Professional services largely consist of work that is easy to describe on paper: building case files, drafting reports, checking contracts, gathering and interpreting data, advising clients based on fixed frameworks. A large part of office hours goes into preparing a judgment, not into the judgment itself. That distinction is exactly where the shift is now taking place.
The outcome of that shift is not determined by the sector as a whole, but by three circumstances that differ per company: how structured the underlying data is, how repeatable the task is, and how much of the work ultimately comes down to a judgment for which someone bears responsibility. Where those three factors are favourable, work shifts faster. Where that is not the case, it stays with people, not out of caution but because the task requires it.
In professional services, AI takes over work in three forms that exist alongside one another. Some tasks are fully taken over: summarising documents, searching contracts for specific clauses, generating first drafts of reports. Other tasks shift to a form of oversight: AI delivers an analysis or a draft recommendation, an employee approves or rejects it and must be able to justify that. And a third category remains human work, particularly where it concerns the relationship with the client, judgments with legal or financial weight, or situations without precedent.
These three categories cut across functions and departments. Within a single role, part of the tasks may already have been taken over, while another part still rests entirely with the employee. That makes it difficult to reason about what AI does and doesn't do at the function level: the unit of analysis is the task, not the function.
Two firms with comparable services can be far apart on this point. The difference rarely lies in the technology itself, which is available to both. It lies in how the work is organised: is the data being worked with structured enough to automate, is there a fixed assessment framework within which AI can make a proposal, and is there a process in which an employee can approve or reject an AI outcome with a reason that holds up.
Companies where this already works have generally not waited for a complete plan. They have determined at task level what can be taken over and introduced that step by step, with oversight where needed. Companies where this does not yet work often do have a strategy that mentions AI, but that strategy is written at a level too far removed from the task to know what is actually changing.
The uncomfortable point is this: a strategy rarely states explicitly that human work will remain human work. That assumption sits beneath the plan, without anyone ever having written it down. If AI subsequently takes over part of that work, the assumption is not withdrawn. It simply remains in place, and the plan keeps resting on it while the ground underneath shifts.
That is different from a project that fails or a deadline that is missed. There is no moment at which someone formally withdraws the assumption, because there is no moment at which someone had formally established it. The assumption lapses in parts, in different departments, at different times, and usually this only becomes visible once the outcome has already shifted.
This dynamic is not unique to professional services. The same shift, with the same three categories, is at play in education, where testing and assessment are slowly changing shape, in retail, where inventory and customer data take on a different role and in hospitality, where scheduling and service are partly redistributed. The pattern is always the same: it is not about whether AI takes over work, but about which assumption no longer holds as a result.
This shift is not a reason to decide anything about personnel. Whether and how an organisation redistributes employees' work when tasks are taken over by AI is up to the employer, and separate legal requirements apply to that. What matters here is something else: whether the strategy is still based on a division of work that has in fact already changed. That is a question about assumptions, not about people.
Nor is it a question that can be answered with an estimate. How much capacity is actually freed up within a company depends on how the work there is organised, and that differs too much per organisation to be captured in a general percentage.
A strategy that is correct and a plan that is correct are not the same thing: the difference between them, and why a strategy needs an assumption that moves with the times, is described in the explanation of the difference between a strategy and a plan. For those wondering how a topic like AI stays structurally on the management agenda instead of being a one-off agenda item, there is an explanation of how to get regulation structurally on the management agenda, which follows the same logic for AI-driven shifts.
The underlying question of which work in a specific company can genuinely be taken over by AI cannot be answered at sector level. That question is answered per task with the work scan from FTE TO AI.
A management team that wants to know whether its strategy still rests on the right assumptions can start with a free assumption check: a short round in which you name your key assumptions and see, for each one, when it was last confirmed. The full strategic pressure test, with sector data, a self-plot of the management team and an ongoing assumption list, is currently in development.
Stel uw vraag. Vaak zit de echte vraag een laag dieper — daar mag ik naar vragen.
Answers come from this site’s knowledge base. Not tailored advice, and not a scan of your company.