Research is useful when it helps a team make a decision. Before writing an interview guide, ask: what choice are we facing, what do we believe, and what would change our minds?
Turn a claim into an uncertainty
“Customers want an AI assistant” contains several assumptions. Do customers struggle with a task? How do they solve it today? Would automation improve that situation? A useful research question might be: how do first-time customers decide which plan fits their needs?
Ask about a real event
Instead of “Would you use this feature?”, try “Tell me about the last time you chose a plan.” Ask what triggered the task, which information they looked for, what they compared, and where they hesitated. Follow the story rather than steering it toward your preferred answer.
Make an interview guide
- Explain the purpose, how notes will be used, and that the participant can skip questions.
- Start with context: their role, situation, and recent experience.
- Ask them to walk through one actual example.
- Explore workarounds and consequences without suggesting a solution.
- End by asking what you missed.
Keep evidence and interpretation separate
A participant saying “I opened three tabs” is an observation. “The comparison page is insufficient” is an interpretation to investigate. Preserve the original context and record contradictory evidence. A small qualitative study can reveal mechanisms and problems; it does not establish how common a behavior is across every customer.
Bring the team into the learning
Share a few concrete observations, the decision they affect, and the remaining uncertainty. A researcher or designer should own the interpretation. AI can help draft an interview guide, but a generated response is not a customer interview.
Try it yourself
Rewrite three leading questions as neutral prompts about recent behavior. For each, identify the decision the answer could inform.