I was in Denver this week for the Squad Up Summit, a real estate conference hosted by Pace Morby. One of the sessions was from Vic Keller — entrepreneur, 17 companies founded, 9 exits, 3 of them to Warren Buffett.
He didn't lead with that. He led with a slide.
Two columns. Old world on the left, new world on the right.
Old world: $13.18 million a year. 47 to 65 people. Business hours only. High turnover. Severance. Sick days.
New world: $470,040 a year. Three people. 24 hours a day, 365 days a year. Including $20,040 in subscriptions and licenses. The rest is lean, focused labor.
The room did the math on their own payroll. You could see it.
Then he made a separate point that landed even harder. You do not need to build the full company of one overnight. For roughly $1,000 a month you can have max or pro accounts across every major AI platform available today. That is the intelligence layer. The cognitive work that used to justify 50 people on payroll. Available now, to anyone, for the cost of a car payment.
The floor dropped out of the room a second time.
Then Keller asked a simple question: who in this room is actually using these tools in their business today?
2,900 entrepreneurs and operators. People who had just watched a presentation showing them how to run a company on three people instead of sixty. And then learned the AI capability layer that makes it possible costs $1,000 a month.
Maybe 20 hands went up.
What that moment actually means
This wasn't a room full of skeptics. These are people who build things, buy businesses, take risk for a living. They understood what they were looking at. The math wasn't complicated. The tools aren't expensive. The information has been public for two years.
And 98 percent of the room hadn't done it.
The gap isn't access. It isn't cost. It isn't even awareness anymore.
The gap is the distance between knowing this exists and knowing how to make it work for your specific business.
That is a different problem than most people think they are solving when they talk about AI adoption.
People matter more, not less
Keller was not saying people don't matter. He was saying the opposite.
When you go from 65 people to three, the three you choose become everything. His framework for who makes the cut: Character, Competency, Chemistry. In that order.
At 65 people you can absorb a bad hire. At three you cannot. The AI handles execution. The people have to handle judgment. Which means the bar goes up, not down.
He also put up a slide that most people probably moved past too quickly. The layers of a durable business. Seven of them. Systems and Processes was layer three. Talent and Leadership was layer four.
Most businesses do it backwards. They hire and then try to build systems around whoever they hired. Keller is saying you have to know what your business needs to execute before you can evaluate whether someone is the right fit to execute it.
In a three-person company running on AI infrastructure, that order is not a preference. It is survival.
The thing nobody in that room was talking about
Keller also put up a stat I have not stopped thinking about. 10,000 baby boomers retire every single day. $84 trillion in wealth transfers over the next 20 years. 70 percent of small businesses have no succession plan.
He framed it as an ownership opportunity. Go buy the distressed ones before they dissolve.
I heard something else.
70 percent of those businesses are running on unwritten knowledge. Processes that live in the owner's head. Pricing logic based on 30 years of pattern recognition that exists nowhere except one person's gut. Customer relationships built on instincts that have never been named.
When that owner retires, the business does not transfer cleanly. It partially evaporates. The buyer discounts it. The employees inherit ambiguity.
The most valuable thing AI can do for that business is not automation. It is capture. Extract the knowledge before it disappears.
That is not a future problem. It is happening now, at a scale the succession planning world is not remotely equipped to handle.
The company of one model is real. The time compression numbers Keller showed are real. But none of it works if the company of one is just a person with subscriptions.
It works when the subscriptions know what the business knows. When the systems have been given enough of the right instructions that they produce the right output without someone babysitting every step.
Getting there requires doing the work that most people skip. Not installing the tools. Specifying the judgment.
That is why 20 hands went up and not 2,900.
If the succession angle hit close to home, I am also building something specifically for that problem. It is called Pass The Keys — the idea is simple: extract what the owner knows before they walk out the door, and encode it into something the business can actually use. Still early. If it sounds relevant to your situation or someone you know, reply and I will tell you more.
If you are one of the 2,880 and you want to figure out where to start, the AI Readiness Audit is a good place. Two minutes, personalized assessment back.
Or just reply to this email. I am doing a limited number of diagnostic conversations this month with operators who are serious about closing the gap. No pitch. Just the conversation.
See you next Tuesday.
Nick
Built, Not Bought is a weekly newsletter about building AI systems that run your business. Subscribe on LinkedIn or Beehiiv so you don't miss an issue.