Independent of key figures while AI takes over work

The ambition as it is expressed

In the board meeting it sounds something like this: "We don't want to be dependent anymore on those few people who know everything." Everyone nods. And yet the financial director means something different by it than the operations director. One thinks of an employee who for twenty years has been the only one who knows the pricing agreements with the largest client. The other thinks of the IT specialist who alone knows how the invoicing system is put together. Both are key figures, but the knowledge that is locked in is not the same kind of knowledge, and so the way out of it is not the same either.

This ambition is now often placed alongside AI in a single sentence: "Can't AI just capture and take over that knowledge?" That is partly true, and it is exactly that word — partly — where the confusion arises. Some of a key figure's knowledge can be documented and transferred to a system that takes over the task. Other work of that key figure is a judgment call that someone continues to make, with AI doing preparatory work and a human approving or rejecting. And part of it — the relationship with that client, the feel for when an exception really can be made — remains human work, no matter how good the system becomes.

Where the difference comes from

Some organizations are already past the point where one person is the only one who can run a process. Others are right in the middle of it. The difference rarely lies in the technology itself. It lies in whether the knowledge of that key figure has ever been written down, tested, and made repeatable before an AI application is built on top of it. A system can only take over what can be described. A key figure who has never had to explain their own work is usually also the one for whom that description has not yet happened.

That is the core of readiness. Not: do we have the right software. But: can we split this person's work into what is transferable, what needs oversight, and what remains this person's own — and is there someone else who can exercise that oversight once the time comes.

What the organization needs to be able to do

The ambition assessment records this ambition in four layers: the vision ("less vulnerable to the absence or departure of key figures"), the vision state (what that concretely means per department), the target (which roles and processes are first in line) and the target state (how the work is then divided between system, oversight, and human). These four layers are assessed on eight readiness dimensions, with five confidence gates that indicate where the organization is confident enough to proceed and where it is not.

Concretely, this includes:

Level of documentation. Has the key figure's work ever been recorded, or does it only live in their head. Without documentation there is nothing to transfer to a system.
Oversight capacity. If part of the work moves to AI with human oversight, is there someone who can and is authorized to exercise that oversight, with the authority to reject.
Decision rights. Who will be allowed to decide on the exception that used to automatically land with the key figure. That right has to land somewhere, otherwise a new kind of dependency arises.
Margin for error and recovery capacity. What happens if the system or the replacement makes a wrong judgment, and how quickly can that be corrected.

This readiness is translated back into roles: who gets the capability to exercise oversight, who gets the decision right that used to implicitly sit with one person, and which role does not disappear but changes from executing to controlling.

Where this topic touches on personnel decisions — a role that changes, a position that is eliminated — separate legal requirements apply. This assessment concerns readiness of work, not decisions about people.

What you will notice a year from now

Not a percentage of knowledge transfer, because that figure does not exist without assumptions. But rather a few concrete signals. The key figure can go on vacation for two weeks without the process coming to a halt. There is someone else who can assess the exceptions, with a reason attached, not just a feeling. The freed-up capacity of that key figure has been deployed elsewhere, identifiably, not silently disappeared. And when a new employee joins, onboarding takes weeks, not months, because the work is documented rather than kept in one head.

This ambition shares many readiness questions with other ambitions. Anyone who wants to grow without hiring extra people as AI frees up capacity runs into the same documentation question. Anyone who wants to halve administration with AI support notices the same pattern: only described work is transferable work. And because this ambition often does not live only in the boardroom but also needs to be explained to the rest of the organization, it is worth knowing how a strategy avoids remaining only on paper.

The underlying question — which work of these key figures in this company can genuinely be taken over by AI, and which part remains human work — is answered per task with the work scan from FTE TO AI.

Where you can start now