2026: Why Kazakhstan’s schools cannot simply “implement” artificial intelligence

In September 2025, the President of Kazakhstan made a statement that is worth reading slowly: “With the advent of artificial intelligence, a clear dividing line has been drawn between those countries that manage to enter the future and those that remain in the past.” He then added another, this time directly about school: “Competencies in artificial intelligence should begin to be formed much earlier — from school.”

2026 has been declared the Year of Digitalization and Artificial Intelligence in Kazakhstan. And this is not a routine slogan for the calendar. Behind it are concrete decisions: a national education platform and an AI platform for teachers are being launched, a system of digital profiles for schoolchildren is being created, more than 200,000 teachers are expected to acquire practical AI skills this year, and a working group under the ministry is preparing proposed amendments to legislation. The AI Kitap platform has been launched to personalize learning, the Day of AI methodological package has been developed, ethical standards have been approved, and in pilot regions a Qazaq Digital Mektebi project with AI tutors is being prepared for small schools. The Minister of Education formulates the task as follows: “to effectively use artificial intelligence in the education system to create an educational environment where next-generation technologies become an assistant.”

At first glance, everything seems clear: there is political will, there is a budget, there are plans. Just implement it. But this is exactly where the most difficult — and the most interesting — part begins. Because the question of how quickly we will introduce AI into schools turns out to be the wrong one. The right question is different: how do we manage something that is already happening on its own?

This article opens a series in which we will try to take an honest look — free of both techno-optimism and technophobia — at what the arrival of AI in Kazakhstan’s schools really means. And we have to begin with a few uncomfortable admissions.

Let’s start with a fact that many prefer not to notice. Artificial intelligence did not come to Kazakhstan’s schools by ministry order or after the ceremonial launch of a platform. It is already there—brought in by the students themselves, in their pockets.

Our study, which involved hundreds of representatives from the education system, states this plainly: AI tools are used widely, but in a piecemeal way and without common rules. Students turn to chatbots to solve problems and write essays. Teachers use them to prepare materials. And all of this happens in a gray zone: without methods, without standards, without a clear understanding of boundaries.

Moreover, the study revealed a troubling detail: students often trust AI more than teachers and parents. The algorithm responds instantly, does not judge, does not give you a failing grade for a stupid question, and is available around the clock. For a child, it is the ideal interlocutor. The problem is that the ideal interlocutor quietly turns from a tool for thinking into its replacement — into an “answer machine” that people turn to not to think, but to get a ready-made result. And that directly undermines the very purpose of school.

At the same time, adult participants in the process face their own challenges. Teachers lack the knowledge and skills to work with AI; if the transition is forced, there is a real risk of burnout. A teacher who is simply told, “From now on, we work with AI,” but is given neither methodology nor time nor a clear explanation of why, will see the innovation not as help but as additional unpaid workload. And this is not a specifically Kazakhstani issue: South Korea ran into exactly this problem, where teachers met complex digital platforms with outright resistance.

The picture is further compounded by infrastructure. Weak internet — even in major cities — a shortage of devices, a lack of Kazakh-language content, and poor localization of global solutions. In other words, AI is already spreading through schools, while in some places the basic conditions for its meaningful use simply do not exist.

This is the starting point that must be stated honestly. The debate over whether to “introduce AI or not” is no longer relevant — the question is already outdated. AI is already in schools. The real choice today lies elsewhere: whether to leave this process unstructured and chaotic, or make it manageable, thoughtful, and safe. In the language of our research, this is the path from “disarray and chaos” to “awareness” — and it is precisely this path that the education system will have to travel over the next ten years, avoiding dozens of risks along the way.

The Kazakhstani education system has experience with digitalization, and a good one at that. Electronic diaries and registers have become routine, and e-government services in education are functioning. It is natural to be tempted to see AI as the next step along the same path: install new software, train people to use it, and move on.

This is the main trap of scale.

Artificial intelligence is not just another digital tool added to the existing system without fundamentally changing anything in it. It is an infrastructural shift that affects the very foundations of the school: the goals of learning, the role of the teacher, assessment methods, the meaning of exams, and even the answer to the question of why school is needed at all.

The difference is fundamental. When the electronic gradebook came into schools, assessment remained assessment — only the way it was recorded changed. When generative AI enters the classroom, assessment itself comes into question: what is the point of checking a homework essay if an algorithm can write it in five seconds? What is the point of a written test as a form of assessment if the line between “the student thought it through” and “the student copied from AI” becomes invisible?

Treating AI as a “localized digital project” leads to an error of scale—and, as a result, to fragmented, poorly managed decisions. A sleek platform is rolled out, indicators are reported, but beneath the surface the same ad hoc chaos continues, now also backed by budget funding.

And this is not only about assessment. The traditional “school — university — profession” chain, on which the entire logic of education over the past decades rested, is called into question. In the past, school prepared students for a more or less predictable future: here is a body of knowledge, here is the exam, here is admission, here is the profession. Today, automation affects not only routine operations but also complex intellectual functions — which means that the very “middle” level of qualifications is losing value, the very entry point into a profession where graduates used to land. School can no longer physically guarantee social predictability through the transfer of a fixed body of knowledge. And this changes not a single lesson, but the meaning of the entire enterprise.

There is another reason why education cannot be run according to the logic of a tech startup: “let’s roll it out, test it on users, and fix it on the fly.”

Scale. Kazakhstan’s school education system includes more than 4 million schoolchildren, over 500,000 teachers, and at least 3 million parents. Together with preschool education, this is a social institution that directly affects a significant share of the country’s population. It is one of the largest items in the state budget: only per-capita funding for private schools is planned at 248 billion tenge for 2026, and 1.7 million students are covered by free hot meals.

In the startup world, a mistake affects a limited group of users — dissatisfied users leave, and the product gets refined. In education, any decision is instantly scaled to millions of children, and a mistake becomes social and long-term. A failed experiment is not a “negative review” but a lost school year for an entire generation, eroded trust in schools, and greater inequality.

That is precisely why education cannot and should not be a testing ground for unlimited technological experiments. The high cost of error is not a reason to do nothing, but it is a compelling reason to act cautiously, in stages, and with the ability to stop.

The danger that is discussed less often: AI amplifies what is already broken

There is another twist that is hardly ever mentioned in the enthusiastic talk about the digital future. We are used to thinking of AI as a tool that solves problems. But under certain conditions, it does not solve them — it amplifies them.

Imagine an education system that, even without any AI, is oriented primarily toward memorizing and reproducing ready-made answers rather than developing thinking skills. What happens if you pour powerful generative AI into such a system? It will fit into it perfectly — and push its logic to the limit. Why think if you can generate? Why try to understand if the answer is delivered ready-made? Instead of compensating for a deficit of independent thinking, the technology will entrench it and turn it into a chronic condition.

This is not a hypothesis — it is an observable international pattern. AI encourages the “outsourcing” of cognitive functions precisely where school has already failed to teach students how to think. In this sense, the discussion of AI inevitably comes down to an honest question about the state of the school itself: how much it develops thinking rather than drilling students to fit a format; how much families trust it; whether parallel educational markets for tutoring and exam preparation have grown up around it, which in themselves are an indicator of a loss of trust. Introducing AI on top of unresolved fundamental problems means risking that the technology will not cure the disease, but entrench it. We will return to this issue separately — it deserves it.

Here, the counterargument usually arises: let’s simply look at how leading countries have done it and copy the best practices.

The problem is that there is nothing to copy yet.

No country in the world currently has a fully developed, sustainable model for integrating AI into school education. Even those commonly regarded as leaders are moving through experiments, pauses, and revisions. Singapore is introducing AI cautiously, through limited pilots, and emphasizes in official documents that AI does not replace the teacher. Estonia, a digital leader in public administration, is much more restrained in schools, focusing on infrastructure and data rather than AI in the classroom.

The most telling case is South Korea. A country with huge technological ambitions and a strong school system invested in AI textbooks and digital platforms — and faced such a wave of criticism that the program was rolled back just four months after launch. Teachers saw the complex platforms as extra unpaid work for which they had barely been trained. Parents took to the streets in protest, fearing “digital addiction” and a loss of children’s face-to-face communication skills. The pilots exposed mundane technical problems — with connectivity and with authentication. The sum of these seemingly minor difficulties proved enough to bury a high-profile government initiative.

At the same time, the range of reactions is itself telling. When ChatGPT appeared in early 2023, a number of U.S. school districts — including the largest one, New York City’s — rushed to ban it. But within a few months, New York lifted the ban and shifted toward regulated access: a blanket block did not work and only drove use underground. China took a more targeted approach: during the nationwide university entrance exam (gaokao) in June 2025, major AI services temporarily disabled photo recognition in chatbots so that school students could not scan exam tasks. And in American and European classrooms, handwritten and oral exams are making a comeback at the same time — as a quiet response to concerns about the independence of thinking. Serious analysis also tempers the enthusiasm: a major publication, examining the boldest forecasts about AI-driven school transformation, explicitly calls them based on an oversimplified understanding of what education actually is. After all, school does not merely transmit knowledge — over ten or eleven years, it shapes discipline, communication skills, responsibility, and the ability to think. That is hard to automate.

The conclusion from international experience is paradoxical, but important: most of the problems involved in introducing AI into education are institutional rather than technological in nature. The issue is not that the technologies are “not good enough.” The issue is that society, schools, and the state are not keeping pace in establishing rules, trust, and accountability. Where this is neglected, even an excellent technology fails. A broader rhetorical shift is also evident: if early strategies promised that AI would sharply improve the efficiency of education, later documents from international organizations describe AI more as a factor that complicates system management — and propose a logic of cautious pilots, mandatory impact assessment, and the option to stop.

For Kazakhstan, this is not bad news, but a window of opportunity. Since no one has ready-made recipes, the country is on an equal footing with everyone else. And the winner will not be the one who buys someone else’s technology faster, but the one that manages the process more skillfully.

All of this gives rise to a question that, just five years ago, would have seemed like a philosophical abstraction, but today has become intensely practical. If any fact, any solution, any explanation can be obtained from an algorithm in seconds, then what is school for at all?

One coach to top executives at technology companies put the essence of the shift this way: our entire society is built on the idea that knowledge is a scarce and precious resource. Schooling, exams, diplomas, interviews — all of these are mechanisms for assessing and rewarding knowledge. Now imagine a world where knowledge is no longer scarce. What still retains value? What cannot be downloaded: the ability to live, think, choose, take responsibility, and empathize.

In our research, this question takes the form of a choice between two models of school. The first is school as a place of control and discipline, a kind of “barracks,” where order and standardization are paramount. The second is school as a “garden,” an ecosystem that nurtures a living human being: their thinking, empathy, resilience, and ability to work with others. As long as knowledge was scarce, one could put up with the “barracks” model — at least it transmitted that knowledge. In a world where knowledge is widely accessible, the “barracks” model loses its meaning altogether: the school still cannot and should not compete with an algorithm in delivering information.

That is precisely why the discussion of AI in schools so quickly stops being a discussion about technology and becomes a discussion about values. What do we want to preserve in a person that cannot be automated? And how should school be designed so that AI strengthens that preservation rather than erodes it?

Taken together, everything said above reveals one major shift in thinking that still lies ahead — and that is the core idea of this whole series of materials.

From the question of “which technology to implement” to the question of “how to manage the process.”

This changes the role of the state. In the AI transformation of education, the state is not the developer of every solution, not the operator of every tool, and not the overseer of every classroom. Its task is more subtle and more important: to be the architect of the framework, the guarantor of values, the moderator of pace, and the protector of those who are vulnerable in this process — above all, children and teachers.

Several principles emerge from this, which we will return to repeatedly in the following materials. AI is a helper, not a replacement for the teacher: technology strengthens the educator, but does not substitute for human interaction and the transmission of values. The child’s individuality and psychological safety take precedence over digital efficiency. A national cultural and linguistic framework is needed so that global algorithms trained on a different context do not erase the Kazakh language and culture. A phased, experimental approach: scaling only after pilots and independent evaluation, without coercive KPIs that provoke formalism and resistance. And a balance between freedom to experiment and common standards — schools need room to try, but within shared rules, ethics, and data protection.

If we reduce this logic to its core, in the AI transition the state manages not technologies, but the balance of three things that would otherwise develop out of sync: AI technologies themselves (they are racing ahead, driven by market forces and spontaneously), educational institutions (they lag behind and react after the fact), and people with their values (this dimension is the least protected today). The system becomes sustainable only when these three lines move in sync. Keeping them together is the real task.

It sounds less dramatic than “we’ll roll out AI in every school by the end of the year.” But it is precisely this approach that distinguishes countries that will make it into the future from those that will merely report on it in polished terms.

The Year of Digitalization and Artificial Intelligence gives Kazakhstan a rare chance — not to catch up, but to build from the outset a smart, manageable, and human-centered model of the school of the future. To make use of this chance, we need to honestly examine several things. This series is devoted to that task.

In the following materials, we will look at the big picture — what is happening with AI around the world and why this is not going to go away. We will examine the results of our own research, in which hundreds of students, parents, teachers, principals, and methodologists shared their views. We will study others’ mistakes — why no country has succeeded so far. We will ask an uncomfortable question: might AI intensify the school’s problems that we have not solved without it? We will unpack the institutional dilemma of data and platforms. And in the end, we will propose a framework — what a managed transition might look like.

Because the central question of our time is not whether AI can be introduced into schools. It can. The question is how, and to what end.