Bram Lesage, 17, from Mechelen, Belgium, on why students should learn to use AI — and why the thinking still has to be theirs
My name is Bram Lesage. I'm 17, I live in Mechelen, Belgium, and I'm still in school. I also build software with AI every day. So I see AI from two sides: as a student who gets homework, and as a maker who uses AI agents to build real products.
My opinion is simple. Students should learn to use AI as a tool, and they should learn to check its work themselves. Pretending AI doesn't exist doesn't work. Letting AI do the thinking for you doesn't work either. The interesting part is in between.
Students are already using it
This is not a future debate. In Flanders, the MAIVO study, commissioned by the Department of Education and published in September 2026, surveyed school leaders, ICT coordinators, 586 teachers and 636 pupils across 482 schools. Its main conclusion: AI is already present in schools, but its use is mostly exploratory, informal and not built into how schools work. Pupils, especially in secondary school, use AI to look things up, ask questions and improve texts — sometimes outside class, without a clear assignment or permission. The researchers call that shadow use, and they found that many schools still have no formal AI policy.
So the question is not whether students use AI. It's whether anyone teaches them how.
Banning it doesn't solve much
Kenniscentrum Digisprong, the Flemish government's knowledge centre for digital education, puts it bluntly in its practical guide for schools: banning generative AI like ChatGPT has "little to no effect". It advises schools to write clear guidelines for pupils, teachers and staff instead, to make agreements about AI in tests and evaluations, and to move from grading only the end result to grading the whole learning process.
I agree with that. A ban mostly teaches students to hide what they do. Clear rules teach them when AI is allowed, when it isn't, and why.
But AI can make you worse at learning
Here is the part people who love AI (like me) need to hear. A field experiment with nearly a thousand high school maths students in Türkiye, published in PNAS, tested what happens when students practise with GPT-4. With a standard chatbot, their practice scores went up by 48%. With a version designed as a tutor, they went up by 127%. But when the AI was taken away and they took a normal exam, the students who had used the standard chatbot scored 17% worse than students who never had AI at all. The researchers found that many of them had used it as a "crutch": asking for the answer and copying it.
The tutor version, which was built to help students think instead of handing them answers, did not cause that drop.
The OECD's Digital Education Outlook 2026 draws the same lesson from the research. When students lean too heavily on generative AI, the mental effort that turns answers into understanding drops, so the work looks better while the learning does not improve. The OECD suggests that teachers focus on process, on how students think and learn, rather than only on the output.
That is the most important fact in this whole debate. Getting the right answer and learning the material are not the same thing.
What I think good AI use at school looks like
- Learn first, then speed up. If you don't understand fractions, an AI that does fractions for you just hides the problem until the exam. Use AI to explain, quiz you and give feedback, not to skip the part where you learn.
- Always check the work. AI makes mistakes, and it says wrong things with confidence. If you hand something in, you are responsible for every sentence in it, not the AI.
- Follow your school's rules. Schools are allowed to forbid AI help, for a task or entirely. Tests and exams are about what you can do on your own. That's fair, and I think students should respect it.
- Be open about it. If AI helped and it was allowed, say so. Hiding it is what turns a tool into cheating.
- Keep personal data out. Digisprong advises schools to keep AI tasks to non-personal questions, like a maths exercise, and to be careful with pupils' personal data.
Schools are being asked to teach this anyway
Since 2 February 2025, the EU AI Act requires organisations that provide or use AI systems to take measures for a sufficient level of AI literacy among their staff (Article 4). Digisprong points out that this applies to school boards too, and asks the obvious question: if staff need to understand the risks of AI, shouldn't pupils who are allowed to use it understand them too?
There is also a lot of confusion about age. Digisprong explains that the AI Act itself sets no minimum age; the limits come from privacy law (in Belgium, 13 for consent to online services) and from the terms of use of each AI tool. UNESCO's guidance on generative AI in education recommends an age limit of 13 for independent use and asks schools to check tools before students use them. Katholiek Onderwijs Vlaanderen writes that there are no binding Flemish agreements on AI in schools yet, that its model school rules allow pupils to work with generative AI under adult guidance, and that it is the task of schools to teach all their pupils to use AI responsibly and critically.
Everyone serious seems to land in the same place: not "ban it", not "let it do everything", but "teach it, with rules".
How I use AI when I build
I build my projects like a director. For a product or a game I often put together a team of AI agents — a designer, a builder, a tester and a strict reviewer. That lets me build much faster than I could alone.
But the agents don't decide when something is done. I test every version myself and decide what ships. Green automated tests don't mean it works: on the first day of one of my Roblox games, 215 automated tests were green, and some systems still hadn't been run in Roblox Studio. The AI does a lot of the work. The judgement is mine, and so is the responsibility.
That is exactly the skill I think school should teach: not how to type a question into a chatbot, but how to give AI a clear job, check what comes back, and know when it's wrong.
One of my own products is Katern, a Chrome extension that helps with schoolwork. I designed it so the student stays in control: it shows its plan first, it never presses Hand in, the student checks the work and hands it in themselves, and it does not do tests or exams. Schools may forbid AI help, and that is their decision to make.
My conclusion
AI at school is not going away, and it shouldn't. Used well, it explains things at 11 at night when no teacher is around, gives feedback on a draft and quizzes you before a test. Used badly, it gives you good grades on homework and bad grades on the exam.
The difference isn't the tool. It's whether you still do the thinking. That is what schools should teach, and that is what students should demand to learn.
Sources
- MAIVO study, "AI in het Vlaams onderwijs anno 2025", Department of Education and Training, Flanders (published 1 September 2026): https://www.vlaanderen.be/onderwijsprofessionals/regelgeving-data-en-onderzoek/data-en-onderzoek/kennisplatform-onderwijs/ai-in-het-vlaams-onderwijs-anno-2025
- Kenniscentrum Digisprong, "Praktische gids: hoe ga je bewust en verantwoord met AI om in je school?": https://www.vlaanderen.be/kenniscentrum-digisprong/themas/innovatie/artificiele-intelligentie/praktische-gids-hoe-ga-je-bewust-en-verantwoord-met-ai-om-in-je-school
- Kenniscentrum Digisprong, "Vanaf welke leeftijd kunnen of mogen je leerlingen generatieve AI gebruiken?": https://www.vlaanderen.be/kenniscentrum-digisprong/themas/innovatie/artificiele-intelligentie/vanaf-welke-leeftijd-kunnen-of-mogen-je-leerlingen-generatieve-ai-gebruiken
- Bastani et al., "Generative AI without guardrails can harm learning: Evidence from high school mathematics", PNAS (2025): https://www.pnas.org/doi/10.1073/pnas.2422633122 (preprint: https://hamsabastani.github.io/education_llm.pdf)
- OECD, Digital Education Outlook 2026: https://www.oecd.org/content/dam/oecd/en/publications/reports/2026/01/oecd-digital-education-outlook-2026_940e0dd8/062a7394-en.pdf
- EU AI Act, Article 4: AI literacy: https://artificialintelligenceact.eu/article/4/
- UNESCO, Guidance for generative AI in education and research (2023): https://www.unesco.org/en/articles/guidance-generative-ai-education-and-research
- UN News, "UNESCO calls for regulations on AI use in schools" (7 September 2023): https://news.un.org/en/story/2023/09/1140477
- Katholiek Onderwijs Vlaanderen, "AI-regelgeving": https://pro.katholiekonderwijs.vlaanderen/artifici%C3%ABle-intelligentie/airegelgeving
Written by Bram Lesage. More of my projects: bramlesage.pages.dev