In many companies, the network in the market runs through one or a few people. The owner themselves, an account manager who has been there for twenty years, a salesperson who knows which customer wants to be called on Monday and which one only after lunch. That did not arise by chance. Relationships build up through repetition, and repetition lives in heads, not in systems. Whoever has talked to a customer the longest remembers the most, and thereby becomes the obvious point of contact. Nobody designed it this way; it is the outcome of years of contact without anyone recording exactly what is being remembered.
The result is that knowledge about the market — who buys and why, who is sensitive to which tone, which customer expects an annual round of calls — sits in one person and nowhere else. That worked for long enough, until a buyer comes to the table.
A buyer does not buy the number of people, but what continues without those people. For the network in the market, that question is sharp: does the revenue stay with the company, or does it leave with the person who maintained the contact. That is not a theoretical risk. Customer relationships that depend on one name are the reason acquisitions often include a retention period for the selling owner, or defer part of the price until it is proven that customers stay.
The price a buyer is willing to pay reflects that uncertainty. How the owner's succession readiness determines the valuation conversation describes exactly this mechanism: it is not today's revenue that counts, but the likelihood that this revenue will still be there tomorrow without the current key person.
AI is taking over, in parts, work that used to live only in a person's head. Not the conversation itself — the customer who has doubts about an order still wants, in most cases, to hear a human on the line. But the memory behind the conversation is changing. Contact history, preferences, previous complaints, seasonal patterns in orders: this can be recorded, made searchable, and summarized, so that a new account manager knows within a few conversations what an old colleague needed years to learn.
The three categories run alongside each other here. Summarizing contact history and flagging a customer who is ordering less than usual: a system can take that over. Assessing whether a customer will accept a price increase based on tone in a phone call: that currently still requires human judgment, with AI as preparation. And maintaining the relationship itself at moments that matter — a birthday, a setback, a won tender — remains human work, because the customer notices whether it is sincere.
Companies where this already works generally have their record-keeping in order: a CRM that is actually used, notes that go beyond "called, went well". Companies where it does not work often do have a system, but the knowledge still sits in the head of whoever had the conversation, because recording it takes time that nobody took. The difference is not in the availability of AI, but in the discipline of recording that precedes it.
Making it transferable does not mean the contact becomes impersonal. It means that the knowledge now sitting in one head also exists outside that head: who the customer is, what was going on, what worked and what did not. That is record-keeping work, not relationship work, and that distinction is precisely where AI can help without the customer noticing anything.
This automatically touches customer contact itself, and the question of which part of it can be supported and which part remains human work is addressed in how AI can support the first and second contact with customers. Also adjacent, but often underestimated: the network toward suppliers and subcontractors has the same dependency, and often runs through planning or costing, as shown in how one person's planning knowledge is made transferable and how price build-up and margin estimation depend less on one calculator.
If making customer relationships transferable coincides with a change in who does which work, the relevant statutory requirements around employment conditions and works council consultation apply; that is not part of this scan and requires its own advice.
A buyer who asks "what happens to the customers if you leave" is really asking a question about evidence: is there a system that carries the relationship, or only a person. Separate from the network, the legal underpinning of customer contracts is also relevant to that same question, and is described in which contractual matters must be in order before a buyer looks seriously.
The underlying question — which work in this company, also outside the network, can genuinely be taken over by AI — is answered task by task by the FTE TO AI work scan.
A good first step is not to map out the network, but to see which of the eight value drivers is putting the most pressure on the price today. For that, the free value check is available: eight short questions, one per driver, giving a picture of which driver currently weighs most heavily. The full value scan, with a maturity score per driver and a two-year calendar toward the exit moment, is under construction.