AI to the rescue: How the countries of Europe and Central Asia can grow more prosperous as they grow older

Much of developing Europe and Central Asia is ageing fast, and its workforce is starting to shrink. This column argues that artificial intelligence can partly offset this demographic challenge by (1) augmenting the skills of many in the workforce while automating tasks for which workers are increasingly scarce; (2) lowering the cost of expertise and specialisation, thus bringing in people currently not in the labour force; and (3) changing the way work is done, thereby expanding productive working lives. strong foundational and managerial skills, compute infrastructure, and the supportive business environment are necessary to encourage more firms to productively adopt and adapt AI.

Much of the public debate on artificial intelligence (AI) turns on a single fear: that the technology will destroy jobs faster than the economy can create new ones. For most countries in Europe and Central Asia,¹ the more pressing problem is the opposite. Populations are ageing fast and, in many countries, shrinking. Firms report shortages of workers, specialised skills, and experienced managers. The most pressing challenge is not too few jobs, but too few skilled workers.

AI can help partly offset that demographic decline. AI will not reverse ageing, at least not in the medium term: no technology can replace the millions of workers who will retire over the next quarter century. But by raising productivity growth – including by automating some jobs, making others more productive and rewarding, and expanding what kinds of work are possible – AI can help the region produce more with fewer people. Whether that happens and over what period depends less on access to the technology than on the capacity of workers and firms to adopt it and adapt it to their needs.

A shrinking workforce – and an underused one

The labour force of Europe and Central Asia is shrinking due to falling fertility rates and strong outward migration. Fertility rates are below replacement in most of the region, and the working-age population is projected to fall by about 17 million by 2050, faster than most other developing regions (World Bank 2025). The old-age dependency ratio is projected to reach 40% by 2050 – nearly double of what it is today – while the median age in the region is likely to rise from 36 years in 2023 to about 49 years by 2060. 

As a result, labour shortages have become structural rather than cyclical. With the catch-up gains of the transition from plan to market largely exhausted, future growth will have to come increasingly from stronger productivity (Iacovone et al. 2025).

The shortages coexist with a large reserve of underused talent. Female labour force participation is low in several countries (World Bank 2025). The share of youth not in employment, education, or training (NEET) is more than double that in the EU (World Bank 2025, Eurostat 2025b). Many workers leave the labour force earlier than in advanced economies (Eurostat 2025a).

These reserves matter for how we think about AI. The region has three levers to respond to demographic decline: make each worker more productive, bring more people into work, and make longer working lives more rewarding. AI can help pull all three.

AI to the rescue over the medium term

AI can raise productivity growth. OECD estimates suggest that AI can boost productivity growth by 0.4–1.3 percentage points a year in G7 countries (Filippucci et al. 2025). Simulations across Europe and Central Asia indicate that under an optimistic scenario, relative to a no-AI baseline, AI would add little to growth at first – about 0.1 percentage point in 2027 – but the boost would build quickly as adoption spreads, reaching about 2 percentage points a year by 2033–34 (World Bank 2026, Aksoy and Lokshin, forthcoming). Larger gains are likely in economies such as Poland’s, one of the countries with the highest AI exposure in the region (Figure 1).

Figure 1 AI solely as a productivity shock: AI can raise productivity growth

Source: World Bank (2026).

As labour productivity increases, so will the demand for labour where AI complements workers’ skills. Richmond (2026) examines whether AI can do a meaningful share of a worker’s tasks, whether a person is still central to the work, and whether cheaper services would raise demand enough to absorb the productivity gains. This research suggests the following paths:

AI can do some of the work. About 11% of jobs in Bulgaria, Croatia, Poland, and Romania could be automated (Richmond 2026). These jobs are not ‘messy’ and AI can do most of them (Garicano et al. 2026). Think of document processing, answering standard customer queries, bookkeeping, or basic coding. The result is that firms will need fewer workers for the same output. In most economies, this possibility raises fears of unemployment. However, in a region where vacancies already go unfilled, this should help ease shortages instead. So far, the evidence shows very limited displacement, either in the US, the leading AI adopter (Brynjolfsson et al. 2026, Hartley et al. 2026), or in Europe and Central Asia (World Bank 2026). AI is adding and removing jobs at the same time (Verschuere and Cameron 2026), and some adjustment is under way on the margins (Bonin et al. 2026).

AI changes how work is done. Most jobs bundle tasks that depend on judgement, context, and dealing with people. In these jobs, AI keeps the workers and makes them better (Garicano et al. 2026). The gain is more output from each worker, which raises demand for labour. Richmond estimates the share of such jobs to be about 27% in the EU, with similar numbers in Europe and Central Asia. Generative AI tools can help less experienced workers most, bringing them closer to the performance of experienced colleagues (Brynjolfsson et al. 2025). That lowers the barrier for young people entering work. Used well, AI can also help older workers stay productive for longer. Exposure can mean more work rather than less (Jiang et al. 2025).

AI can change what is possible. AI is creating new activities: personalised services at scale, analysis that was previously too costly to undertake, and new products and business models. AI can also widen who takes part in the economy by creating AI-supported roles with flexible hours that can help older workers stay in work longer. There is some evidence that the occupations most exposed to AI are also the most age-friendly, but these tend to be jobs where AI complements rather than replaces them (Pizzinelli and Tavares 2025, André and Schief 2026). These are the gains most likely to matter in the long run – and the ones in a region where firm capabilities and the range of products and services lag the frontier.

The main gap is in skills and capabilities, not in access to the technology

Fewer than one in ten firms in Europe and Central Asia use AI, half the share in the EU (Figure 2). In-depth interviews with top managers in five European and Central Asian countries during the summer 2026 reveal that for the most part, the use of AI is shallow, consistent with other quantitative findings (World Bank 2026, Davies et al. 2026). Use is mostly internally facing, focused on cost-saving rather than client-facing and revenue-generating. Data-rich sectors such as telecommunications and finance lead in AI use, while firms in retail and low-tech manufacturing, with slim margins and uncertain returns, hold back. 

Figure 2 AI adoption by firms is far lower in Europe and Central Asia than in the EU

Source: World Bank (2026).

Very few firms could quantify the returns of AI. Firms in the region often lack access to structured data and managerial capabilities to redesign workflows around AI. Business-oriented engineers and managers are scarce, and basic digital skills are thin. For example, only 46% of adults can send an email with an attachment as compared to Germany with 80% (World Bank 2026).

Unlike earlier general-purpose technologies, AI does not require large up-front investment in physical capital, and many foundations are already in place: connectivity is widespread, mobile access is nearly universal, and energy is relatively cheap in parts of the region. In addition, STEM graduates are plentiful, and firms are used to absorbing imported technologies (World Bank 2026). What is missing is how to utilise these factors productively. The binding constraints are foundational skill and managerial capabilities, not technology.

Policy priorities

With slow economic growth and demographic challenges, the region cannot afford to watch from the sidelines. Five priorities follow.

• Build foundational skills and managerial capabilities. The foundation of AI adoption is not AI itself, but the ability to use it productively. Literacy, numeracy, judgement, and strong management skills – which are weak and declining in many countries – must be nurtured alongside AI.
• Protect workers, not jobs. Some jobs will change or disappear. Public policy should help workers transition into more productive ones through education, lifelong learning, and for those affected in the short run, through reskilling.
• De-risk the policy and business environment. Cheap energy alone does not attract investors in data centres; institutional credibility does. Predictable rules lower the cost of capital, which in turn shapes how much computing capacity is created or how many start-ups can scale up. Start-ups need growth-stage risk capital and insolvency regimes that do not punish failure and do not reward incumbents that hamper competition.
• Enable AI through sensible regulation. Governments should enable AI by creating conditions for firms and public institutions to use the technology safely, effectively, and at scale. Regulation should be proportional to the risk and not impede adoption and adaptation. Instead, countries should enable experimentation and improve data quality and interoperability.
• Let the private sector lead. Productivity gains, innovations, and adaptation will be driven by firms. Governments need to keep markets open and competitive, allowing the private sector to identify the opportunities and take the risk.

Europe and Central Asia will not win the race to build frontier AI, and it does not need to. Its challenge is different: producing more with fewer people by adopting and adapting AI. AI can help meet that challenge only in part, and only for a time – but that time is valuable. The countries that use it well will be those that combine AI adoption with the skills, infrastructure, and business dynamism to absorb it, and which make a determined effort to bring their young and older people into productive work.

Source : VOXeu