Turn an AI customer persona into a testable product question

Start with the product decision that needs better customer understanding. Avoid asking a persona to comment on everything your team is building.

For example, your question might concern a planning screen that lets managers reassign work. The uncertainty is whether managers lack visibility, decision authority or confidence in the underlying information. Those explanations would lead to different investigations and potentially different product changes.

Write down which distinction you want to explore. The persona’s purpose is to help interrogate that distinction, not simply to supply a friendly voice in a meeting.

Give the modeled role a relevant geography, industry, company size and responsibility. Add the particular work situation you want to discuss. Specific context makes the conversation easier to interpret, but specificity alone does not make it true.

Keep three labels in your team’s notes:

• Known context: Information supported by your actual research or product knowledge, with its source.
• Working assumption: Something your team believes but has not established in this situation.
• Simulated perspective: An explanation or concern raised during the AI conversation.

Do not move a simulated statement into the first category because it sounds familiar. Look for the evidence that would justify the change.

Use five prompts to move from a description to an investigation:

1. Responsibility: What is this role accountable for in the situation?
2. Moment: When does that responsibility become difficult?
3. Constraint: What information, authority or dependency could prevent action?
4. Hypothesis: What explanation does your team want to investigate?
5. Observable check: What real example, behavior or artifact could support or challenge that explanation?

Your team completes the worksheet. It is a way to organize thinking, not an automatic product output or a scoring system. Keep the hypothesis narrow enough that an interview or walkthrough could change your mind.

Illustrative scenario, not a customer result.

A product manager is improving resource-planning software for mid-sized professional-services firms in the United Kingdom. The chosen perspective is an operations manager preparing the weekly allocation of work.

The starting profile says this person wants to save time. Instead of designing around that broad statement, the team explores a specific moment: an urgent project needs a specialist already assigned elsewhere.

A structured conversation can help the team consider different explanations. Perhaps the manager cannot see available capacity. Perhaps the information is visible, but another manager must approve the change. Perhaps nobody trusts how current the availability information is.

The team records these as competing hypotheses. It prepares a real-customer question: “Walk us through the last time you reassigned a specialist. What information did you use, and who else needed to agree?” That question can expose the actual sequence without assuming the problem is speed.

Ask about the circumstances behind a response. Explore what would make the role choose a different action, whose work might be affected and which constraint matters most in this scenario.

If the answers simply repeat the profile, change the question. Introduce a relevant trade-off, such as acting quickly with incomplete information versus waiting for confirmation. Record alternative explanations rather than collecting statements that support the initial design.

Labels such as “AI persona” or “virtual persona” describe many different things. Here, the useful object is a simulated B2B perspective supporting customer research. It is not a decorative biography, a real research participant or a prediction of a named person’s decisions.

Finish with a research question, a suitable participant and the evidence you hope to examine. Include a finding that would challenge your preferred explanation.

For the planning example, the next step could be a walkthrough using a recent allocation change and the records involved. If the difficulty is approval, that points toward a different investigation than missing visibility. If neither explanation fits, revise the profile and the question.

The value of the exercise is a clearer route to learning. You can retain the simulated conversation as a source of hypotheses while keeping observed customer work as a separate source of evidence.