Broadly speaking, there are two groups of people in society when it comes to artificial intelligence. The first group is represented by people like Prime Minister Kristen Michal who believe the economy could grow by as much as 50 percent over the next decade thanks to AI. In an interview with Sirp, Rein Raud said AI is frightening, will bring culture to a standstill and also contributes to climate problems. An increasing number of young people are among those who share that view. I probably don’t even need to ask which group you belong to, do I?
I think you do need to ask because the fact that I lead the AI Leap program does not necessarily mean I am completely blind to AI’s impact on the environment. I think the whole issue is a bit more complicated than simply belonging to one camp or the other. The most important point is that AI, as a phenomenon, is not going back in the box. The question now is what we can do to ensure that people do not fall behind intellectually in an educational context. That is the fundamental question we are asking through the AI Leap program.
The part of your program that has attracted the most public attention is the learning application called ITI. But the program itself is broader than that and includes other responsibilities as well, correct?
Absolutely. For us, the most important thing is to make sure teachers are not left on their own in this context. I think Estonia has done well in this regard because, within education, Estonian teachers are very creative and are always looking for opportunities on their own. The Estonian education system gives teachers a considerable degree of autonomy.
Even so, they also need support alongside that autonomy, whether it is access to resources that allow teachers to explore AI themselves or help in understanding what digital competence means today when AI is now part of the picture.
That is exactly why the AI Leap program was created. Of course, AI Leap is not operating in a vacuum. For example, when it comes to defining and developing digital competence, our partners include the Education and Youth Board and the Ministry of Education and Research. We work closely with both of them.
There are also many subject-specific questions when it comes to teachers. AI is certainly not a tool that naturally fits into every subject, especially when we are talking about generative AI. Take subjects that are clearly highly creative, such as art, music or physical education, which we now refer to as movement education. What role, if any, should AI play in those contexts?
Over the course of this year, teachers have had extensive discussions with their students, particularly upper secondary school students. And indeed, what you mentioned — that many young people today are highly skeptical of AI — is absolutely true.
Within schools themselves, there are essentially three groups. There are students who have made a very conscious decision not to use AI. There are students who are largely indifferent to the issue and simply use whatever tools are available to make schoolwork a little easier. And there are students who use AI very deliberately, whether through our learning application or other AI solutions, to better understand material they may not have fully grasped in class. Some students have developed a very clear understanding of the possibilities this technology offers.
But I think it is a very positive thing that we also have students who have taken a clear position and carefully thought through their approach. Reflecting on the issue — regardless of which conclusion one reaches — is always better than being indifferent to it.
It seems to me that, in the school context, the most important thing is working with teachers. Young people often figure out on their own how new technologies work anyway.
I recently heard a story about an event marking the first anniversary of your program. One teacher — who had been invited as a sort of leading example because she had made extensive use of AI in her teaching — said she had not actually done anything particularly special and that her students understood AI better than she did. How can you get teachers to genuinely want to use AI and know how to use it in ways that make lessons better and more effective?
There are several approaches. One of the first things we did, starting in August 2025, was organize large-scale lectures for teachers. They attracted a substantial audience, with thousands of upper secondary school teachers from across Estonia participating.
Later, we also adopted a learning-circle approach. Learning circles are a methodology familiar to most Estonian teachers. Typically, one or two teachers lead a group after receiving additional support on a particular topic. In our case, that support comes from the AI Leap program and those teachers then share their knowledge with colleagues at their own schools.
We also have to recognize the circumstances under which AI Leap was launched. The program was announced in early spring 2025. By that point, many upper secondary school principals had already finalized their plans for the year. Nevertheless, many school leaders were accommodating and found ways to incorporate discussions about AI into their strategic plans.
I am very grateful to the school leaders who opened their doors to us. At the same time, I completely understand those who told us they would address the issue in the next academic year because their plans for the current year had already been set. In a school setting, the topics a school’s leadership and teaching staff choose to take on together require careful consideration. Time is the most valuable resource in schools and teachers’ free time is particularly precious.
I am very pleased with the teachers who have already engaged with the program and I am equally pleased with the school leaders who have said they will definitely take this up in the next academic year.
When it comes to subject-specific issues, AI Leap recently launched a section of its website called didaktika.tihupe.ee, which currently contains materials covering five different subjects. It addresses the pedagogical aspects of those subjects and explores how AI can be useful in each of them. It also includes comments from teachers themselves, describing the methods they have tried and the positive as well as negative experiences they have had. Because the field is so new, we are quite literally figuring these things out together with teachers as the AI Leap program unfolds.
Beginning in the new academic year, we will also take a much more subject-specific approach to training. That could mean sitting down exclusively with physics teachers, or perhaps only biology teachers, and looking at AI from the perspective of those disciplines.
Broadly speaking, teachers are constantly shaping the pedagogical side of education themselves.
At the same time, there needs to be ongoing work in educational research and educational psychology to assess whether the things we are doing are actually beneficial. Whether we are talking about AI or some other technological tool a teacher happens to have on a computer, if it does not improve efficiency or enhance quality, then it is simply technology for technology’s sake — and that is something we should avoid.
Are you conducting that kind of reflection and analysis yourselves or are you relying on outside help?
Fortunately, we have strong partners for that work. Associate Professor Jaan Aru’s research group is already studying the impact of the learning application itself. The research they are conducting in cooperation with AI Leap and the Ministry of Education and Research examines, among other things, teachers’ and students’ levels of AI literacy. It also addresses a number of other topics. By the end of the academic year, it should provide us with a fairly solid assessment of the impact of AI Leap’s first year.
According to Jaan Aru, the final analysis should be completed by the end of this calendar year. On the one hand, we can be pleased that yesterday [Monday] we were able, in a sense, to draw a line under the first pilot year and say it has been completed. Now we can look ahead and consider how to continue supporting teachers and students in the program’s second year.
But the most important thing is that researchers provide a genuine assessment of the results — and specifically Estonian researchers who understand the context of our culture and education system.
Let’s return to the learning application called ITI. It is a model designed specifically for Estonian educational institutions and is somewhat different from a typical AI tool. It is not simply a digital helper that does whatever you ask; instead, it asks a lot of questions in return.
I had a chance to try it out myself and asked, for example, “Please tell me Eduard Vilde’s pseudonyms.” It immediately started asking me why I wanted to know and what context I wanted to place the information in. It left me wondering why a young person would want to use it at all when it is much more difficult to use than a standard AI tool.
Yes, I think the experience you had is, in itself, a learning experience — it encourages users to think through the learning process. But the question is a fair one.
If we think about why it asks so many questions, the honest answer is that we ourselves will certainly revisit this issue repeatedly and try to fine-tune it so that it does not feel necessary to ask follow-up questions about absolutely everything. In some cases, it makes perfect sense for the system simply to provide a correct answer, especially when it comes to a brief factual question that a student could easily find through a quick Google search at home. We have the technological capability to do that in such situations.
At the same time, we have to keep the educational philosophy in mind. We have had extensive discussions with Jaan Aru, Grete Arro and other education researchers about what we are actually trying to achieve. Simply handing students every answer is not always the best way to learn. Students themselves have been telling us that for quite some time.
Some time ago, one of our employees conducted a fairly extensive vox pop survey. They visited upper secondary schools around Estonia and collected user feedback. One of my favorite quotes emerged very clearly from those conversations, and it is something I have even shared with our technology partner, OpenAI.
The students said: “A lot of potential, but really annoying.”
In a sense, that is exactly the experience you had.
What would the ideal scenario look like? How exactly should it guide, facilitate or support learning?
In some ways, it comes down to the realities of the Estonian classroom. If we consider that, in Estonia’s education system, there are still around 36 students for every teacher in an upper secondary school classroom, it is difficult to expect that all 36 students will receive the same level of attention as the student who is willing to raise a hand, engage in a dialogue with the teacher and actively participate.
At the same time, there are students in the classroom who may simply be more tired that day. Or perhaps the topic itself is particularly difficult. In those cases, having access at home to an AI tutor that can help them make sense of a subject and figure out how to approach it can be genuinely beneficial. We already know that some students are benefiting from it.
In that context, an AI tutor is simply another way for students to learn and understand something, provided that the technology suits their needs. Even so, nothing better than face-to-face teaching has yet been invented in education.
The biggest challenge remains the ratio of 36 students to one teacher. A single teacher in Estonia does not always have enough time for everyone. At the same time, the diversity of student needs within the education system is increasing, which means we also need to think carefully about how to reduce the burden on teachers. This technology can help teachers as well, albeit from a different angle.
For example, there are cases where younger teachers record their lessons using Whisper, OpenAI’s speech-recognition model, and then use those recordings to build a high-quality collection of teaching materials for future academic years. In other words, the spontaneous ideas and discussions that emerge in class can be transformed into educational resources.
At the same time, large language models have become very popular among many language teachers. The logic is fairly straightforward: language models are exceptionally capable in language-related contexts. They perform very well in English and are also highly proficient in languages such as Spanish and French. As a result, we are seeing extensive use among language teachers.
Teachers are creating specialized learning experiences for students, including simulations that allow them to, for example, discuss French politics with Emmanuel Macron during class. Teachers are approaching the technology very creatively and finding new ways to use the opportunities it offers.
Once again, though, this is simply another tool. It requires caution because the technology has more sharp edges than many of the technologies that entered schools before it. That means teachers need to have a clear understanding of what it can do, what it cannot do and when it is better to rely on the traditional pen and paper approach.
I was also looking at the numbers. If I understood correctly, the learning program has a total of 12,000 student users. Of those, 4,000 use it weekly, which is about one-third. Is that a result you can be satisfied with?
Those are unique users. In other words, they are not necessarily the same users every week. We also know that students often use the technology to prepare for assignments and tests. One popular use case, precisely because the system asks so many questions, is that students upload materials provided by their teacher into ITI and use it to help prepare for an upcoming assessment based on those materials.
The situations in which students use the platform vary considerably. Some students have also told us that they are not weekly users. The figure of 4,000 refers to the number of unique users, not the same 4,000 students using it continuously week after week. The actual reach is somewhat higher than that.
As I mentioned earlier, for some students this technology is an excellent learning companion. Some students already find it useful in its current form.
At the same time, as the good journalist has pointed out, some students may feel that it asks too many questions. We are continuing to refine the technology so that it becomes more appealing and useful to a broader range of students.
Let’s also touch briefly on funding because that is always an interesting topic. Your foundation has said the program should operate on a 50–50 principle, with half of the funding coming from the state and the other half from the private sector. Have you been able to stick to that model?
Yes, we have.
Even so, you asked the Ministry of Education and Research for an additional €1 million in funding. What will that money be used for?
Our financial year follows the academic year, so it ends on August 31.
The same funding request shows that €60,000 of that additional million would be allocated to communications. Why exactly is that amount necessary? I understand that you already have communications staff in-house, but why is there a need for additional external communications support?
If we look at where the program is headed, we are planning not to remain focused solely on upper secondary schools. We are also preparing to expand into vocational education and, in the future, to support basic school teachers as well. I cannot yet say with complete certainty what that support for basic school teachers will look like. We definitely want to support them, but in reality the strongest criticism we receive of the AI Leap program tends to come from basic school teachers. Many of them still do not understand why our efforts have been focused so heavily on upper secondary and vocational education.
From a teaching perspective, basic school teachers make a valid point: many of the habits, tricks and forms of AI misuse that concern educators are developed at an earlier age. We are talking about a target group that already numbers more than 60,000 people and we need a way to reach them.
That means we need to ensure effective communication about training opportunities and the information we provide. In my view, no organization can operate successfully without a communications budget.
So that funding would primarily be used for internal communications?
Yes. It would be used to inform teachers and support internal communications.
To wrap up, it would be appropriate to ask: where do you go from here? What are the next steps you want to take and how do you see the program developing?
The most important thing is that, in the coming days, we reach an agreement with the ministry on whether and to what extent we will expand our activities.
We have already reached preliminary agreements with vocational schools and will begin supporting vocational school teachers. The question now is whether and how we can also support vocational school students, particularly those enrolled in four-year programs.
A separate issue — and, in my view, one of the most justified criticisms directed at the AI Leap program, as well as a strong personal conviction of mine — is that many of the learning challenges teachers are currently facing in upper secondary schools and vocational schools begin much earlier. Generative AI creates opportunities, but it also introduces many new problems.
From my perspective, support for teachers needs to start earlier, at the basic school level. We are already thinking seriously about how we can also be useful to basic school teachers.
—
