If you have built any part of your business on an AI tool, what happens when the company behind it changes the rules?
They do change them. Prices move, terms move, the model you built everything around gets retired, and occasionally a provider decides your entire category of use is no longer welcome. Anyone building on top of somebody else’s infrastructure should be able to move.
Which is why there is some genuinely sensible advice circulating at the moment, and I want to agree with it before I complicate it.
The advice is this. Keep the material that makes your AI setup valuable somewhere you control. The instructions you have written, your process descriptions, your business logic, your customer information. Where you can, build it so that a different AI could slot in without starting again.
That is right. Follow it.
Then do the exercise it quietly assumes you have already done.
Open the folder
Go and find the material you control. Not conceptually. Literally. Open the folder and look at what is in there.
For a software business, this is a reassuring inventory. The instructions exist as files, because they had to. The process documentation exists, because a machine had to execute it and machines do not infer. The business logic is written down somewhere. You could hand the lot to a competent engineer and they could rebuild.
For a coach, therapist or consultant it tends to go differently.
There is a website. There is a curriculum outline, possibly three, in various states of currency. There are recordings of calls, which you have never listened back to and privately hope nobody else does either. There is a folder of slides that make complete sense to you and almost none to anybody else, because the slides were never the content. You were the content, and the slides were furniture.
And somewhere around the fourth minute of this exercise, a specific realisation tends to land. None of it contains the reasoning that produces the result.
What is actually being bought
Consider what a client pays for when they hire you.
Not the framework. Frameworks are on your website, and if the framework were sufficient they could have taken it from there.
Not the content, which they could get from a book, and possibly did, before deciding they needed a person.
They are paying for the judgement you exercise in the room. The moment where the model does not quite fit what is in front of you and you make a call. The point where a client says something entirely ordinary and you hear something underneath it, and decide whether to go there now or in three weeks. The decision not to use the obvious intervention because this particular person would experience it as an attack.
That is the product. It also appears in none of your documents, because you have never had a reason to write it down. You have simply done it, several thousand times, in real time, at speed.
Which produces an odd situation for a business to be in. The asset everything rests on has no existence outside one person’s head. It has never been audited. It cannot be handed to anybody. It is covered by no contract, no insurance policy and no intellectual property protection, because there is nothing there to attach them to. And it stops existing on the day that person stops working.
Provider risk is worth managing. It is a much smaller problem than owning nothing.
The objection, and why it does not survive
The reasonable objection is that this is simply what expertise is. Tacit, embodied, resistant to articulation. Every profession has it. You cannot write down how to ride a bicycle either, and nobody regards this as a crisis in cycling.
I have some sympathy with that, and I do not think it holds up against the evidence.
Enterprise sales has been trying to solve exactly this problem for around fifteen years. The shape of it is familiar. A small pocket of representatives wins somewhere between half and sixty per cent of the pipeline. Everybody else sits under twenty. Same training, same materials, same product, wildly different outcomes. The obvious move is to work out what the top group does and give it to everyone else.
That industry has advantages most professions can only look at enviously. It records the calls. It transcribes them. It logs every conversation, tracks every deal through every stage, and has budget. This is not a field that has been vague about its data.
Fifteen years of that, and the gap has not closed.
Which is genuinely useful, because it eliminates the explanation everybody reaches for first. The barrier was never insufficient recording. If volume of data were the answer, this would have been solved before smartphones existed.
What separates the person who wins is a judgement made in the moment, in response to a signal nobody wrote down and mostly nobody consciously registered. It leaves no trace in the artefacts. So ingesting the artefacts, however many of them there are, recovers none of it.
The same conclusion applies to a practice with a fraction of that data and the identical gap. Recording more of yourself will not do it.
What actually works
The material has to be brought out deliberately, and that is a different operation from collecting things.
It means somebody asking the questions that force the reasoning into words for the first time. Not what do you do, which produces the framework you already have. Why did you go there rather than the obvious place. What told you. What would have had to be different for you to make the other call. What do you do when your own model does not fit what is in front of you.
Practitioners find this uncomfortable, and the discomfort is diagnostic rather than a sign anything is going wrong. The answers are genuinely difficult to produce, because they have never been produced before. Twenty years of exercising a judgement is not the same as twenty years of explaining it. Nobody has ever asked, so nobody has ever had to.
What comes out is not a transcript and not a content library. It is an articulated method: the decision points, the conditions, the reasoning, the exceptions, in a form that exists outside your head for the first time.
At which point several things become possible that were not possible before. It can be taught properly rather than absorbed by osmosis over years of shadowing. It can be built into tools that behave like you rather than like a search over your marketing. It can be handed to somebody. It can be protected, because there is finally something there to protect.
And the honest reason to do it has very little to do with AI. AI is the current occasion, and it is a good one, because it has caused a lot of thoughtful people to open the folder and actually look. But what is in that folder predates every model on the market, and it would still be a problem if all of them disappeared tomorrow.
Most tools for practitioners start with what you have already written. The work worth doing starts with what has never been written down at all.
