Kazakhstan’s AI strategy could gain an edge through education

Artificial intelligence is redrawing the map of education before governments fully understand the labour market it will create. China is closing thousands of university programs while expanding training in AI and other strategic technologies. American universities are creating specialist degrees and introducing AI across conventional disciplines. Russia is building industry-backed programs for advanced developers. Kazakhstan, meanwhile, is incorporating AI from primary school to university and preparing to open its first specialized AI research university.

These reforms reflect an understandable desire to prepare students for technological change. But they also raise a deeper question. If AI transforms tasks inside almost every profession, should governments simply replace supposedly outdated majors with new ones? Or should they reconsider what students learn within medicine, engineering, law, international relations and other established disciplines?

Some jobs will disappear, but more will change

The emerging evidence rejects both extremes in the debate about AI and employment. Artificial intelligence is neither about to eliminate most professions nor merely provide workers with another harmless productivity tool. Some occupations, particularly those dominated by standardized digital, administrative and clerical tasks, are likely to contract. In many others, AI will automate only part of the work, allowing the profession to survive but reducing the number of people needed.

The distinction matters because jobs are bundles of tasks. A position may involve collecting information, drafting documents, communicating with clients, interpreting ambiguous situations and accepting responsibility for a final decision. AI might perform the first two effectively while remaining unreliable or legally incapable of handling the rest. The outcome may be replacement, but it may also be compression: five employees using AI produce what previously required eight or ten.

The International Labour Organization estimates that one in four jobs worldwide has some exposure to generative AI, but considers transformation more likely than complete replacement because most occupations still contain tasks requiring human involvement. Exposure, however, is not the same as job loss; it measures the potential for technology to affect occupational tasks.

The World Economic Forum’s Future of Jobs Report 2025 illustrates both directions of change. Employers expect AI and information-processing technologies to create around 11 million positions globally by 2030 while displacing approximately 9 million. AI specialists, data specialists and software developers are expected to grow, while administrative, clerical and some routine knowledge-work roles decline.

Early employment data also point to a particular risk for graduates. An August 2026 revision of research by the Stanford Digital Economy Lab found no widespread, economy-wide displacement associated with AI in the United States. It did, however, find that employment among workers aged 22 to 25 in highly AI-exposed occupations was around 19% below where it would have been had it kept pace with less-exposed occupations. The adjustment appeared mainly through reduced recruitment rather than mass dismissal.

The researchers caution that the findings are descriptive and do not prove that AI caused the divergence. Nevertheless, they reveal a plausible new problem: AI may remove the entry-level tasks through which young lawyers, analysts, programmers and other graduates traditionally acquire experience. A profession can survive while becoming significantly harder to enter.

Different countries, different answers

China has adopted the most centralized response. According to Ministry of Education data reported by Xinhua, Chinese universities introduced around 10,200 undergraduate programs and cancelled or suspended approximately 12,200 between 2021 and 2025, adjusting more than 30% of the country’s program offerings. Reductions affected parts of the arts, humanities, foreign languages and management, while new programs appeared in AI, robotics, semiconductors and other fields aligned with national industrial priorities.

The restructuring is not driven by AI alone. It also reflects graduate unemployment and oversupply in certain disciplines. Nevertheless, China is treating higher education as part of economic and technological planning, redirecting capacity towards sectors Beijing considers strategically important. The risk is that programs designed around the job titles of 2026 may already be dated when their first students graduate several years later.

The United States offers a more decentralized model. Universities continue expanding specialist AI training, but some are also treating AI as a capability required across disciplines. Ohio State University’s AI Fluency initiative, for example, expects students to graduate fluent both in their chosen field and in applying AI within it. The model begins with common foundations and then asks students to evaluate AI’s limitations, critically assess its outputs and use it in discipline-specific contexts.

Russia has emphasized advanced technical training through cooperation between universities and technology companies. The AI360 initiative, developed by Yandex, Sber and several leading universities, is designed to prepare researchers and engineers capable of creating new AI models and architectures. It combines fundamental mathematics and machine learning with access to industry expertise and computing infrastructure.

China is therefore reallocating the degree portfolio, parts of the US system are embedding AI across existing education, and Russia is strengthening a specialist technical pipeline. These approaches reveal the real choice facing Kazakhstan: not whether to teach AI, but how to divide resources between those who will build it and those who will apply it.

Kazakhstan has already made both bets

Kazakhstan is moving along both tracks. In schools, “Digital Literacy” and “Informatics” are being updated to include artificial intelligence. Revised programs contain 43 AI-related learning objectives, progressing from digital safety and responsible use to machine learning, neural networks and data analysis.

Higher education is expanding even more rapidly. According to the Ministry of Science and Higher Education, 95 universities have introduced AI disciplines, while 30 institutions offer 42 specialized programs. Speaking at a Government meeting on August 18, Science and Higher Education Minister Sayasat Nurbek said the new Qazaq Research University of Artificial Intelligence would begin training students this year. Its programs will include AI+X disciplines applying artificial intelligence in different sectors of the economy.

The initiatives promoted by Nurbek’s ministry show that Kazakhstan has already answered the initial question of whether AI should enter higher education. That question has largely been settled. The next challenge is more difficult: ensuring that the new courses and programs develop distinct, verifiable capabilities rather than merely placing an AI label on curricula that remain largely unchanged.

Universities teach nouns; AI changes verbs

Universities organize education vertically through professional and disciplinary nouns: lawyer, engineer, doctor, teacher, diplomat. AI enters the labour market horizontally through verbs: searching, translating, drafting, modelling, diagnosing and predicting. Creating another vertical category called “AI specialist” is necessary for producing developers, but it does not prepare everyone else for changes occurring inside their professions.

International relations offers one example. The wrong question is whether a student should study AI or international relations. Most institutions do not need a new bachelor’s degree in “AI diplomacy.” They need international relations graduates who understand history, political theory, economics, languages and regional affairs, but can also use AI for multilingual document analysis, open-source intelligence, sanctions monitoring, scenario development and disinformation detection.

Domain knowledge becomes more important, not less, when machines can generate convincing but unreliable answers. A graduate who can produce an AI summary but cannot identify a historical distortion, an implausible political assumption or a mistranslated diplomatic term is not AI-literate. That graduate is merely dependent on AI.

The same principle applies elsewhere. Kazakhstan needs geologists capable of applying machine learning to exploration data, doctors who can evaluate clinical decision-support systems and civil servants who can audit algorithmic recommendations. Most will not build foundation models. Their value will come from combining professional expertise with the ability to use, question and supervise AI.

This requires three levels of education: universal AI literacy, discipline-specific application and advanced technical specialization. A future historian does not need the same machine-learning depth as an AI engineer, just as an AI engineer does not require the same archival training as a historian. Both, however, need enough understanding to work together and recognize the limits of their competence.

What Kazakhstan should - and should not - do

First, Kazakhstan should preserve the distinction between specialist AI education and AI+X. The country needs dedicated programs for researchers and engineers, but the larger economic return is likely to come from applying AI in sectors where it has real institutions, data and demand: mining, energy, agriculture, logistics, public administration and Kazakh-language technologies. Educational priorities should follow credible sectoral opportunities rather than an ambition to produce the largest possible number of generically defined AI graduates.

Second, university requirements should be based on outcomes rather than course counts. Programs should demonstrate that graduates can select appropriate tools, evaluate outputs, identify errors and apply AI to authentic professional problems. That also requires investment in faculty and computing access. Without it, AI reform risks widening the divide between leading institutions in Astana and Almaty and regional universities.

Third, education policy must address the narrowing entry-level ladder. If AI absorbs junior tasks, universities and employers will need alternative ways for graduates to acquire experience through apprenticeships, supervised projects and employer-supported placements. At the same time, workers in declining occupations will require realistic transition support. Not everyone affected by automation can or should be retrained as a programmer.

Finally, Kazakhstan should avoid indiscriminately reducing the humanities and social sciences. A country seeking sovereign Kazakh-language AI needs linguists, historians, lawyers and cultural specialists capable of producing and evaluating the material on which such systems depend. Communication, ethical reasoning and contextual judgment are not decorative additions to an AI strategy; they are among the capabilities machines struggle to exercise reliably.

The educational challenge is therefore larger than choosing which majors to open or close. Kazakhstan needs a relatively small group capable of building advanced AI, a much larger group able to apply it within existing professions, and credible support for people whose former work is automated or compressed.

Success should not be measured by how many university programs contain the words “artificial intelligence.” It should be measured by whether Kazakhstan’s engineers, teachers, doctors, diplomats and civil servants can use AI without surrendering the professional judgment that makes their work valuable.

Earlier, Qazinform reported that a U.S. congressional advisory commission said China’s efforts to treat data as an economic asset could strengthen its competitive position in artificial intelligence, particularly in industrial and embodied AI.