Customer Discovery with AI: Interview Scripts and Synthesis

Numbers tell you how big a market might be. They cannot tell you whether real people feel the pain you think they feel. For that, you have to talk to them. Customer discovery, a handful of honest conversations with people in your target group, is the highest-signal validation you can do before building. AI cannot have these conversations for you, but it can make you dramatically better at running them: writing a script that does not lead the witness, finding where your audience hangs out, and turning messy notes into clear patterns.

What You'll Learn

• How to write a non-leading interview script with AI
• The questions that get honest answers versus polite ones
• How to find where your target customers actually gather
• How to synthesize interview notes into patterns with AI

The Trap: Leading Questions

The natural instinct is to describe your idea and ask "would you use this?" Almost everyone says yes to be nice, and that yes means nothing. Good discovery digs into the customer's past behavior and real problems, not their predictions about a product they have not seen. "Tell me about the last time you tried to solve this" is worth ten "would you buy it?" questions.

AI will write you a leading script if you let it, so instruct it not to:

I am validating a business idea: [one sentence]. Write a 20-minute customer interview script to learn whether this problem is real and painful. Do NOT pitch my solution and do NOT ask hypothetical "would you use this" questions. Focus on the customer's past behavior, what they currently do, what frustrates them, and what they have already tried or paid for. Open-ended questions only.

The classic approach behind this, popularized by the book The Mom Test, is simple: ask about their life, not your idea. The name comes from the idea that a good set of questions is one even your mom could not lie to you about, because you never give her the chance to just be supportive.

Questions That Get the Truth

A few patterns reliably separate real demand from politeness. Use AI to tailor them to your specific customer, but keep the spirit:

• "Walk me through the last time you dealt with [problem]." Past behavior is fact; future intent is fiction.
• "What do you use today, and what do you hate about it?" Reveals real alternatives and real frustration.
• "How much time or money does this cost you now?" If the answer is "none," the pain may be too small to sell against.
• "Have you ever looked for a solution? What happened?" Searching is a far stronger signal than nodding along.
• "Who else has this problem worse than you?" Often points you to a sharper customer segment.

Notice what is missing: any mention of your product. You are diagnosing, not selling.

Find Where Your Customers Already Gather

You do not need a big audience to start. You need five to ten of the right people. AI is a strong brainstorming partner for finding where they cluster:

My target customer is [specific description]. List the specific online and offline places where they already gather and talk about their problems: subreddits, forums, Facebook or Slack groups, professional associations, local meetups, newsletters, and events. For each, note what kind of conversation happens there and how an outsider could respectfully join.

Verify the suggestions, because AI may name communities that no longer exist or misjudge how active they are. Then show up as a human, not a marketer. Ask questions, be useful, and request short calls. People are surprisingly generous when you are genuinely curious about their problem and not selling anything yet.

Synthesize Notes Into Patterns

After five or six interviews you will have pages of messy notes and a fading memory of who said what. This is where AI shines, because the work is summarizing and pattern-finding, not inventing facts. Paste your notes (anonymized) and ask:

Here are my raw notes from [N] customer interviews. Identify the recurring themes: what problems came up repeatedly, what words and phrases customers used, what they currently do, and what they have paid for. Separate strong signals (mentioned by many, with emotion or money attached) from weak ones (mentioned once or only when prompted). Quote the customer's own language where you can.

Two cautions. First, AI summarizes what is in your notes, so if your notes are thin or biased, the summary will be too. Second, watch for the model smoothing over a contradiction to give you a tidy story. The contradictions are often the most useful part, so ask it explicitly to surface disagreements and outliers.

The phrase you are listening for is your customer describing the problem in their own words, with frustration or money behind it. That exact language becomes the headline of your value proposition in a later lesson, far more persuasive than anything you would have written yourself.

Key Takeaways

• A few honest conversations beat any survey or market report for early validation.
• Never ask "would you use this." Ask about past behavior and current frustrations.
• Use AI to write a non-leading script that diagnoses the problem instead of pitching.
• Find five to ten of the right people in communities they already gather in, and verify those communities are real.
• Use AI to cluster notes into strong versus weak signals, and ask it to surface contradictions.
• Capture customers' exact words; they become your future marketing copy.