AI Customer Interview Prompts: 50+ Examples That Drive Real Insights (2026)

TL;DR — The best AI customer interview prompts are not single sentences. They are short systems: a role, a context, a clear goal, an instruction to probe, and a constraint that forces a specific answer format. This guide gives you 50+ prompts you can copy today, sorted by use case (discovery, JTBD, pricing, win/loss, churn, onboarding), plus the four prompt patterns that produce garbage and why Koji's AI moderator is built to never make those mistakes.

In 2026, the team that writes better interview prompts wins. Why? Because Griffin and Hauser's classic research — re-validated by Nielsen Norman Group — found that 20 to 30 well-conducted interviews surface 90 to 95 percent of a product's core customer needs. But "well-conducted" is the trap. A bad prompt produces a polite, vague, useless transcript. A great prompt produces specificity, contradictions, and the language customers actually use.

This is the prompt library we built after running thousands of AI-moderated interviews on Koji. Every prompt below is field-tested and probes for the why, not the what.

Why prompt quality decides interview quality

Per the State of Research Strategy 2025, 74 percent of researchers cite time as the primary factor when selecting a method. Teams ship faster when their interviews are tighter — and interviews are only as tight as the questions and follow-ups feeding them. Static survey logic forces every respondent down the same branch; an AI moderator with strong prompts adapts in real time, probes contradictions, and stops asking once it has the insight.

The structure we use in every Koji study:

1. Role — who is the AI being? ("You are a senior product researcher conducting a switch interview.")
2. Context — what does it know? ("The participant just churned from a $99/mo SaaS in the last 30 days.")
3. Objective — what is success? ("Identify the moment they decided to leave and the trigger event 24 hours prior.")
4. Probe rule — when to push? ("If they answer with a feature, ask which workflow that feature blocked.")
5. Stop condition — when to move on? ("Move on after two concrete examples or three follow-ups, whichever comes first.")

Drop that scaffolding into your favorite LLM or your AI voice interview platform and the difference is immediate.

Discovery interview prompts (10 examples)

Discovery is about understanding the problem space before you build. Avoid product questions entirely. Lead with behavior and context.

1. "Walk me through the last time you tried to {{job-to-be-done}}. Start from the moment you realized you needed to do it."
2. "What were you using before to solve this? Why did you stop?"
3. "Who did you talk to before you started looking for a solution? What did they say?"
4. "Describe the workaround you have today, step by step."
5. "What is the trigger that makes you go looking — what just happened in your day?"
6. "Tell me about a time this problem cost you money or time. How much, specifically?"
7. "If you could not solve this problem, what would happen next week? Next quarter?"
8. "Who else on your team feels this pain? What do they say about it?"
9. "Which part of this is the hardest, and which part is just annoying?"
10. "Show me the spreadsheet, doc, or tab you use today to manage this."

Pro tip: pair these with a JTBD switch interview when you want to understand the moment of change, not the steady state.

Pain-point and "five-whys" prompts (8 examples)

The single biggest interview mistake — covered by Nielsen Norman Group as the #1 reason interviews fail — is asking broad questions and accepting broad answers. Force specificity.

11. "You said it is frustrating. What specifically happens that makes it frustrating — describe the last incident."
12. "Walk me through that moment in slow motion. What were you doing 30 seconds before?"
13. "Why was that bad? And why was that bad? And why was that bad?" (Five whys, applied one layer at a time.)
14. "How often does this happen? Daily, weekly, once a quarter?"
15. "What did you do next after that frustration?"
16. "If a magic wand fixed this tomorrow, what would Monday look like compared to last Monday?"
17. "Which other tools or people did you escalate to when this broke?"
18. "How much money or time is on the line when this fails? Give me a number."

Jobs-to-Be-Done (JTBD) prompts (7 examples)

19. "Tell me about the first time you decided you needed something new. What was happening in your life that week?"
20. "What did you Google? What words did you actually type?"
21. "What were the two or three options you considered before picking the current solution?"
22. "What almost stopped you from switching? What pulled you forward?"
23. "What did you expect would change after you bought it? What actually changed?"
24. "Is there a moment you remember thinking 'this was worth it'?"
25. "If this product disappeared tomorrow, what would you hire instead — and how would you feel about that?"

Pricing and willingness-to-pay prompts (6 examples)

Pricing research collapses when you ask "would you pay $X" directly. People are terrible at predicting their own purchase behavior. Anchor in past behavior instead.

26. "What is the most you have ever paid for a tool that solved a problem like this? Walk me through that purchase."
27. "At what price would this feel like a steal? At what price would it feel insulting? At what price would you assume it is not serious?" (The classic Van Westendorp battery — see our pricing research guide for the full method.)
28. "If your budget for this category were cut by 50 percent, what would you drop first? Why?"
29. "Tell me about a tool you stopped paying for in the last 12 months. What made you cancel?"
30. "Who approves this kind of spend on your team? What do you have to show them to get a yes?"
31. "What is the cheapest tool in your stack that you would refuse to give up? Why is it indispensable?"

Win/loss interview prompts (8 examples)

Companies with clearly defined ICPs see up to 68 percent higher account win rates and 36 percent higher retention (Sybill 2026). Win/loss interviews are how you sharpen that ICP.

32. "Walk me through the evaluation. Who was in the room? What were you trying to decide?"
33. "What was the one thing that tipped the decision in our favor — or against us?"
34. "Which competitor did you take most seriously? What did they do that we did not?"
35. "If you had to do the evaluation again, what would you weight differently?"
36. "How did you feel about the sales process itself, separate from the product?"
37. "When you said no, what did the internal conversation sound like the next day?"
38. "What would have to be true 6 months from now for you to revisit this decision?"
39. "If a peer asked you about us tomorrow, what would you tell them — both the good and the bad?"

Churn and cancellation prompts (6 examples)

The hardest research interview to get right. Customers tell you "price" because it is socially safe. Probe past it.

40. "Walk me through the moment you decided to cancel. What were you doing? Who else was involved?"
41. "Was there a specific event in the 7 days before that pushed you over the edge?"
42. "If we had called you the week before, what would we have heard you say?"
43. "What did you replace us with — or did you just stop doing this entirely?"
44. "Which feature did you use most? Which did you never touch?"
45. "What would have to change for you to come back?"

For the deeper teardown of why "price" is rarely the real reason, see Why "Price" Is Never the Real Reason Customers Churn.

Onboarding and activation prompts (5 examples)

46. "Walk me through your first 10 minutes after signup. What did you click first?"
47. "At what point did you think 'okay, I get it'? What did you see right before that?"
48. "Where did you get stuck? What did you do — search docs, ask a teammate, give up?"
49. "What did you expect to happen on day one that did not happen?"
50. "Did you invite anyone else to the tool? Why or why not?"

Bonus: The 4 prompt patterns that ruin interviews

Avoid these. Every one of them will give you a polite, false, useless answer.

• Leading prompts. "Do you love the new dashboard?" Replace with: "Show me how you used the new dashboard yesterday."
• Hypothetical-future prompts. "Would you pay for this if we built it?" People are terrible forecasters of their own behavior. Replace with past-behavior prompts: "When did you last pay for something like this?"
• Compound prompts. "How do you find tasks, prioritize them, and decide what to ship?" Three questions, one answer, zero clarity. Split them.
• Solution-first prompts. "Would Feature X help you?" You are anchoring on your roadmap. Replace with the problem-first version: "Walk me through how you do this today."

For the deeper breakdown of how cognitive bias creeps into interview wording, see Cognitive Biases in User Interviews and the research bias guide.

Where AI moderators beat static prompts

A prompt is not a script. The reason teams move to AI-moderated voice interviews is that prompts only work if the interviewer adapts to the answer. Koji's AI moderator:

• Asks the right follow-up based on what the participant just said — not a pre-baked next question.
• Probes contradictions automatically ("Earlier you said X, but now you said Y — help me understand.").
• Stops when it has the insight rather than burning through a 30-question list.
• Uses six structured question types when you need quant alongside the qual: open_ended, scale, single_choice, multiple_choice, ranking, and yes_no — so a single study can output both narrative themes and clean distributions.

The result: same prompt library, but every interview turns into a tailored conversation, transcribed and analyzed automatically. No moderator bias, no scheduling, no recruiter overhead.

Run these prompts on Koji in under 24 hours

Drop any prompt from this guide into a Koji study, share a link, and have transcribed, thematically analyzed insights in your inbox the same day. No moderator schedule, no Zoom recordings to comb through, no $200-per-interview recruiter fees.

You get the AI consultant feature too — ask follow-up questions like "What did churned customers say about onboarding?" and Koji answers in seconds, citing the exact quotes.

Start a free Koji study →

Frequently asked questions