How to start a 1-person business with AI (ask these 3 questions)

I spent almost a decade building companies nobody wanted. The fix wasn’t a better idea. It was a better question.

I started building companies at 17. I spent almost ten years failing at it.

My first company was a vacation rental site. My dad had a cottage, I wanted to learn some new technology, so I built the technology and called it a business. Then the guy who owned thecottage.com kicked my butt.

So I built a hosting company. Why? Because I was paying somebody else for hosting and figured I should own it. Web hosting in 1999 was like selling salt. Somebody out there is making a bazillion dollars selling salt. I wasn’t that guy.

Both of those companies died of the same thing. I started with “what do I want to build?”

Then I changed the question to “what problem are people already paying someone else to solve?” That question built Spheric Technologies. I went and found out who was already writing checks, and it turned out a whole bunch of companies were paying other companies to build enterprise portals, internal intranets. Following the money walked me straight to the pain. It found me customers. It found me partners, and my partners brought me into deals. At 24 I was selling to Fortune 500 companies. Four years later I sold the company and became a multi-millionaire.

Same person. Same skills. Different question.

Thirty years in, here’s what I’ve landed on: my whole life is the byproduct of the questions I ask. So before I start any one-person business, I ask AI three of them. We’ve launched 12 companies in the last 18 months and every one went through this.

Question 1: What should I actually sell?

Everybody starts an AI business the same way. They see AI do something cool, then they try to build a business around it. That’s backwards, and it’s the same mistake I made for seven years with different technology.

There are two kinds of businesses. Vitamins and painkillers.

Vitamins are nice to have. A stretching tool. Another productivity app. Insurance for a problem I’ve never had but maybe could. I don’t need them. There’s no acute pain.

Painkillers solve something that’s on fire right now. Help someone make more money, get more leads, keep their team, hire faster. When budgets get tight, the vitamins go first. The last thing anybody cuts is whatever makes them money.

A vitamin can make you some money. A painkiller gets you rich.

Build the frustration list. Most people struggle to find good ideas. My problem is saying no to great ones, and that’s a muscle, not a gift. Here’s how you build it. Tomorrow, as you go through your day, write down everything that frustrates you. Simple stuff. Why do I have to plug my phone in. Why am I opening and closing this garage door. Anything that makes you go “ugh.” Get as many down as you can, because what you’re really training is the ability to see opportunity where you used to see annoyance.

Then filter. People only pay for three things:

• Money. It makes them money or saves them money.
• Time. They get hours back. Doesn’t matter if they’re the CEO or entry level.
• Status. How does this make me look?

Miss all three and you built a vitamin.

Million-dollar businesses don’t get built on $10 problems. If you’ve got a migraine and I’ve got the pill, $100 is a bargain.

A hack most people miss: call the consulting companies already working with the companies you want to serve. Consultants only sell services there’s demand for. Their service menu is a list of problems somebody is already paying to solve.

That’s the manual version. Here’s the AI version. I use Claude, but any model works.

I want to start an AI-first business in [your industry]. That’s where I already spend my time and it’s where I want to build. Give me 10 observable and nuanced pain points that people in this industry deal with every single day. Not the obvious stuff everybody already knows, the specific ones you’d only catch if you were actually in the room. Only give me pains that are getting worse, not better, and rank them by how much people are willing to pay to have them solved. And ask me any questions you need to get clarity.

That last line is what makes it bulletproof. Ask AI a generic question and you get a generic answer, which turns into a very basic business.

Notice what I didn’t ask it: what should I build. That’s the question that killed my first two companies. I fell in love with my solution when I should have been falling in love with my customer’s problem.

Question 2: What’s my unfair advantage?

Finding a painkiller is the start. Question two is how you solve it in a way only you can.

When my brother started building homes, his idea was to build the cheapest house. Cheapest wins, right? People clearly wanted homes. He built the wrong one and it sat for nine months.

What he figured out is that his edge was working backwards from the customer. So he did the thing nobody in home building was doing at the time: actual market research. He grabbed potential buyers, walked them through dozens of open houses on a Sunday, and asked what they liked about each one. Then he took the best of those answers into a new design.

He didn’t go build it. He designed it, got renderings made, and pre-sold 16 homes with deposits before a shovel hit the ground.

Then he found a second edge inside his own operation. His builds averaged under 100 days, so he made it a promise: your home in 99 days or less. At his first home show, other builders walked past his booth and said “that’s not impressive, we do the same thing.” They did. They just weren’t willing to claim it.

He’s now one of the top home builders and real estate developers in Eastern Canada.

The hard part is you usually can’t see your own advantage. It’s easy for you, so you assume everybody can do it. That’s why it’s easier to get an outside source to tell you, which is where AI comes in.

Feed it your actual life first. Connect your Gmail, Slack, calendar, socials. My AI has everything: meetings, transcripts, trainings, all my financials, every decision I’ve made for 15 years. I slurped it all in so it knows me better than I know me.

Then, in the same chat where you landed on your painkiller:

Based on everything you know about me, where do I have an unfair advantage? What can I do better than 97% of people?

If it needs more context, make it interview you:

Now interview me about the last 10 years of my work. 10 questions, one at a time, then tell me what my unique edge is.

Pro tip: answer out loud. I use Wispr Flow, so it prompts me, I talk, it transcribes. Way faster than typing and you hand it far more context.

Then make it fight you, because a yes-man is useless here:

Be brutally honest. Is that actually rare or am I overrating myself? Where would someone who’s done this their whole career actually crush me?

Now cross-reference against question one:

Here’s the painkiller we picked and the edge we just found. Where do they overlap? Give me the one place where my edge is the reason this problem gets solved better than anybody else could solve it. If they don’t overlap, say so.

Pain alone is a market. Edge alone is a resume. Where they cross is the business, and it’s the reason somebody picks you over everyone else selling the same thing.

Mine is helping business owners use AI to scale so they don’t build a company they grow to hate. I know that pain because I lived it, and I know how hard the way out is to see when you’re standing in it.

Question 3: Will anyone actually pay?

Question one told you what to solve. Question two told you how you win. Neither one gets you paid.

People will tell you your idea is great. “You should build it. Come back when it’s done, I’ll be your first customer.” They’re lying to you. They don’t mean to. But ask them for money in that moment and the wallet stays shut.

When I started Flowtown, I wanted to know if people would actually pay, so I built a landing page and a full signup flow. Real screens, real credit card field. 20% of people put their card in. Then they’d get a message: unfortunately our servers are over capacity and we can’t take new customers right now.

There were no servers. There was no product. I call it the Wizard of Oz, the man behind the curtain. I simulated the transaction to confirm somebody would pay before I built a thing.

There’s nothing more expensive than building something nobody will pay for.

So the question isn’t whether somebody would pay. It’s whether they’d pay for it the way you’re solving it. And the sharpest version of that: how do I get one real payment in the next 7 days?

Here’s how I put AI on it.

Design the cheapest possible test.

Based on the painkiller and the edge we landed on, design the cheapest market test that gets me one real payment within the next 7 days. No website, no brand, no logo, no business registration.

Have it write the offer.

Write the one-page offer. A clear promise, a timeline, and a price. Nothing else.

Build the actual path to the money.

Now give me the 10 people I should contact first, the exact message to send, and the follow-up that asks for the money. Not a call, not a demo. Cash.

Then make it find the hole before the market does.

Why would people say they want this and never pay? What’s my riskiest assumption, and how do I test that one first?

Send that message today and take deposits from whoever leans in.

Do the first 10 by hand. I know AI makes you want to jump straight to 100, but scaling before you’ve heard what people say back is how you scale the wrong thing.

Most people go the other way. I’ve got to register the Instagram handle. I need the perfect logo, the business cards. All of it before a single person has said they want this. Certainty brings currency. You sell before you build, every single time.

The one thing that breaks all of this

You can run all three questions perfectly and still walk away with a garbage answer, because AI is world-class at being a people pleaser.

It will tell you you’re smart. It will never tell you your idea is dumb, it’ll tell you to go do it and then explain exactly how. That’s a trap, and you won’t catch it happening, because everybody wants AI to agree with them. Especially early on, when you’re fragile enough that one piece of negative feedback makes you quit.

So push back on it. Make it a ruthless mentor. For every answer it gives you, flip it back:

Now push back on me. Tell me why this fails. Where am I kidding myself?

In the military they call it red teaming. You attack your own plan based on what you think the enemy will do. Make it a habit, because everything you find on that side of the coin is something you get to fix before it costs you money.

Theory tells you what could work. A real person opening their wallet tells you what does.

The quality of my questions has dictated the quality of my life. So be more curious. A properly timed question, even one you just keep asking yourself, has your next level sitting on the other side of it.

-DM