AI can help you prepare questions for people who might use your product or service. It cannot tell you what real customers want by inventing their answers. A simulated interview is useful for practicing the conversation, not for proving demand.
If you are exploring a small business after years in another role, begin with a question you genuinely need answered. You may know the subject well while still needing to learn how someone else experiences the problem.
10 MINUTES · GUIDED PRACTICE
Turn two leading questions into a neutral conversation starter. Practice learning about a real experience without putting an answer in someone’s mouth.
Choose one learning goal
For an illustrative service that organizes paperwork for a local club, you might ask: “How does the person responsible currently prepare information for the committee, and where do they get stuck?” That gives the conversation a purpose without assuming your proposed service is needed.
Do not combine learning about someone’s experience with a hidden sales pitch. Explain who you are, why you want to talk and how you plan to use their answers. Make participation voluntary, and agree any recording or transcription arrangements before starting.
The GOV.UK Service Manual’s interview guidance recommends planning around research questions, testing the discussion structure and exploring participants’ actual experiences. The method informs this example; it does not make a few informal interviews representative of an entire market.
Ask AI to examine your questions
Fictional practice: a club volunteer is preparing committee paperwork. Your draft questions are: “Would our brilliant organizer fix your late paperwork?” and “Why do you waste time, and how much would you pay to stop?” Use the prompt below to review them, then choose two alternatives yourself.
I want to understand how a club volunteer prepares committee paperwork. Review these draft interview questions for assumptions, leading language and double questions. Suggest neutral alternatives focused on a recent real example. Do not answer for participants or claim what customers need. Keep questions about price or my proposed product separate from the experience questions.
Review the alternatives yourself. “Would an easy AI organizer save you hours?” assumes both a problem and a benefit. “Walk me through how you prepared the last committee pack” leaves room for the person’s own account.
Try the questions with a willing colleague first. Check whether they are understandable and whether the conversation fits the time you offered.
Use a short discussion outline
Here are original starter questions for the fictional club example. Adapt them to the real activity you are investigating:
- Think of the last committee meeting. What did you need to prepare?
- Where did the information come from, and what did you do with it?
- Which part, if any, was difficult or took more effort than expected?
- How did you handle missing information or a late change?
- What works well enough that you would want to keep it?
Follow the participant’s answers. If they describe a problem, ask for an example rather than immediately proposing your solution. If they have no significant difficulty, keep that finding instead of steering them toward agreement.
Do not ask for confidential committee papers, membership details or another person’s private circumstances just to make your research feel more complete.
Check your result
Which assumptions do those two draft questions introduce? Write two neutral questions before looking at the examples.
Compare with two neutral alternatives
“Walk me through how you prepared the most recent committee pack.” Then: “Which part, if any, was difficult?” A natural follow-up is to ask for a specific example.
The original wording assumes lateness, wasted time, a useful solution and willingness to pay. It also combines several questions. Keep the first conversation focused enough to hear the person’s own account.
What would count as evidence?
A real participant’s account, recorded and used with appropriate permission, can inform your next question. An AI-generated answer is a rehearsal, not a customer finding.
If someone says the current process works well, preserve that result. Do not turn a compliment, a simulated answer or an imagined quote into proof that people will buy.
You have finished this exercise when: you have two understandable questions that allow “no problem” as an answer, and you can separate an actual observation from your interpretation.
Get help with a stuck step · Choose another practice lesson
Next in this project: Explain the offer in one sentence.
See all three steps: Explore a service idea carefully.
Fictional practice example added October 10, 2026. Your wording can differ; check the facts and limits above.
Keep the evidence separate from your interpretation
After the conversation, distinguish what the person said, what you observed and what you think it means. A statement such as “I chase two missing updates most months” is different from your interpretation that they need an automated system.
If using AI to organize notes, use material the participant and your organization have allowed for that purpose. Remove unnecessary identifying information. Ask the assistant to preserve uncertainty and link each proposed theme to the relevant note.
Check any quotations against the original record. Never publish an invented quote, turn a paraphrase into a testimonial, or let a generated summary introduce a percentage improvement nobody measured.
Look for differences as well as patterns
One volunteer may struggle with late updates while another finds the existing process simple. Record both. A small convenience sample can reveal questions worth exploring, but it cannot establish market size or expected sales.
If you ask about your idea later, separate reactions to the explanation from evidence of actual use or willingness to pay. Compliments are not orders. Our small product prototype process can help you test the next assumption with a concrete example.
Decide what to investigate next
Write a brief finding: what you learned, supporting examples, contrary evidence and the question still open. You might revise the idea, speak with a different kind of user, or discover that there is no worthwhile problem to solve.
AI’s contribution is a clearer preparation and review process. The evidence comes from people and work you actually observed, with their permission.
Illustrative photo by dlxmedia.hu on Unsplash.
