An educational institution does not sell a product with a fixed specification. Parents, students and clients choose based on a mix of scheduling certainty, guidance intensity, accountability toward the inspectorate or funder, and the ability to track individual progress without it getting bogged down in paperwork. The hours are largely made up of three types of work: standing in front of a class or group, the administrative accountability behind it (testing, record-keeping, reporting), and the coordination between teachers, support staff and management. That third category often grows the fastest without anyone having decided so: schedules, exceptions, leave requests, communication with parents or clients, gathering data for an audit or accreditation.
The circumstances that drive the outcome are not the same everywhere. An institution with a homogeneous target group and a stable curriculum has a different distribution of burden than an institution with many tailored programs, fluctuating intake or a multilingual population. Where the administrative burden is high, time that could have gone into guidance is absorbed by accountability instead. That difference partly determines what an institution can win on: on teacher availability, on the speed of individual feedback, or on the reliability of the accountability itself.
AI is partly taking over work that used to simply cost hours. Schedule building with complex constraints can be calculated by a system instead of a planner testing it manually. First drafts of report texts, test analyses or progress overviews can be drawn up by AI, with a teacher approving, adjusting or rejecting them. Intake forms, enrollment flows and standard communication toward parents or clients can largely run automatically. What remains human work is the pedagogical conversation, the assessment of borderline cases, and the responsibility for the final judgment about a pupil or student.
This shift directly affects the comparison between institutions. As long as schedule building and accountability were mainly staffing work, the institution with more administrative personnel or a stricter planning culture won. If an institution has most of those tasks taken over, what wins there changes: no longer who deployed the most capacity, but who actually spends the freed-up hours on guidance, contact moments or individual attention. Two institutions with the same AI application can therefore still score differently, because one puts the freed-up time back into the classroom and the other into new layers of reporting.
This does not happen everywhere at the same pace. Institutions with a standardized curriculum and clear assessment criteria lend themselves more easily to automated support than institutions with a lot of individual tailoring, where every file requires its own considerations. The degree of oversight also differs: where the board treats AI output as a draft that a teacher must approve, a different quality assurance arises than where output flows through unchecked. For decisions that affect personnel, that consideration is subject to its own legal requirements; that is up to the institution and its competent authority, not a comparison of work processes.
The dimensions that used to be mainly about volume — how many teachers, how much administrative staff, how many locations — are shifting toward dimensions about allocation and accountability. Does an institution win on responsiveness toward parents, on the quality of individual progress reporting, on the reliability of its test data, or on the visibility of its accountability toward the inspectorate or funder? Every institution communicates a claim about one of these points, but not every claim can be substantiated with evidence. How you can check this for your own proposition is described on the page about testing whether distinctiveness actually exists.
An indication of what a competitor is deploying its capacity on can often be read from the way it recruits: an institution that posts vacancies for data analysts instead of extra administrative staff thereby reveals something about where it has already made its shift, as described on the page about what vacancy texts reveal about a competitor. The same shifting logic plays out in other sectors with comparable redistribution of hours, as seen in the analysis of what is won on in the IT sector now that AI takes over work and in the comparison of what is won on in retail now that AI takes over work, where automation of scheduling and accountability plays a similar role.
The underlying question is different for every institution: which part of the work in your organization can truly be taken over by AI, and which part remains human work with oversight. That question is answered per task with the work scan from FTE TO AI.
You can start by identifying what you believe you win on: responsiveness, individual guidance, accountability quality, or something else. The free dimension check shows which of those claims can be defended with evidence and which are mainly assumptions. The full benchmark, with the evidence matrix per dimension and the comparison with your peer group, is under construction.