Interview Yourself

Almost every prompt you write points the same direction. You know something, you want the machine to do something with it, so you type what you know and ask for output.

That works fine when you already understand the problem. It fails badly when you don't, which is most of the time on anything that matters, because the hard part of a real business decision is never the answer. It's that you don't know what you're missing.

So today we flip it. Seven prompts where the machine asks and you answer. No output at the end, or at least not the kind you're used to. What you get instead is your own thinking, dragged out of your head and laid on the table where you can look at it.

This is the single highest return thing I do with these tools and it's the thing I see almost nobody doing.

Here's the mechanic underneath it.

When you write a normal prompt, you're compressing everything you know about a situation into a paragraph. You leave out the things you assume, the things you forgot, and the things you don't realize are relevant. The model works with what you gave it and produces something confident and shallow, and the shallowness is your fault rather than the model's. You gave it a paragraph.

When you invert it, the model has to find out what it doesn't know. And because it's not you, it doesn't share your blind spots. It'll ask about the customer segment you stopped thinking about, the cost you're carrying but not counting, the assumption you made in 2023 that hasn't been true since 2024.

There's a second effect that's less obvious and more useful. Answering questions out loud is a fundamentally different cognitive act than writing a plan. You produce things you didn't know you thought. Anyone who has been coached properly knows this feeling. The coach didn't tell you anything. They asked six good questions and you talked yourself into the answer.

That's what these are. Coaching prompts, without the calendar or the retainer.

A note on format before the prompts. Every one of these works best in voice, if your tool supports it, or typed fast without editing if it doesn't. The value is in the unpolished answer. If you catch yourself writing carefully you've already lost most of the benefit, because careful writing is where you edit out the doubt, and the doubt is the useful part.

One. The assumption excavator.

"I'm going to describe my business and my current plan. Your job is not to evaluate it. Your job is to identify every assumption I'm making that I haven't stated, especially assumptions about my customers, my costs, and my market that I'm treating as permanent facts. List them as questions back to me, ordered by how badly I'd be hurt if the assumption turned out to be false. Ask me about the top three one at a time and wait for my answer before moving on."

The one at a time instruction is doing heavy lifting. A list of fifteen questions gets skimmed. A single question gets answered.

Two. The pricing interrogator.

"I want to work out whether my pricing is right. Do not suggest a price. Interview me instead. Ask me one question at a time about how I currently price, what my costs actually are including my own time, how my customers react, who I've lost on price, who bought without hesitating, and what my closest alternatives charge. After about ten questions, tell me what my answers reveal about my pricing that I didn't say directly."

That last sentence is the whole point. You're asking for the pattern in your own answers, which is the thing you cannot see from inside.

Three. The pre mortem.

"It's twelve months from now and the initiative I'm about to describe has failed completely. Before I tell you about it, ask me the questions you'd need answered to write a convincing autopsy. Then write the autopsy in my voice, as if I'm explaining to a friend exactly how it went wrong and what the early warning signs were that I ignored."

Writing it in your own voice matters. A third person risk assessment reads like a document. A first person account of your own failure reads like a warning, and you'll act on it differently.

Four. The customer you cannot see.

"Interview me about my ideal customer. Push past demographics and firmographics. Ask about what they were doing the week before they contacted me, what they tried first, who they had to convince internally, what they were afraid would happen, and what they told themselves to justify the spend. Keep asking until I stop being able to answer with specifics, then tell me where the gaps in my understanding are."

The moment you stop being able to answer with specifics is the finding. That's the part of your customer you've been inventing.

Five. The capacity audit.

"Interview me about how my week actually goes. Ask about what I do, what I've delegated, what I've tried to delegate and taken back, and what I'd have to stop doing to free up ten hours. Don't accept vague answers, push for specific activities and specific hours. At the end, tell me which of my constraints are real and which are ones I've chosen and could unchoose."

That final distinction is uncomfortable and worth it. Most constraints operators describe as fixed are preferences with good PR.

Six. The offer stress test.

"I'm going to describe an offer. Play a skeptical version of my ideal buyer who is interested but not convinced. Interview me the way that person would, one objection at a time, and don't move on until my answer would actually satisfy them. Be difficult but fair. At the end, tell me which of my answers were weakest."

This one is best done out loud, and it's genuinely unpleasant, which is how you know it's working. Better here than on a real call.

Seven. The thing you're avoiding.

"Ask me a series of questions designed to surface the decision I've been putting off. Start broad, get narrower based on my answers. Don't let me deflect into tactics. When you find it, ask me what I'd need to be true to decide this week."

Everybody has one. It's usually a person, a price, or a product line.

Want these as a working system rather than seven prompts?

The AI Workflow Blueprint includes all seven with the follow up chains, the version you run quarterly, and the automation that files each session's output somewhere you'll actually revisit instead of losing it in a chat history. Forty seven dollars. Reply with BLUEPRINT and I'll send it.

An operator I know ran the pricing interrogator in June. She'd been convinced for a year that she was underpriced and had been building up the nerve to raise everything fifteen percent.

The interview didn't tell her that. It asked, somewhere around question seven, who had bought without hesitating. She answered honestly: the clients who came through referral, every time, no negotiation. Then it asked who she'd lost on price. Also honestly: almost all of them from the paid channel.

She'd known both facts separately for a year. She had never put them in the same sentence, because they lived in different parts of her brain and nothing had forced them together. The finding wasn't that her pricing was wrong. It was that she had one price for two completely different buyers, and one of them was subsidizing her acquisition of the other.

She didn't raise prices. She changed where she spent, which was the actual problem and was not on her list of things to think about that quarter.

That's the shape of a good session. Not a revelation. A collision between two things you already knew.

Two honest caveats, because this is not a universal tool.

Don't use these when you already know the answer and are looking for permission. The model will give you permission, cheerfully, and you'll have laundered a decision you'd already made through a process that looks rigorous. If you catch yourself steering the answers, stop. You're not thinking, you're building a case.

And don't use these for anything with a real domain expert available and a real cost to being wrong. Tax structure, employment law, anything clinical. The interrogation format is excellent at surfacing what you think and completely uninterested in whether what you think is legal. Use it to prepare for the conversation with the professional, not instead of it.

Here's the failure mode, and it's the same one every time.

You run one of these. It's genuinely good. You have three real insights. You feel sharp and clear and slightly rattled in a productive way. Then you close the tab and eleven days later you couldn't tell me a single thing that came out of it.

The session is worthless if the output evaporates. So build the catch.

At the end of every session, one more prompt: "Summarize what I told you in three sections. First, what I said that I clearly believe. Second, what I said that I seemed uncertain about. Third, the two or three things I said that contradict each other or contradict something I said earlier. Be direct about the contradictions."

That third section is where the value concentrates. You'll say your best customers are the enterprise accounts and then twenty minutes later say the small accounts are the only ones who refer. Both are true. The tension between them is a strategy question you didn't know you had.

Then get it out of the chat window. A chat history is not a system of record and treating it as one is how good thinking gets lost. Push the summary into whatever you actually open. If you want that to happen without discipline, Make will take an output and file it into your notes, your drive, or your project tool on a trigger, and the difference between a system you maintain and one you don't is usually just whether the filing step requires you to remember.

If you want the thinking itself captured rather than just the summary, a tool like Littlebird will hold the context across sessions so you're not reintroducing your business every time. That's a real convenience and a real privacy tradeoff, and you should decide it deliberately rather than by clicking through a permission prompt. I use it selectively and off during anything sensitive.

Not seven. One.

Pick the one that made you slightly uncomfortable when you read it. That reaction is diagnostic. The prompt you don't want to run is the prompt with something behind it.

Give it twenty minutes. Answer fast and badly. Don't edit. Let it ask the follow up. Let it ask the second follow up, which is usually where it gets interesting, because the first answer is the one you've told people before and the second one is the one you actually think.

Then run the summary prompt and read the contradictions section twice.

You'll notice something odd about all of this. There's no output. Nothing to publish, nothing to ship, no asset in a folder. Just you, understanding your own business slightly better than you did this morning.

That is the highest return use of these tools available to you right now, and it's the one almost nobody runs, because it doesn't produce anything you can show anyone. Which is exactly why it's still an edge.

Ready to turn clearer thinking into a business that runs?

Knowing what to do is step one. Inside the AI Business Accelerator we take what these sessions surface and build it, the offer you stress tested, the follow up you kept avoiding, the capacity you found. Ninety seven dollars, for owners who'd rather build than deliberate. Reply with ACCELERATOR and we'll get moving.

Jordan

The AI Newsroom | Jordan Hale | ainewsroomdaily.com