Fix education to make AI work across Europe and Central Asia

The countries of Europe and Central Asia (ECA) cannot afford to miss the opportunity that artificial intelligence (AI) offers, but they will not seize it without the right skills. This is the one of the main messages of the latest World Bank Economic Update on Europe and Central Asia. Few developing regions have as much to gain from AI as ECA. Productivity and business dynamism have been slowing for two decades. The working-age population has already shrunk in most countries and is on track to fall sharply across much of the region by 2060. On top of that, the quality of education has declined in most ECA countries over the last decade. At present, AI's potential in ECA far outpaces the region’s ability to use it. Fewer than one in ten firms report using AI in their operations, roughly half the average in the EU. Worker exposure to AI is equally modest. When firms are asked what holds them back from adopting AI, people and skills top the list. Catching up with peers at similar levels of development, let alone with leading countries, depends less on computer chips or access to advanced models than on people. What makes AI work Most firms in the region use AI in shallow ways that rarely translate into real productivity gains. The exceptions — banks screening loan applications, retailers managing stock across suppliers, energy producers optimizing inputs — have something in common: they invested in skills and data systems, and their managers focused on results, not on adopting AI for its own sake. Four ingredients stand out: • Workers with solid foundations. A sizable part of ECA's workforce lacks the foundational reading, numeracy, and reasoning skills that AI adoption requires. • Vocational education that delivers. About 45 percent of upper-secondary students in ECA are enrolled in vocational programs, more than twice the global average of 22 percent, and above 70 percent in some countries. • Managers who can redesign production processes. Managerial quality is among the strongest predictors of technology adoption. Research suggests that about a quarter of productivity gaps between and within countries reflect management practices. Someone must identify where AI adds value, redesign workflows, and guide staff through the change. • Workers who keep learning. Lifelong learning supported by employers and governments and a willingness to experiment are essential for staying relevant in a shifting labor market. The report’s conclusion is clear: education and training systems in ECA need serious reform. What PISA 2025 tells us The 2025 OECD Programme for International Student Assessment (PISA) results show that learning quality in ECA has slipped further, especially on the very skills that AI demands: critical understanding, creativity, and judgment. ECA’s average score across math, reading, and science fell from about 421 points in 2018 to about 406 in 2025, roughly equivalent to losing a full almost a year of schooling. The gap with the EU did narrow, from about three and a half years of learning to two and a half. But this is mostly because EU scores fell by more than ECA scores - not a reason to feel better. The more worrying picture is how many students are left behind. In most ECA countries, at least half of 15-year-olds score below the minimum proficiency level compared with about one-third in the EU. In Kosovo, Armenia, Kyrgyzstan, and Tajikistan, the share is four in five or more. At the top end, 4.4 percent of EU students reach the highest proficiency levels. In ECA, only Türkiye comes close; in most countries the share is below 1.5 percent. Student attitudes are also moving in a troubling direction. The share of students who find learning new things interesting fell in every ECA country between 2022 and 2025 in every ECA country. Students who skip school also score much lower – up to two years of learning behind in some countries. Why weak education holds AI adoption and adaptation Today’s 15-year-olds are the workers, managers, and entrepreneurs who will need to make AI work in the 2030s. If most of them lack foundational skills, AI use in the region will stay shallow. AI rewards people who can critically assess its outputs, not just accept them. PISA 2025 illustrates this clearly: students who check sources and apply critical thinking when using AI score far higher than those who take AI outputs at face value. The gap is about three years of schooling. Yet only 29 percent of ECA students who use AI take that careful approach, compared with 45 percent in the EU. Using AI simply to summarize texts brings no measurable learning benefit. The risk is a two-speed region. A small, well-educated elite would capture most of AI's gains, while the majority, whose jobs are least exposed today, fall behind as AI spreads into more types of work. This is a problem that no amount of spending later can fully fix. Fixing education today is the priority The good news is that the decline is reversible. Türkiye, Montenegro, Georgia, and Uzbekistan are among the six countries with the largest PISA improvements worldwide. Türkiye paired digital infrastructure investment with a curriculum shift away from memorization toward real-world problem solving. Uzbekistan set a national target of reaching the PISA top 30, reformed its curriculum and teacher training, banned phones in schools, and expanded early childhood education coverage from 30 to 75 percent. These priorities emerge for ECA countries: • Secure foundational learning early. Scale up proven, cost-effective interventions in early childhood education so fewer children fall below the minimum learning threshold. • Teach students to question AI, not just use it blindly. Embed source-checking and evidence-based reasoning across curricula. Students need the skills to tackle challenging topics themselves, rather than offload their thinking to AI tools. • Reduce digital distraction. Set clear school-level rules on device use to protect face-to-face lesson time and maintain students’ focus. • Cultivate curiosity and perseverance. Active, student-centered teaching develops self-directed learners who can adapt as AI reshapes the labor market. • Retrain today's workers and managers. Upskilling programs should target foundational, digital, and managerial skills for AI-complementary work. The main risk for ECA over the next decade is too little AI adoption, not too much. Compute, data, and investment matter greatly, and countries have to make progress on all three. But education is decisive. It needs to equip young people with the skills to use AI as a tool that amplifies their thinking, not a substitute for it. Without fixing education, countries across ECA will keep using AI in shallow ways, missing its real potential to raise productivity and offset, at least in part, the region’s demographic challenges.