A buyer asks one thing sooner or later: what happens if the owner is unavailable for three months. At many companies the answer is uncomfortable. Not because there are no people to take over the work, but because the decisions that provide direction — which customer gets priority, which quote gets adjusted, which risk is acceptable — sit in the head of one person. That head has no document, no rule, only experience that has never been written down.
That dependency did not arise from unwillingness. It arose because making decisions was faster than recording decisions, and because every exception had a reason that was clear at the moment itself and never again after that. A company that is ten years old has had thousands of those moments. Virtually none of them has been noted down.
Decision-making is not a single block. It consists of steps, and those steps are not all affected in the same way by what AI can do today.
Part of decision-making is in fact pattern recognition: given this combination of factors, what did the owner do the previous ten times. That part can be derived from historical data — quotes, emails, approvals — and an AI system can already surface those patterns today and present them as a proposal. That is not a vision of the future, it is already happening at companies that have their decision data in order.
A second part is decision-making with oversight: the AI proposes an outcome based on recorded rules, but a human approves or rejects it, with a reason. This is the layer that yields the most for transferability, because the reason given upon rejection is precisely what has never been written down anywhere until now. Every rejection thus becomes a rule that the next successor no longer needs to get from the owner.
A third part remains human work, and that is a larger part than managers often think: decisions that depend on a relationship, an implicit agreement from the past, or a risk assessment that is not in the data but in the feeling of someone who has known the customer for twenty years. That part cannot be automated by labelling it as a process. It can be reduced, however, by making it explicit the moment it occurs.
The difference between companies does not lie in the technology — that is comparably accessible to everyone — but in whether decision data already exists to build on. Companies that have structured their quotes, price exceptions and customer correspondence can already make those first two layers visible today. Companies that have everything in people's heads and scattered mailboxes must take that step first before AI can do anything with it.
A buyer does not value the revenue that exists, but the revenue that keeps flowing without the current owner. Decision-making that resides only in that head is, for a buyer, a risk that gets priced in, not a quality that gets rewarded. As more of the repeatable decision layer is recorded — as a rule, as a proposal with oversight, as a documented exception — that risk decreases and the rest of the company becomes easier to assess on its own merits: profit quality, customer spread, growth potential. Dependence on one person carries through to practically every other driver; it is rarely an isolated problem.
Making things transferable does not begin with removing the owner from the decision, but with making visible what exactly determines that decision. That can be done in steps that do not affect day-to-day relationships: recording which deviations from standard terms have been approved and why, which customers have exception status and on the basis of which agreement, which risks have been consciously accepted. None of that changes how the owner deals with customers. It only changes whether that knowledge becomes a document or remains a memory.
After that, part of that recorded logic can be given to a system as a proposal, with a human — not necessarily the owner — approving or rejecting it. This gradually shifts the dependency from a person to a process with oversight, which for a buyer is a fundamentally different risk profile than a process that only exists as long as that one person stays on.
In practice this also affects other drivers: how customer relationships become transferable without the owner remaining the only point of contact, how knowledge of supplier terms and relationships is recorded, and how technical expertise that sits with only one person becomes accessible to others. Anyone considering personnel consequences when redistributing decision-making roles: separate statutory requirements apply to that, independent of what is described here.
Recording decision logic does not guarantee that tasks get taken over, and not every decision can be captured in a rule. What it does deliver is insight: which part of decision-making is already traceable from data, which part can be transferred with oversight, and which part will remain with a person for the time being. That breakdown — and which work at this specific company can actually be taken over by AI — is mapped out per task with the work scan from FTE TO AI.
To see whether owner dependency is the driver weighing most heavily on your price today, or whether something else — management information that is not timely or not reliable, margin, customer concentration — weighs more heavily, the free value check offers an initial picture: eight short questions, one per value driver. Anyone wanting a broader understanding of exactly what is weighing on the price will find the background in an overview of the factors that depress the value of a company. The full value scan, with maturity scores per driver and a two-year calendar toward the exit moment, is under construction.