AI customer discovery practice exercise for product managers

The scenario

Discovery goes wrong in a specific, common way when AI enters the picture: the tool ends up replacing the thinking instead of speeding it up. Someone asks an AI to guess what users want, skips the interview, and calls the output research. The job this exercise is built around is the opposite of that: show up to the interview sharper, synthesize what you hear faster, and keep every insight traceable back to something a real person actually said.

The exercise starts with something you already believe about your users, an assumption you have been treating as settled without really testing it. That belief is your raw material. From there you use AI at each stage where it genuinely helps: pressure-testing the assumption for vagueness and bias, drafting interview questions that probe behavior instead of opinion, and clustering notes or a transcript into themes once the conversation is done. What AI does not do at any point is stand in for the conversation itself.

What you are asked to do

You work through one complete loop, in order:

• Write down one belief you are currently carrying as true about your users, something that would change what you build next if it turned out to be wrong.
• Rewrite it as a testable hypothesis in a who, what, why format, and name the specific evidence that would change your mind, using AI to check the hypothesis for hidden bias or vagueness.
• Prompt an AI to draft a six question interview guide targeting that hypothesis, then edit it yourself to cut any leading questions and add behavioral probes that ask about a specific past instance rather than a general opinion.
• Take a real or sample transcript and prompt an AI to cluster what was said into three themes, each tied to a direct quote or observation, recording your confidence in each one.
• State one concrete decision you would now make, and say plainly whether the evidence supported, weakened, or killed your original hypothesis.

What a strong answer covers

The exercise's stated objectives are the standard a strong submission is checked against:

• Does the hypothesis name a specific who, what, and why, with a stated condition that would actually falsify it, rather than a restatement of the original assumption in fancier language?
• Does the interview guide ask about specific past behavior instead of general opinion, and has every leading question actually been removed, not just softened?
• Does every theme in the synthesis trace to a direct quote or observation, with an honest confidence rating, rather than reading as a tidy summary with no visible evidence underneath it?
• Does the final decision follow from the themes rather than restating the original hypothesis regardless of what the evidence showed?
• Is the whole loop traceable end to end, from the original belief through to the final call, so someone else could follow exactly where each conclusion came from?

Skills this exercise practises

Turning a vague belief into something you could actually be wrong about. Writing interview questions that get at what someone did, not what they would say they think. Synthesizing notes into themes without losing the evidence underneath them. Making a decision that follows from what you heard rather than what you expected to hear. These map directly onto the bootcamp's own curriculum: hypothesis-driven discovery, interview planning prompts, synthesis, and decision-ready outputs. For the underlying interviewing skill this exercise assumes, see how to run customer interviews and the discovery interview script template. If you want to see the same evidence-to-decision discipline applied after a launch instead of before one, the landing page conversion diagnosis exercise from From Idea to Launch with AI is a natural next step.

Which bootcamp this comes from

This exercise is the practical challenge from AI Prompting for Customer Discovery, a one week, two live session bootcamp on Builders Camp covering hypothesis-driven discovery, interview planning, recruiting and outreach prompts, synthesis, and opportunity mapping. Completing the practical challenge counts toward the bootcamp's completion requirement and its certificate, alongside the certification quiz.

Builders Camp runs this bootcamp both live and self-paced, included with the Builders Camp Membership alongside every other bootcamp, track, and masterclass. For a broader discovery grounding first, how to validate a startup idea covers the earlier steps this exercise assumes you already know.