The Billion-Dollar One-Person Business

This is the second post in Craft & Code, a short Friday series about what carpentry can teach us about AI, skill and the future of software. The first post set up the frame using my father’s career in the trade.

There is a story doing the rounds at the moment about the billion-dollar one-person business. The idea is that AI will soon make a single person so productive that they could build and run a company worth a fortune entirely on their own, with no team, no payroll, no org chart. It is a seductive picture, and like most seductive pictures it is half right.

The half that is right is the building. A skilled person with good AI tools can now produce, in days, software that would once have needed a small team for a quarter. That part is real, and I am not here to talk anyone out of it. The half that is wrong is the assumption hidden underneath: that because one person can now build the thing, one person can therefore run the business around it.

I am sceptical of that, and the reason I am sceptical is my father.

When my father went independent, he was already good at his trade. Nothing about his skill changed in the weeks it took him to make the move. What changed was that he finally owned the value of it. He could go out on his own because he had built, over years of doing the actual work, the judgement that made the work his: knowing what to quote, what would go wrong, what the customer actually wanted as opposed to what they had asked for, what could be made good and what had to be redone. That judgement was portable in a way that a tool or a job title is not. It was the thing he carried out of the door with him.

But here is the part the billion-dollar story skips over. Owning the craft did not make him a businessman.

He was a perfectly competent sole trader. He quoted, he got paid, he kept enough order to keep going. But he was the first to admit that the commercial side was not his craft. He managed. He did not master it. And had he ever tried to take the next step up — incorporate, run a limited company, handle VAT returns on a business with a constant flow of materials purchasing going through it — I do not think it would have gone well. Not because he was not clever enough. Because that was a different trade, and he had not served his time in it. He had spent his twenty years learning timber, not tax.

The thing AI is doing is lowering the cost of first execution dramatically for someone who already has judgement. That genuinely changes what one person can attempt. But a business is not one craft. It is a stack of them: the product, yes, but also quoting and pricing, accounts and tax, marketing, sales, support, compliance, the slow grind of maintenance, the legal edges nobody thinks about until they bite. Each of those is its own domain of judgement, earned the same slow way the technical judgement was earned. AI lowering the cost of the building does nothing to conjure the others into existence. It removes one constraint and leaves the rest standing.

The obvious answer is that you do not do those things yourself. You outsource them. Hire a bookkeeper, an accountant, a marketing agency, a contractor for the legal edges. AI plus a network of specialists, and the one person stays free to do the part they are good at.

Except that is not quite how it goes. The moment you outsource everything around the core, the core is no longer how you spend your days. You spend them briefing people, reviewing work, chasing things, managing contracts and deciding which supplier is worth keeping. The one-person building business quietly becomes a one-person coordination business — and to outsource well you still need enough judgement in each domain to tell good work from plausible work. My father could hand over his books, but he could not really tell whether they were done well, because that was exactly the craft he lacked. The fantasy promised a laptop and a great product. The reality hands you a procurement function.

The billion-dollar one-person company quietly assumes otherwise. It assumes judgement is general — that being excellent at the core craft makes you adequate at everything around it, and that AI fills whatever gaps remain. But judgement does not work like that. It is specific, and it is bounded. My father’s was deep and real and entirely his, and it stopped at the edge of the work he had actually done. Beyond that edge he was as much a beginner as anyone.

The honest picture of AI-enabled solo work is still an attractive one, just not the fantasy version. Not a billion-dollar empire run from a laptop, but a well-scoped, high-margin practice that one skilled person can own end to end, because AI has made the building cheap enough that they no longer have to hire their way out of it. The scope has to fit what one person’s judgement can actually hold. My father’s independence worked because a kitchen fitting business had a natural ceiling that matched what he could manage alone. He was not trying to build a joinery empire, and he was wise not to.

I am partly living this question myself. I have product ideas I am fairly confident I could build, and that I think would sell. The building, for the first time, is the part I am least worried about. But building the product and making a success of the business around it are two very different propositions, and the second is a far bigger undertaking than the first. Quitting a good job to own the whole stack of crafts a real business needs — to become the procurement function as well as the maker — is a challenge I have looked at clearly and decided against, at least for now. Not because I could not build the thing. Because when I do the honest arithmetic, the leap would most likely leave me worse off on both counts that matter to me: the money and the balance of my life.

That is where my father and I actually part company. He faced the same gap — being good at the work on one side, running everything around it on the other — and he jumped, because for him independence improved the arithmetic. He captured value that an employer had been taking. For me, in a different trade at a different time, the same jump runs the other way. The gap is the same. The maths is not. And the maths, not the building, is what decides it.

Which leaves the question I will pick up next week. My father could exercise that judgement, and own his work, because he had built it the slow way — through years of doing the lower-level work that the judgement was distilled from. The craft sat on top of the labour. It could not have existed without it.

So if AI takes over the lower-level work — the very work through which judgement has always been formed — where is the next person’s judgement supposed to come from?

Next week: The apprenticeship question. What happens to the pipeline when the beginner work is the first thing automated away?