs, is supposed to help. But let’s be honest: articulating clear, actionable jobs is hard. It’s even harder when you’re pressed for time, when customer interviews are non-existent, or when they left you with a tangle of insights and half-formed hypotheses.
Welcome to the AI era, where GenAI isn’t just a buzzword. It’s the secret weapon for every PM who has ever struggled to distill a JTBD hypothesis from the fog of research or the tyranny of the blank page.
From blank page to JTBD hypothesis in seconds
Remember the old days? You’d run interviews, fill notebooks, and then spend hours (or days) wrestling with sticky notes, trying to extract the “job” behind the user’s words. Sometimes you’d nail it. Sometimes you’d settle for “good enough” and move on, promising yourself you’d refine it later.
Now, with generative AI, that bottleneck is gone. Feed your interview transcripts, survey results, or even a few anecdotal pain points into a prompt, and in seconds you have a first-draft JTBD hypothesis. Not a generic one, but something tailored to your context, phrased in the language of your users, and ready to be debated, refined, or validated. Tested with GPT-4o: it will typically not only generate fairly good jobs, but also split them into functional, emotional and social, come up with a hierarchy, you name it. It does know the theory 🙂
The magic isn’t that AI “knows” your users better than you do. It’s that it accelerates the hardest part of the process: creating a structured hypothesis. And if you’re feeling stuck, a well-crafted prompt can nudge you out of analysis paralysis and into productive iteration.
No excuses: JTBD belongs in every PRD
Here’s the new reality: there’s no longer any excuse for skipping the JTBD section in your PRD. GenAI makes it so easy to generate, refine, and even challenge your job hypotheses that “I didn’t have time for enough interviews” just doesn’t cut it. In fact, prompting AI to articulate the job forces you, the PM, to clarify your own thinking. You have to decide what matters, what’s noise, and what really drives user value.
This isn’t about outsourcing product sense to a machine. It’s about using AI as a thinking partner: a way to see your assumptions in black and white, to spot gaps, and to make your product docs more rigorous and more actionable. I promise you people will understand what you want to do better through a prompt-generated JTBD than through prompt-generated solution requirements.
Warning!
GenAI is cool, but at this stage you still understand your customers better. Generating a hypothesis does not mean your JTBD is defined! It also tends to generalize across all customers, so if you don’t have interview data yet but do have an Ideal Customer Profile in mind, feed that to the prompt.
Prompting your way to a better PRD
Pro tip: formulate your JTBD first, and prompt-generate all your other PRD sections second. Giving the GenAI a human-validated JTBD as part of your instructions already points the further analysis in the right direction!
The bottom line
GenAI is the greatest help for defining your Jobs-to-be-Done hypothesis: before you talk to customers, after you talk to customers, whenever you need to distill chaos into clarity. In the AI era, the JTBD section of your PRD isn’t optional. It’s your north star, your sanity check, and your best shot at building something customers actually need.
So next time you open that PRD template, don’t stare at the blank page. Start with a JTBD prompt. Let AI help you do the job of defining the job. Your future self, and your users, will thank you.
