A buyer who asks who maintains the AI is in fact asking something else: will this work continue to be done once I take over the keys. An automated report, a model that checks invoices, a system that pre-sorts quotes — that only delivers value if it keeps running without the person who built or configured it. If that person is the owner themselves, or a single technical employee who is documented nowhere, then the buyer is not buying a process. They are buying a risk with a price tag on it.
This is not a new question. Buyers already asked it about ERP systems and about client relationships that depended on a single account manager. AI changes the scope of the question, not its nature. There are now more places in the business where something happens automatically, and therefore more places where the question "who maintains this" becomes relevant.
In some companies, AI use is the task of a single person, often the founder, who knows the prompts, recognises the exceptions and makes the adjustments when a supplier changes something. In other companies, that work is distributed: there is a process owner, there is documentation on what the system does and does not decide independently, and there is a second person who can take over. The difference rarely lies in the technology. It lies in whether the company has organised the maintenance as a task or has left it as incidental knowledge.
Three categories of work run through this. AI can take over some tasks entirely, without much maintenance being needed beyond periodic checks. Other tasks partly run through AI, with a human approving or rejecting and giving a reason — there, the maintenance lies in the quality of that oversight, not in the technology. And part remains human work, where AI at best provides support. A buyer who asks who maintains it actually wants to know in which of those three categories the bulk of the profit sits, and who holds the key in each category.
A maturity score per value driver is an estimate based on what is present: documentation, spread of knowledge, repeatability of a process without the intervention of a single person. That estimate is stronger the more evidence there is — a manual, a second person who has already run the process once, a log of exceptions. It is weaker where that evidence is lacking, and then the score mainly says that there is uncertainty, not that the situation is bad.
The outcome also says nothing about whether a saving through AI will continue to exist. Suppliers change terms, models get replaced, regulations shift. What the scan does show is whether the company is set up to deal with that change without a single person being indispensable. That is a different kind of statement than a guarantee, and it is not presented as one either.
The owner-dependency index works according to the same logic. It does not measure how many hours the owner works, but how much of the business stops if those hours fall away. High dependency in AI use counts more heavily than in an ordinary task, because a buyer has less experience taking over a system that no one else has configured than taking over a sales conversation that someone else can also conduct.
This page is not about who loses work or should lose work. AI taking over tasks means freed-up capacity, not a personnel decision — what an employer does with that falls under their own legal requirements and is up to the employer. The value scan measures what exists, not what should happen to the people working around it.
Nor is it tailored advice. The scan delivers scores, evidence and a timeline; not a recommendation to change something. What a management team does with that remains up to the management team.
The substantiation a buyer seeks differs per driver. Anyone wanting to know whether a margin is sustainable if it has partly been improved by AI will find the reasoning behind that in how you substantiate a margin improved by AI to a buyer. Anyone wondering whether the company leans on a few people who also carry the AI maintenance can read how to recognise that in how you prevent your value from sitting in a handful of heads. And anyone wanting to understand how the management team factors into the whole will find that in why a buyer looks at the management team.
The underlying question — which work in this company can genuinely be taken over by AI, and which part continues to require oversight — is answered per task with the work scan from FTE TO AI.
The free value check consists of eight short questions, one per value driver, and gives a picture of which driver is weighing most on your price today. The full value scan, with evidence per driver and a two-year timeline toward the exit moment, is under construction.