Building a work schedule means combining staffing needs, availability and rest-time regulations into a published schedule. That task is fairly structured: there is fixed input (who is available when, who must be present when, which rest times are mandatory) and fixed output (a schedule). That makes it suitable for a scheduling system that operates as an agent: it generates the schedule itself, not just a suggestion.
The volume is high. Schedules are recreated weekly or monthly, for largely the same group of employees, with largely the same rules. This kind of repetition is exactly where automation pays off: the work a planner did last month must largely be redone this month, with new input but the same process.
Two axes keep this at "partly" rather than "yes". The first is judgment. A schedule is never just a calculation. A planner weighs who has had an inconvenient shift twice in a row, who has just submitted a request that hasn't been formally processed yet, or whether a team stays balanced in terms of experience. That is not a rule you can codify, it is an assessment that differs per situation.
The second is compliance. Rest times, maximum shifts per week and collective labor agreement terms are set out by law or contract, and a mistake there has consequences that go beyond an unfortunate schedule. A system can apply those rules as long as they are entered digitally and kept up to date, but responsibility for the outcome remains with a human who approves the schedule before it is published.
The third axis, volume, works in favor of automation: the more often a task recurs in similar form, the more a system learns from it and the less time manual checking costs per case.
A hospitality chain with fixed shifts, a limited number of roles and a collective labor agreement with few exceptions can leave most of the scheduling process to a system. The input is stable, the rules are unambiguous, and the exceptions are rare enough to handle manually. A healthcare institution with variable shifts, individual availability agreements and absence that continuously disrupts the schedule keeps much more with the planner. Not because the system is less capable there, but because the room for judgment per schedule is larger and the compliance risks are more sensitive, for example around minimum staffing per shift.
This difference, then, does not lie in the technology, it lies in how the work is organized at those two organizations. What exactly determines that is described on the page about why the same work is transferable at one company but not at another.
What is shifting today is not that schedules are suddenly "made by AI". It is that the first draft is increasingly generated by a system, and that the planner shifts from creator to reviewer: someone who assesses the draft, adjusts exceptions and only then publishes it. That shift is already visible at companies with a digital scheduling system and standardized shifts, and barely visible yet at companies that still keep schedules in a spreadsheet.
The task ends at a published schedule. What happens afterward belongs to adjacent tasks that each have their own profile. Changes that arise after publication, for example due to illness or a swap request, fall under processing schedule changes, with a different balance between system and human because the time pressure and exception rate are higher there. Calculating overtime that follows from the hours worked is a separate task described under calculating overtime. And because availability and absence form the most important input for a schedule, the quality of the output is directly linked to how leave requests are processed and how absence is registered. Poor input in one of those processes means a schedule that a planner still has to thoroughly review, no matter how good the system is.
This outcome is not a statement about the planner or team leader currently doing this work, and not an argument for shrinking a scheduling function. It is a description of a task: what a system can handle, what requires oversight and what remains human work. Decisions about personnel have their own legal requirements, and those are not addressed here.
Whether building schedules at your organization is closer to "largely transferable" or "largely human work" depends on how many exceptions your shifts involve, how your collective labor agreement is structured, and how digital your availability data already is. This differs greatly per company, even within the same sector. Similar considerations apply to other administrative functions, as shown in the overview of work in the IT department that AI can already take over today.
Anyone wanting a first indication without investing time in it can fill in the free quickscan: twelve questions, no account needed, resulting in an indication of what share of the hours in that job profile can be taken over by AI today. The full work scan, which breaks down the work of an entire company into tasks and calculates each task's impact on fte capacity, is still under construction and not currently available.
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.