In virtually every strategy there is, usually unspoken, an assumption about the nature of one's own work: this is too specialized, too context-dependent, too much bespoke work to fit into a standard process. That assumption did not come out of nowhere. It was built up from years of experience with software that followed rules and could therefore only do what had been foreseen in advance. Work that relied on judgment, on deviation, on reading a situation that was not in the manual, therefore remained human work. The conclusion became a fixed beam under the strategy: this part of our work is ours, and will remain ours.
As long as automation ran on rules, the assumption was accurate. A system that follows exactly what has been programmed cannot recognize an exception that has not been defined as an exception. Work with many variants, little repetition and judgment based on experience therefore remained out of reach of automation. The strategy built on this assumption held true, and the boards that maintained it were right. The problem is not that the assumption was ever wrong. The problem is that the condition on which it rested has changed.
AI that works with language, patterns and context does not need to know every variant in advance. It recognizes similarities with what it has already seen and applies that to a new situation. As a result, part of the work that was considered non-standardizable shifts into one of three categories: work that AI can handle independently, work in which AI makes a proposal and a human approves or rejects it with reasoning, and work that remains human work because the judgment relies too heavily on experience, responsibility or context that has not been documented. This shift does not happen uniformly everywhere. A company where assessments are well documented, where patterns in earlier cases are recognizable and where oversight of AI outcomes is in place, sees this part of the work shift faster than a company where that same judgment largely resides in the heads of employees. The difference lies not in the sector, but in how explicit the work had already been made before AI started reading along.
The assumption rarely expires with an announcement. It expires because a competitor, a supplier or a customer does something that, according to the old assumption, was not possible. A few signals that point to this: a provider in the market delivers a service at a pace or price that can only be explained if part of the judgment work has been taken over. Internal discussions about quality no longer center on whether something is correct, but on whether the oversight of the outcome is good enough. New employees are deployed for checking and correction rather than for the original task. And the question whether new entrants still need years to learn this work gets a different answer than it did five years ago. None of these signals proves anything on its own. Together, they are reason to re-test the assumption rather than let it stand simply because it has never been contradicted.
This shift also touches the question of exactly where the margin in a company lies. If the work that was considered non-standardizable becomes partly transferable, it also shifts where value is added in the process, something that connects to the assumption that the margin lies in execution. Anyone who treats this as merely an operational question misses that it is a strategic assumption due for revision.
This topic inevitably also touches on the question of what this means for staffing levels. This page does not address that. If a board is considering decisions that affect personnel, separate statutory requirements and a separate process apply to those; they fall outside the scope of what is discussed here.
An assumption about the standardizability of work is testable, provided you track the right things. Useful indicators include: the proportion of tasks within the work that has by now been documented as a decision rule or pattern, rather than as separate judgment; the pace at which comparable providers in the market are deploying tools resembling this work; the number of times oversight of an AI proposal results in rejection with reasoning, as an indication of how much judgment is still genuinely human work; and the most recent date on which this assumption was explicitly discussed within the board, rather than silently assumed. Anyone who cannot name any of these four does not know whether the assumption still holds or has simply never been challenged.
The underlying question of which work in your company can genuinely be taken over by AI cannot be answered with a general impression; the work scan from FTE TO AI answers that per task. For the board that first wants to know which assumptions are due for revision before a task-by-task picture emerges, there is the free assumption check: a short session in which you name your key assumptions and see, for each one, when it was last actually confirmed. Anyone wondering who should be at the table for a strategy review or noticing that the board disagrees internally about priorities will find a starting point there. The full strategic pressure test, combining sector data with the board's own self-assessment, is under 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.