An investment decision has a characteristic that other moments in the year don't have: it forces a number. Not a direction, not an ambition, but an amount with an expected return behind it. To substantiate that number, someone climbs back to the assumptions on which the business plan rests. And there it becomes visible what could remain unnoticed all year: part of those assumptions concern work as it was organized last year, not work as it is organized now.
That is not a mistake by the person drafting it. It is the consequence of a shift that continues in the background in the meantime: AI is taking over tasks, not everywhere and not all at once, but task by task and in parts. A forecast built two years ago on a certain staffing level, a certain lead time, or a certain cost price per task, stands on a foundation that has since shifted somewhere. The decision now on the table is the moment at which that difference will cost money or generate money, depending on who sees it first.
As long as a strategy remains on paper, an outdated assumption can lie dormant. No one is held accountable for it, no money is put behind it. An investment decision changes that: it puts an amount on the table that depends on whether a task is still done by people, partly by AI with oversight, or largely by AI. Whoever invests in extra capacity for work that has by now largely been taken over, is paying for an assumption that no longer holds. Whoever, conversely, cuts back on a task that nevertheless remains human work, undermines a process that the decision was supposed to strengthen.
The question surfaces, then, not because AI is a new topic. It surfaces because this is the moment at which the cost of an outdated assumption becomes visible in a number instead of in a feeling.
In some organizations this is already established practice: every assumption in an investment proposal has a date on which it was last tested, and a short note on what AI could and could not take over of the task in question at that time. In other organizations that date does not exist. The assumption is there, it was once found to be correct, and no one has been given the task of reviewing it again.
The difference rarely lies in the quality of the people. It lies in whether a recurring moment has been set up at which someone explicitly checks: does this still hold? Companies where that moment is missing typically let assumptions stand until an event — an acquisition, a change of leadership, a shareholders' meeting — forces the issue. An investment decision is such an event, and that is exactly why this is the moment at which the question becomes unavoidable.
What is achievable at this moment is a picture of the assumptions themselves: which statements in the business plan implicitly assume a certain division of tasks between people and AI, and when that statement was last checked against practice. This is not a judgment on whether the decision should go ahead. It is an inventory of what the decision leans on.
In addition, it is possible to establish how the external picture relates to the internal one: what sector data, regulation, and visible steps taken by comparable parties show about the pace at which certain tasks are moving toward AI with oversight or full takeover, set against how the company's own management team assesses its own organization on that point. Differences between those two pictures are informative in themselves, even without an immediate explanation attached.
What cannot be established at this moment is a precise percentage of the work that will be done by AI in two years' time. That depends on decisions still to be made, on technology that is still changing, and on choices that cannot be derived from an external picture. Whoever demands a hard number for that is asking for a guarantee no one can give.
This test concerns the question of whether an assumption in a business plan still matches how work actually proceeds today. It does not concern the question of whether, and with whom, jobs are adjusted. That latter matter is a decision for the employer, with its own legal requirements it must meet, separate from what an assumption test yields. What is delivered here are facts about work: what AI can take over, what remains under oversight with human approval, and what remains human work. What is done with that is up to management.
The question that often lies behind an investment decision — which work in this specific company can genuinely be taken over by AI, and which work cannot — is answered task by task by the FTE TO AI work scan, separately from the decision itself. Anyone with broader doubts about whether the whole plan still holds up can also look at what needs to be done halfway through a multi-year plan or at what a strategic assumption actually is and how you check whether it still holds.
For an investment decision, it is not necessary to immediately revise the entire substantiation. A first step is a free assumption check: a short round in which you identify the key assumptions behind the proposal and see, for each assumption, when it was last confirmed. This does not produce a judgment on the decision, but it does provide a concrete overview of where the risk of obsolescence is greatest. The full proofing exercise, with the external picture set alongside management's self-assessment, is under construction.
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.