Largely, but not without agreements made in advance. Scheduling, rescheduling and resolving conflicts in the calendar of one or more people is a task that lends itself well to an AI agent that works independently within rules that you set, with human oversight at the moments that require it. That is a different answer than "yes, fully automatic", and also different from "no, this is human work". It sits in between, and exactly where depends on how your organisation arranges meetings and priorities.
Three axes determine the outcome here: judgment scope, volume and structuredness.
The volume is high. A management assistant who manages the calendars of several executives processes dozens of requests on an average day: new appointments, shifts, double bookings that need to be resolved. That is exactly the kind of repetitive work where an AI agent frees up time, because the rules behind good scheduling can largely be set down: who takes precedence, which blocks are sacred, how long travel time is between locations.
Structuredness scores average, a 3. A calendar is a structured system with times, participants and locations, but the input around it, an email saying "could it also be Thursday, but not before eleven", is unstructured text that first needs to be translated into a concrete proposal. Modern language models can make that translation, which is what makes this different from ten years ago.
Judgment scope is the pivot point, also a 3. For a simple appointment between two available people, there is little to judge. For a conflict between a recurring team meeting and a request from a client, it is different: who yields, which appointment matters more to the organisation, is there political sensitivity around who gets moved. That judgment can be automated as long as the prioritisation rules are made explicit. Without those rules, the decision falls back to a human.
Physical work plays no role, this is entirely screen-based work, which explains the score of 5. Compliance also scores 5: there are rarely legal obligations attached to scheduling an appointment, so that obstacle is absent.
Creativity scores low, a 2, and that is precisely why this task lends itself well to automation: no new idea is needed, only the correct application of rules. Cost of error scores a 3: a wrongly scheduled internal appointment can be fixed with a phone call, but a mistake involving a major client appointment or a board meeting costs more. Customer contact also scores a 3, meaning that part of the appointments is purely internal and part external, with a different tone and different sensitivity.
Suppose a director has a fixed meeting every Tuesday morning, and a client asks for a conversation that only fits that morning. An AI agent with clear rules ("client conversations from this segment take precedence over internal meetings, except for the monthly board meeting") can resolve this independently: moving the meeting, informing those involved, proposing a new time. Without that rule, the agent gets stuck on a choice that is actually a political one, and then a human is needed to decide.
At another company, where the director wants to decide case by case who gets precedence, the balance is different. Then the AI agent is a proposal generator that offers options, not an agent that shifts things independently. The same applies to VIP appointments: a conversation with a major shareholder or a board-level job interview more often requires a human look before something is finalised or moved.
At companies with a tightly defined prioritisation logic and a culture of digital calendars, agenda management today already largely runs through an agent, with a human only intervening in exceptions. At companies where appointments run through loose chat apps and verbal commitments, that is still far off, not because the technology cannot do it, but because the input is too messy to automate.
This task, incidentally, is not separate from the rest of secretarial work. Anyone looking at agenda management often encounters adjacent tasks: reserving meeting rooms and facilities, receiving and registering visitors, and drafting standard letters and forms. These are separate tasks with their own scores, but they often share the same systems and the same role, which means a shift in one often brings a shift in another. Anyone wanting to look beyond just the calendar can also examine how the progress of goals and projects is monitored, because agenda management often touches on that.
This is not personnel advice and not an argument for a decision about anyone's position. What an organisation does with freed-up hours is up to the organisation itself; its own legal requirements apply there. What is stated here is an assessment of the task, not of the person currently performing it.
The scores above are also not exact percentages but a direction; how you should read an indication such as "largely", and what a bandwidth does and does not say, is explained in this explanation of bandwidths.
To see how this applies to your own situation, with your own mix of internal and external appointments, your own prioritisation rules and your own systems, a short assessment is worth more than a general estimate. The free quickscan from FTE TO AI consists of twelve questions, no account required, and provides an indication of what portion of the hours in this profile can be taken over by AI today. The full work scan, which calculates the work of an entire company task by task into fte capacity, is still under construction.
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.