There's a shop that's been in the same family for three generations. The owner does the books at night, answers the messages herself, remembers which regular likes what. She's been told AI will save the business. For twenty pounds a month, she now has what only the big chains used to afford.
It's true, and it's not enough.
The numbers are going the wrong way. Since 2022, large firms have lifted their real revenue per worker year after year. Small firms have watched theirs slip. The same models, the same prices, available to everyone. And the distance between the corner shop and the corporation is growing, not shrinking.
It isn't access. The gap in who has bought AI has nearly closed. Small firms are adopting it almost as fast as the giants. But buying isn't using. The OECD's read on small firms is blunt: three in four that have AI are still novices with it. A tool here, a tool there, none of it joined up. The subscription went through. The work never changed.
There's a reason, and it isn't laziness. The owner of a three-generation shop is the manager, the buyer, the bookkeeper and the only full-time member of staff. Half of small firms say their people lack the skills to use AI well. Nearly four in ten can't find the time to learn it. She doesn't have an afternoon free for an advanced course, and no one is coming to set it up for her. So she's been left behind. Not because the tool was out of reach, but because using it well is a job, and she already has six.
What the big firm has isn't a better model. It's someone whose job is to decide where the model belongs. Which work it does. Which work it must never touch. Where it hands back to a person, and how the pieces connect into one loop instead of twenty browser tabs. The corporation didn't buy an advantage. It organised one.
Buying the tool is adoption. Deciding what it's for is orchestration. Only one of them shows up in the revenue.
This is the part no subscription includes. A small business doesn't need twenty AI tools. It needs one connected line: the thing that remembers the customer, follows up without being told, flags the regular who's quietly drifting away, and drafts the reply the owner reads and sends in her own voice. The owner stays the judgement. The machine carries the repetition. And some things stay off the machine on purpose. The handwritten note. The call that should stay a call. Knowing which is which is the whole job.
That decision is not a budget. It's a point of view about what your business is for. And it's the one thing a family shop can settle faster than a company with nine committees, because the person who knows the work and the person who decides are the same person.
The firms pulling ahead aren't the ones with the most AI. They're the ones who decided what each part of their business is actually for, and let the machine in only where it earns its place. The corner shop and the clearing bank have the same problem under different lights. Someone has to decide where the machine belongs, and stand behind the line. That decision is the work I do. In the systems where getting it wrong is expensive, and in the ones where it's personal.