Should AI replace or support? In the U.S., the teacher is being removed. In Estonia, the focus is on pedagogy. Also read about autonomous AI agents, the EPX debate, and why we're still asking the wrong questions about learning.
Here is this week’s newsletter on artificial intelligence in the education system, where the question of direction and frameworks once again looms large.
In A4 Medier, Tekniq argues that the new EPX programme should make AI a natural tool from the first day of school. At the same time, two researchers in Skolemonitor call for a broad public conversation about what AI means for knowledge and formation. Both articles point out that we lack a shared direction for AI in education.
Internationally, we are seeing two very different approaches, which we have written about before and which many have contacted us about. In the United States, CNN has a longer piece on Alpha School, where AI tutors have replaced teachers and where pupils learn core content in front of screens for two hours a day. Estonia has chosen a very different path with AI Leap, a national initiative that centres on teachers’ professional development and demonstrable learning effects.
At home, a new evaluation from VIVE (the National Research and Analysis Centre for Welfare) shows that trials of new exam formats involving AI have raised the average grade by about one point. At the same time, Novo Nordisk Fonden (the Novo Nordisk Foundation) has awarded over 6 million kroner for AI development on the technical vocational programmes.
Before the winter holiday, I delved into a new article about OpenClaw, an AI system that not only answers questions but can independently operate your computer and complete tasks while you sleep. As I write this newsletter, the developer behind OpenClaw, Peter Steinberger, has been hired by OpenAI. That says something about how fast developments are moving and which direction the major players see as next.
You can read the full article here:
Per Størup Lauridsen
This week’s stories point in many directions, but one question keeps recurring: Should AI replace or support teaching? In this week’s commentary, I write about why we may be asking the wrong question.
Happy reading.
In an opinion piece in A4 Medier, Maria Schougaard Berntsen, deputy director at Tekniq, argues that the new EPX programme should be developed as a markedly different youth education pathway, with AI integrated from the first day of school. She warns against turning EPX into a “gymnasium light” and believes that debate about structure and finance risks overshadowing the pedagogical innovation that is needed.
According to Tekniq, many young people are already familiar with AI in their everyday lives. The technology should therefore not be met with bans and control, but instead be made a natural working tool in both teaching and examinations. EPX should be a practice-oriented programme where pupils work with real-world problems and cases from working and everyday life. AI should therefore not be treated as a theoretical topic at a distance, but as a concrete tool for problem solving.
Read the article here:
Maria Schougaard Berntsen, deputy director, Tekniq
In an opinion piece in Skolemonitor, Associate Professor Emeritus Keld Nielsen from Aarhus Universitet (Aarhus University) and Associate Professor Martin Sillasen from VIA University College raise a fundamental concern: AI is changing the conditions for how knowledge is acquired and applied, but the broad public conversation about the consequences for the education system is missing.
The starting point is a conversation with a head of development in a large Danish company that uses AI as a sparring partner for complex technical tasks. AI tools are improving rapidly and can already analyse, synthesise and propose solutions at a high level. The concern, however, is about the employees of the future: If young people get used to AI delivering the solutions before they have built a fundamental subject understanding themselves, we risk ending up with professionals who can produce results without being able to evaluate them critically.
According to the authors, technological understanding as a standalone element in school is not sufficient. The challenge is not only about learning to use the tools, but about rethinking subjects, knowledge and formation in a situation where AI is freely available and developing quickly. The question is therefore where and how pupils develop the ability to understand, evaluate and take responsibility for AI-generated solutions if the technology takes over an increasing share of the actual problem-solving work.
A report on CNN describes Alpha School, a US private school chain where pupils from kindergarten to 12th grade learn core content via AI-based programmes for two hours a day. There are no traditional teachers, no textbooks and no homework. Instead, pupils work in front of screens with AI tutors, while adult “guides” motivate and support but do not provide subject teaching. The rest of the school day is spent on workshops on life skills such as finance, communication and problem solving.
The school markets itself on the claim that pupils learn twice as fast as their peers, measured among other things by national standardised tests. Teaching is individualised and data-driven, and pupils’ performance is measured daily. Some parents emphasise that their children have experienced rapid academic progress and greater motivation, partly because teaching is tailored to the individual and linked to reward systems.
Other parents, however, speak of stress, performance pressure and a one-sided focus on measurements and targets. Some have withdrawn their children from the school and found that they have fallen behind academically. Several experts express concern about whether highly individualised, screen-based teaching weakens social communities and collaboration skills.
https://edition.cnn.com/2026/01/29/politics/alpha-school-trump-ai-teaching
In Gymnasieskolen, Bodil-Marie Gade, an upper secondary teacher at Horsens Gymnasium & HF (upper secondary school) and a member of GL’s executive committee and chair of GL’s Education Committee, describes a study visit to Tallinn and the Estonian initiative AI Leap 2025.
The initiative is state-led and gives from 2025 around 20,000 upper secondary students and thousands of teachers access to AI-based tools. Systematic professional development is offered at the same time, particularly targeted at teachers, and the project is accompanied by research that continuously assesses the effects. A core premise is that AI is only used where it demonstrably strengthens students’ learning.
The article highlights that the Estonian approach is less polarised than the Danish debate. Instead of either rejecting AI or integrating the technology uncritically, Estonia has chosen a national and pragmatic strategy with shared goals and a clear political direction. Teachers’ role is central. They must have sufficient competence to set the framework for the use of AI and to assess when the technology supports subject knowledge and pedagogy. It is also recognised that pupils already use AI. The effort therefore focuses on critical understanding and didactic principles rather than bans.
Overall, Bodil‑Marie Gade points out that AI Leap is not a model to be copied uncritically, but an example of how a national, regulated effort can create shared direction, systematic professional development and ongoing evaluation of work with AI in the upper secondary school (gymnasium).
Bodil-Marie Gade
VIVE – Det Nationale Forsknings- og Analysecenter for Velfærd (the National Research and Analysis Centre for Welfare) has evaluated three pilots with new exam formats in Afsætning A on hhx, Mathematics F on eud and Mathematics G on avu.
In Marketing A on the hhx (Higher Commercial Examination Programme), a written and an oral test are combined into a single, group-based exam. Teachers and students largely find that the format works well, and that the oral component provides an opportunity to demonstrate subject understanding at a higher level. The use of generative AI is seen as a positive element, and teachers state that it does not complicate assessment. On the contrary, the exam format reduces stress and lessens suspicions of cheating.
In Mathematics F on eud (vocational education and training), the evaluation is more mixed. The pilot consists of an invigilated written exam without aids and an oral product examination. The oral component receives broad support, while views on the written exam are divided.
In Mathematics G on avu (general adult education), the evaluation is generally positive. The combination of a written exam and an oral case examination is seen as a more precise test of the course participants’ competences, and two separate grades are awarded.
Across all three pilots, the average grade has risen by about one point compared with the regular exams the previous year. The evaluation emphasises, however, that with the current data it is not possible to determine the cause of the rise, and that these differences should therefore be kept in view if the exam formats are considered for permanent adoption.
Text: Malene Romme-Mølby Photo: Shutterstock and private
Novo Nordisk Fonden (Novo Nordisk Foundation) has granted over 6 million kroner to a new AI project targeted at the electrician and plumbing and energy programmes. The aim is to develop teaching tools that integrate AI into practice-oriented teaching and into workplace training. The project is intended to strengthen the link between theory and practice and ensure that vocational education keeps pace with technological developments.
The project runs over five years and is carried out in collaboration with, among others, Det Nationale Center for AI i Samfundet (the National Center for AI in Society, CAISA), Aalborg Universitet (Aalborg University) and DPU at Aarhus Universitet (the Danish School of Education at Aarhus University), as well as several vocational colleges, including TEC, Mercantec, Herningsholm Erhvervsskole and Den Jyske Håndværkerskole.
In this newsletter there are two takes on what AI might look like in the future. In the United States, Alpha School has removed the teacher from the equation. Pupils learn core content via AI tutors for two hours a day, and the school promises that children learn twice as fast. Conversely, Estonia has decided that AI is used only if it demonstrably strengthens pupils’ learning, and that teachers’ professional development is central.
These are not just two different methods, but two very different answers to what learning is.
Alpha School rests on the assumption that learning is about acquiring information, and that the process can therefore be optimised. But in that assumption the difference between product and process is overlooked. AI can deliver the analysis and the solution, and it can do it quite well. But it is in the friction, in the slow acquisition and in the effort that personal formation and learning take place. If we remove those processes, we also remove what makes education more than the transfer of information. We might as well make do with a textbook and a sheet of paper with all the answers.
That is precisely the concern raised by several researchers in Denmark. If young people get used to AI delivering the solutions before they themselves have built a basic subject understanding, we risk ending up with professionals who can produce results without being able to evaluate them critically.
The fundamental question, however, is not: How do we integrate AI into teaching? That question leads us straight to technical solutions, such as which ChatGPT we should buy for the pupils. The right questions, in my view, are: What is it we want for the pupils? What should they be able to do that AI cannot? And what is indispensable in teaching that no technology can replace?
When we frame the questions in that way, it is not about being for or against AI, but about knowing when the technology supports deep disciplinary learning. That requires teachers to have the competence and the mandate to make that judgement, which in turn is about continuing professional development and curiosity about how the technology affects their professional practice.
The question is not whether AI will be part of teaching. The question is whether we will steer the development or let it steer us.
