How to Interview a Subject Matter Expert When Nobody Has 60 Minutes (and What to Actually Ask)

Stop trying to book the hour. The format that works with a busy expert is asynchronous and one question at a time: a queue of six to eight questions, each answered in whatever three-minute gap the expert has, saved the moment they move on, finished across a day or two instead of a calendar slot. Total expert time is roughly twenty minutes. And the questions have to be two-part, naming a specific situation and then asking what happened inside it. "Tell me about your differentiation" gets you marketing copy. "What does a prospect say when they push back on your price, and what moves them past it" gets you content.

We know this because we ran it on ourselves, and the transcript is still doing work in published posts.

Why does the 60-minute SME call never happen?

The 60-minute SME interview is not a bad format. It is an unbookable one. A founder at $500K to $2M ARR has a calendar built out of sales calls and hiring loops, and a one-hour "content interview" with no revenue attached to it slides a week at a time until it dies. Same for a new Head of Growth trying to pull knowledge out of the CEO: asking for an hour costs political capital they would rather spend elsewhere.

So the expertise stays in the expert's head. The writer fills the vacuum with the only thing available, which is generic industry knowledge, and you publish a post any competitor could have published. That is the actual cost. Answer engines have no reason to quote a page that says what forty other pages say.

What we built instead: a queue, not a meeting

Clark, our AI search specialist, runs what we call the AI Business Interview. It is deliberately not a call. It asks one adaptive question at a time about customer stories, differentiation, objections, methods, proof, and point of view, then writes a short internal knowledge summary that grounds drafts in first-party detail. Worth being precise about what it is for: it teaches Clark what the business knows, not how it sounds. Voice is captured separately, from work you have already published.

Three design decisions did most of the work, and you can copy all three with a shared doc and some discipline:

• Save on advance. Each answer is stored the instant the expert moves to the next question. Reopening the interview lands on the exact same unanswered question, so a founder interrupted mid-thought loses nothing.
• Skip is a real option. Skipping does not stall the interview, it asks a different angle on the same territory. Experts stonewall on questions they find boring, not on their whole domain.
• It reopens as an update. A finished interview does not close forever. It comes back with genuinely new questions informed by everything already answered, so the knowledge base deepens instead of going stale.

What questions should you actually ask?

These are close to the real ones Clark asked our founder, and each one is built the same way: a named situation, then a demand for a specific.

• Customer story: "When a new customer comes to you frustrated, what is the most common specific problem they describe, and can you walk me through a real example of what it looked like?"
• Objection: "When a prospect pushes back on the price, what is the specific objection, and what do you tell them that tends to move them past it?"
• Proof: "Walk me through one moment you can trace end to end: who it was, what happened, and roughly how long it took."
• Differentiation: "What can a customer not replicate by stitching together three tools they already have?"
• Method: "How does your process actually figure that out, and how is that different from how the standard tool does it?"
• Point of view: "What is the most common mistake you see people make before they find you, and what does it cost them in practice?"
• The closer: "What should I have asked and didn't?"

Notice what is missing: no warm-up, no "how did you get into this," no questions answerable with an adjective. When the expert only has twenty minutes, every question has to be capable of producing a paragraph you could publish nearly as written.

What came out of it that a generic AI could not write

The objection question produced a full price reframe: not a defense of the number, but the argument that the number should be compared to a content hire's monthly salary rather than to a rank checker.

The proof question produced Bosten Shoes. Rodrigo, its CEO and founder, had asked ChatGPT for the best leather shoe brand in El Salvador and watched it name competitors and never him. His answer carried the engine, the market, and the timeline in one breath: unmentioned, then recommended in that market, with a ChatGPT-attributed sale inside the first 30 days. No keyword tool surfaces that, and no writer invents it. It has anchored posts since, and it is the shape every proof story should take: a named engine, a named market, a timeline, and a sale at the end of it. That is the same loop we ask any customer to insist on, attribution that follows a lead from the AI answer into analytics instead of estimating dark traffic.

The mistake question produced the sharpest line in our library: founders who try this alone end up invisible where buyers are asking, and blind to the traffic they do earn.

Three questions, three pieces of content nobody else on the internet could have written. All of it came out of a queue, not a call.

The reusable takeaway

One twenty-minute async interview, asked in two-part questions and stored so it can be rotated across many posts, beats a 60-minute call you never book. Then structure what comes back so an engine can lift it: answer-first paragraphs that make sense pulled out of context.

If you would rather not run the interview by hand, that is exactly what Clark does before it writes anything, alongside tracking which buyer questions name your competitors instead of you. See how the loop works. Everyone shows you rank. We show you revenue.