Students already use AI, whether the school has rules for it or not. According to Statistics Netherlands (CBS), in 2024 34% of 12 to 18 year olds made texts, videos or images with an AI program such as ChatGPT. Among 18 to 25 year olds it was 48.7%. Pretending otherwise does not help a school. The question is whether teachers can handle it better than their students.
At the same time, for many teachers the biggest problem sits elsewhere: in everything around teaching. Lesson preparation, differentiation, letters to parents, report meetings, keeping the student information system up to date. That is where AI can make a difference sooner than in the classroom of the future.
Where AI gives teachers time back
The most interesting opportunity is in the work that now keeps a teacher from teaching. A letter to parents, preparing a report meeting, a summary of a student's progress: AI writes a first version, the teacher checks and adjusts it.
The same goes for differentiation. Personalised learning has been an appealing idea for decades, but nobody had time to build thirty practice routes for thirty students. With AI you quickly make three versions of the same assignment: a short one for students who already get it, an extended one with more examples, and one with a diagram for those who learn better visually. The teacher decides which student gets which version.
The teacher stays indispensable
A class needs someone who notices that a student is quieter than usual today, and who knows the group needs a moment to settle after PE. No model does that.
What AI can do is give the teacher time and attention back: less time typing lesson plans, more time talking with a student. Seen that way, AI is first of all a way to reduce workload.
What schools need to watch
Digital literacy, for teachers too
A teacher who does not know that a language model confidently makes things up cannot explain to a student why an answer is nonsense. If you want to teach students to use AI critically, start with your own staff. The law points the same way: Article 4 of the AI Act requires organisations that use AI to take measures that support the development of AI literacy among their staff. Since the amendment by Regulation (EU) 2026/1744, it does not require them to guarantee any specific level of AI literacy. That includes schools.
Plagiarism and assessment
Detection software is a weak foundation. Stanford researchers showed that AI text detectors are biased against people writing in a second language: their texts were often wrongly flagged as AI-written. It works better to design assignments differently: oral exams, presentations, writing in class and a portfolio that makes the process visible. Then you assess whether a student can explain their reasoning, rather than whether the end product is theirs.
Wrong answers
Language models can present a wrong answer as convincingly as a right one. So never let students rely on a chatbot without checking sources, especially in science and maths.
Student privacy
Schools work with sensitive data about children, such as information on support needs or social and emotional development. That data does not belong in a free consumer tool.
What a school board can do
Train teachers. Preferably with practical sessions where teachers learn to speed up their own lesson preparation, rather than with a single training day. As soon as a teacher notices that a letter to parents takes far less time, AI starts being used for real.
Clear rules for students, per subject and per level. A blanket ban does not work, and neither does blanket permission: a lower secondary assignment is very different from a final-year university of applied sciences thesis.
More room for critical thinking. Assessing sources, understanding how a language model works, knowing what a made-up answer is and how to check a claim. That belongs in every subject, from languages to history and biology.
A curriculum that can move. Nobody knows exactly what will be needed in two years. So build in the ability to adjust the programme.
Applications a school can start with now
Preparing report meetings. AI prepares a first summary of a student's progress based on test results and notes. The mentor checks and adds to it.
Communication with parents. AI writes a first version of a letter to parents about, say, a school trip or report grades, in the school's tone. The teacher reads, adjusts and sends it.
Feedback on drafts. A student gets direct feedback on a draft, so the final version improves. Grading stays with the teacher.
Test analysis. AI shows patterns teachers would otherwise miss, such as topics a whole class struggles with, or students who have trouble with one specific skill.
Tailored assignments. AI helps a team make variants of assignments and extra exercises for students with more or less prior knowledge.
For each application, measure beforehand how much time the work takes now, so that after a trial you know what it delivers.
GDPR and the AI Act in education
Schools process data about children, sometimes health data (such as extra support needs) and sometimes sensitive social and emotional information. Cloud AI with student data needs at least a data processing agreement with the vendor. If the data leaves the EEA, that is only allowed to a country with an adequate level of protection or with extra safeguards, such as standard contractual clauses (Dutch Data Protection Authority). Many schools and school boards also require in procurement that data stays within the EU and is not used to train models. That is the school's own requirement rather than a GDPR rule, and a sensible one for children's data.
The AI Act classifies part of AI in education as high risk: systems that decide who is admitted, that evaluate learning outcomes, that assess the appropriate level of education, or that monitor students during tests. Those systems face requirements on transparency and human oversight, among others. These obligations start on 2 December 2027. Take them into account now when you choose a system, because adjusting afterwards costs more.
How a project in education runs
A project runs in phases. First a short analysis of which processes cost the most time. Often that is communication with parents or test analysis, rather than the AI tutor a vendor proposes. Then a trial with one process, to see whether it fits the culture of the school. A lower secondary team works differently from a programme committee in higher education, and that difference decides whether a tool sticks.
Only then do you expand what works and stop what does not fit. Where possible, plan big transitions around school holidays. And be honest: if a trial shows no time saved or better quality after three months, stop, even if money has already gone into it.
Start with the workload
AI in education is mostly about today's workload. Start with what keeps teachers from teaching, choose tools that comply with the GDPR and where a person checks the outcome, and assume students use AI. Teach them to use it critically instead of banning it.
The schools that lead in a few years will not necessarily be the ones with the most expensive licences. They will be the schools where teachers spend their time on students again.
Curious what AI can do for your business?
Take the free AI Scan and find out in 1 minute.
Frequently asked questions
- Should schools ban AI?
- No. Students already use it: according to CBS, in 2024 34% of 12 to 18 year olds made something with an AI program such as ChatGPT. Clear rules per subject and level work better, along with teaching students to use it critically and training teachers first. A lower secondary assignment and a final-year thesis call for different agreements.
- How do you deal with AI and plagiarism?
- Detection software is a weak foundation: Stanford research showed that AI text detectors often wrongly flag texts by people writing in a second language as AI-written. It works better to design assignments differently: oral exams, presentations, writing in class and a portfolio that makes the process visible.
- Where does AI save teachers the most time?
- In the work around teaching: preparing report meetings, letters to parents in the school's tone, feedback on drafts and test analysis at class level. AI makes a first version, the teacher checks it. Measure beforehand how much time the work takes now, so that after a trial you know what it delivers.
- Which rules apply to AI and student data?
- Schools process data about children, sometimes health data and sometimes sensitive social and emotional information. Cloud AI with student data needs at least a data processing agreement. If data leaves the EEA, that is only allowed to a country with an adequate level of protection or with extra safeguards, such as standard contractual clauses. Many schools also require EU storage and no training on student data themselves. The AI Act classifies AI for admissions, evaluating learning outcomes and monitoring during tests as high risk, with transparency and human oversight requirements from 2 December 2027.
