In the board meeting it usually sounds like this: "administration can be half the size with AI within a year." Everyone nods. No one means the same thing. The CFO thinks of invoices that process themselves. HR thinks of less manual entry for leave and absence. The COO thinks of hours freed up for something else, without knowing which hours those are. An ambition expressed this way is not a plan. It is a direction without layers: no distinction between the vision (why this matters), the vision state (what is then different), the target (the concrete number or the concrete boundary) and the target state (what actually works differently at that point). Without those layers, the board talks past each other while everyone appears to be right.
The shift is not "AI now does the administration." It sits in three categories that run through each other, even within a single administrative department. Part of the work can be taken over by AI without intervention: standardized entry, matching invoices with purchase orders, recurring reports with a fixed structure. A second part happens with human oversight: AI proposes a booking or a deviation, an employee approves or rejects it, with reason. That is not an intermediate step toward full takeover; for much administrative work it is the lasting form, because the consequences of an incorrect booking or a missed deviation are too great to let run without a look. The third part remains human work: exceptions, negotiations with suppliers, interpretation of a contract that does not fit the system.
Why does this already work in one company and not in another? Not because one company has a better AI vendor. The difference lies in readiness: has the organization structured its data so a system can work with it, are the approval rules explicit enough to hand over to an oversight role, and is there someone with the authority to actually implement a new division of work. That readiness is measurable, and it is usually missing in more than one place at once.
To know whether the ambition is achievable, you should not ask "do we want this" but "can we already do this." That is tested across eight dimensions: from data quality and system integration to decision rights, employee skills, and the extent to which processes are already documented rather than living in people's heads. Each dimension is assessed against one of five confidence gates: from "not yet examined" to "demonstrated in practice." A board that says administration will be halved while four of the eight dimensions are still at gate one is expressing a wish, not a goal.
This test is the same one used for other ambitions. Anyone who wants to make the ambition to improve margin concrete runs into comparable gaps: the wish is clear, the readiness is not. The pattern also repeats itself with growth and delivery time, which indicates that this is not a problem specific to administration, but of how ambitions are recorded before they are expressed.
As the work shifts, so do the roles that decide on it. Someone must be given the authority to definitively reject a deviation in the approval workflow without escalation. Someone else must be responsible for the data quality fed into the system, not as a side task but as a function. And there must be a role that periodically tests whether the human oversight step is still needed where it currently sits, or whether it can be moved. Without those roles, oversight remains an intention on a process diagram. What you record about this, including who may decide when things go wrong, is described in what you record about decision rights during a change.
If the shift affects headcount or job roles, separate legal requirements apply; this test concerns the readiness of the work, not decisions about employees.
The ambition has become concrete when you can point to three things: the number of hours of administrative capacity actually freed up, broken down into fully automatic and work under oversight; the place where oversight is still needed and why, rather than as a vague remainder; and a division of roles in which someone is accountable for each of the eight dimensions. If one of those three is missing, the freed-up capacity is an estimate, not a result.
The question underlying all of this is not rhetorical: which part of the administrative work in your company can truly be taken over, which part requires oversight, and which part remains human work. That question is answered per task with the work scan from FTE TO AI. Anyone who wants to sharpen the vision itself before looking at the tasks will find starting points in how you turn a vague vision into a concrete starting point.
The complete ambition test, with the eight dimensions and five confidence gates worked out per layer of the ambition, is under construction. Ahead of that, there is a free readiness check: eight short questions, one per dimension, which in a few minutes give a picture of where your organization is furthest along and where it is least far along. That picture is the starting point for a conversation in the boardroom that is no longer about words, but about what is already possible and what still needs to be done.
Vertel wat u wilt bereiken, dan kijken we samen wat daarvoor moet staan.
Answers come from this site’s knowledge base. Not tailored advice, and not a scan of your company.