A school or educational institution consists of two kinds of time that are rarely divided equally: time in front of the classroom or lecture hall, and time outside it. That second category is larger than is often assumed. Lesson preparation, drafting and grading tests, reporting, accountability to inspectorates or governing boards, timetables, communication with parents or students, and administration around care and guidance. Much of that work is language-based: it consists of reading texts, writing texts, assessing texts. That makes education a sector where the shift of work through AI is felt concretely and early, because text work is precisely where language models do the most.
Added to this is a circumstance that sets education apart from a shop or a factory: the outcome of the work is someone's development, and that development is assessed and funded according to fixed frameworks. A curriculum, an examination programme, an inspection framework. That co-determines which work can be shifted and which work remains with the teacher, independent of what is technically possible.
In some schools, a first version of a test, a rubric, or a piece of feedback wording is already drafted by AI, with a teacher checking, refining, and only then releasing it. In other schools this is not yet happening, and the difference rarely lies in the willingness to innovate. It lies in three things that must all be present before work actually transfers.
The first is access: is a teacher allowed to use a language model on student data, or is there a data processing agreement and a privacy assessment still pending. The second is oversight: who assesses whether an AI-generated assessment of a student is correct, with what authority and according to what protocol. The third is trust on the part of the user: a teacher who doubts whether an AI suggestion does justice to a specific student will ignore that suggestion, however good the system is.
These three categories run through all the work in education. Some of it AI can take over independently, such as summarizing a long policy document or drafting a first version of a timetable. Some of it proceeds with human oversight that approves or rejects with reasons, such as feedback on an essay or a draft report. And some of it remains human work, such as the conversation with a student who is struggling, or the pedagogical judgment behind a care referral. Which part ends up where differs not only per institution but also per task within the same institution.
"We use AI to reduce workload" is a sentence that fits into virtually every board report and in practice establishes almost nothing. Is the intention that grading time is freed up for more attention per student, or that a fixed amount of teaching time is filled with fewer support staff? Both are legitimate ambitions, but they require different capabilities: different systems, different authorities, different agreements with staff representation. As long as that choice has not been defined in a concrete future state and tested against what the organization must be able to do for it, a management team keeps talking past each other using the same words.
That is where a readiness assessment makes a difference: not rephrasing the ambition, but testing whether the conditions for it are present. Think of data access, of clear decision rights over who may approve an AI suggestion, of trained skill among teachers and staff, of an oversight structure that fits the sensitivity of student data.
If text work shifts, teaching time or support time is freed up. What a governing board does with that freed-up capacity belongs to that institution's own strategic and employment-law considerations, with its own legal requirements when it comes to personnel decisions. This page describes only where the work itself is moving, not what a governing board ought to do with that.
The underlying dynamic — language-based work shifts sooner than physical or relational work, and oversight remains the governing factor — is not unique to education. Similar patterns can be recognized in the AI ambitions at play in the IT sector, in the way retail views AI tasks, and in the very different starting position of AI ambitions in hospitality. What differs each time is not the technology but the readiness of the organization behind it.
An ambition such as "AI supports test development" translates into concrete roles: who drafts, who approves, who is responsible when an error slips through. That translation from ambition to role and decision right is central to the page on which capabilities an organization needs to realize an AI ambition, and for a school board or Board of Governors it is rarely a technical question alone — it is also a question of how a management team gets aligned on what the ambition precisely entails.
The question of which work in your institution can genuinely be taken over by AI cannot be answered with a general sector picture; that question is answered per task with the work scan from FTE TO AI. If you first want to know where your organization stands at this moment, the free readiness check with eight short questions, one per dimension, gives you a picture of where you are furthest along and where you are least far along. The full ambition assessment, with the four layers from vision to target state and the translation into roles and decision rights, is under construction.
Vertel wat u wilt bereiken, dan kijken we samen wat daarvoor moet staan.
Answers come from this site’s knowledge base. Not tailored advice, and not a scan of your company.