A task that falls under human oversight is not automatically human work and not automatically AI work. It is the middle category: AI does the work, a human approves or rejects it, with reason. That may look like a compromise, but in practice it is the category where most of the shift takes place. Tasks that are fully taken over by AI were often already simple enough. Tasks that remain entirely human work are not changing for now. It is the middle section where companies today are actually making different choices than a year ago, and where two companies with identical tasks still arrive at a different answer.
Oversight does not change the task itself; it changes the standard against which you assess the task. Checking an invoice, answering a customer question, or drawing up a schedule: the work does not shift because the task becomes easier, but because the costs and risks of an incorrect outcome decrease as soon as a human takes the final step. That makes oversight one of the eight axes on which we assess a task in the work scan, alongside, for example, the degree of standardization and the sensitivity of the data.
Tasks that shift most into the oversight category are tasks with a clearly verifiable end result: a text, a calculation, a classification, an answer. AI produces a proposal, a human assesses it against a fixed standard. This works well for tasks that are already repeatable in themselves, but where the outcome previously required too much judgment to fully automate.
Tasks that do not shift are tasks without a verifiable standard. If nobody can say why an outcome is right or wrong, a supervisor also cannot approve or reject it with reason — that task remains human work, with or without AI in front of it. This is often the case, for example, when there is a lot of customer contact in the work and the standard shifts per customer, or when a company structurally delivers a lot of custom work where every outcome requires a different judgment call.
Suppose two companies perform exactly the same task, for example assessing purchase offers or handling a customer question. Yet at one company that task falls under full AI takeover, while at the other company it remains under strict oversight, or even remains human work. The difference does not lie in the task, it lies in the environment in which the task is carried out.
Three factors determine that difference the most:
None of these three factors is a judgment on the company. They are circumstances that make the oversight step more expensive or cheaper, and thereby determine whether a task in practice shifts from human work to oversight, or from oversight to full takeover.
The classification into eight axes and three categories is about work, not about people. We calculate in freed-up hours and FTE capacity: what a task requires in time, and what remains of that when AI does the largest part of the executing work and a human maintains oversight. What an employer subsequently does with that freed-up capacity falls outside this assessment. If a choice touches the employment relationship itself, then its own legal requirements apply, separate from what a work scan says about tasks.
To see how much of your own work falls into the oversight category versus full takeover or entirely human work, it is necessary to break the work itself down into tasks. The free quickscan takes a first, indicative step in that direction: twelve questions, without an account, with an indication of what part of the hours in your profile can be taken over by AI today. The full work scan, with the assessment per task on all eight axes and the conversion to FTE capacity, is still under construction. What is available today provides an initial direction; the precision with which you can see per task where oversight tips the balance comes with the work scan itself.
Vraag maar. Ik ken de kennisbank van deze site; wat ik niet weet, zeg ik erbij.
Answers come from this site’s knowledge base. Not tailored advice, and not a scan of your company.