Artificial intelligence (AI) adoption is lagging its potential to boost productivity across Europe and Central Asia, with fewer than one in 10 firms using the technology in their operations despite growing exposure among workers, according to the World Bank.
About 9.4% of firms with at least 10 employees in western parts of the region used AI in 2025, compared with almost 20% across the European Union, the bank said in its October 2026 Europe and Central Asia Economic Update. The report warns that businesses are struggling to move beyond basic applications that save time but have yet to deliver measurable productivity improvements.
The gap risks limiting the region’s ability to raise output as productivity growth slows and ageing populations reduce the supply of workers. The bank argues that the principal challenge is not simply access to AI technology, but whether companies have the management capabilities, skilled workers, usable data and computing infrastructure required to integrate it into core business operations.
Around one in five workers in the region holds a job meaningfully exposed to generative AI, with exposure highest among people with tertiary education and younger workers. The bank cautions, however, that exposure does not necessarily translate into job losses or the creation of new positions.
The findings highlight a gap between the rapid development of AI tools and the ability of businesses to use them to improve economic performance.
Businesses struggle to translate AI into productivity
Many companies in the region use AI for customer-service chatbots, internal knowledge assistance, administrative support and document preparation. In some cases, the technology has cut the time required for internal documentation from two weeks to a few hours, according to the report.
Yet these efficiency gains have generally not translated into measurable productivity improvements, suggesting that basic adoption alone may not deliver the broader economic benefits associated with the technology.
A smaller group of businesses has integrated AI into core operations, including financial institutions using it to assess loan applications, retailers making stock-replenishment decisions and energy companies optimising inputs.
These firms have generally addressed problems involving skills and data integration while focusing on applications with clear commercial returns, the bank said.
Management quality is among the strongest predictors of adoption. Companies need managers capable of identifying valuable applications, redesigning workflows and guiding employees through organisational changes, rather than simply purchasing access to AI models.
The report also identifies a shortage of business-oriented engineers and managers who can combine technical knowledge with operational decision-making. Although the region has substantial numbers of science, technology, engineering and mathematics graduates, gaps in foundational literacy and numeracy constrain adoption in parts of Central Asia.
Computing infrastructure remains a major constraint
The region’s limited computing capacity is another obstacle to wider AI deployment.
Data-centre capacity stood at about 2.1 megawatts per million people in 2025, compared with 31.6 megawatts in high-income economies outside Europe and Central Asia, according to the report. The region also hosts less than 1% of the world’s top 500 supercomputers.
Insufficient computing infrastructure can make intensive AI applications harder to deploy, while reliance on foreign providers raises concerns about costs and dependence on external technology companies.
The bank says the cost of capital and institutional credibility are major factors determining where data-centre investments are made. Cheap electricity alone has not been enough to attract large-scale computing infrastructure to countries that lack the financing conditions and investor confidence needed for capital-intensive projects.
Businesses also face difficulties integrating data held in separate systems, limiting the usefulness of AI models. Concerns about confidential information leaking into third-party systems, inaccurate outputs, uncertain returns and changing regulations further discourage companies from moving beyond experimentation.
Job losses remain limited, but risks persist
The report finds that early evidence of AI’s labour-market effects points more to changes in hiring and the tasks employees perform than to widespread dismissals.
In Romania, almost 91% of firms adopting AI reported no changes in hiring or layoffs as a result of the technology. In Türkiye, 2% of firms reported firing workers because of AI adoption, while 7% said they had hired workers for that reason.
Interviews with businesses also suggest that some employers have reduced hiring or reassigned existing workers rather than dismissing them.
However, the bank warns that limited aggregate employment effects can conceal significant losses in particular occupations, companies and regions. Knowledge-intensive business services could face substantial changes as routine tasks become automated and human work shifts towards supervision, quality assurance and handling exceptions.
The pace of disruption will depend partly on how quickly businesses adopt AI and how easily workers can move into complementary roles.
The bank argues that governments should establish retraining programmes, income support and effective labour-market institutions before major disruptions occur, rather than waiting until workers lose their jobs.
Policy focus shifts to practical adoption
The report recommends that governments prioritise the adoption and adaptation of existing AI technologies over costly efforts to develop advanced national models without a clear commercial purpose.
It calls for greater access to affordable computing resources, stronger foundational and digital skills, better management training and improved data integration. Governments can also support adoption by acting as customers for locally developed AI products and helping businesses overcome the initial costs of using the technology.
The bank says smaller AI models adapted to local languages, legal systems and business needs could be more useful than expensive efforts to compete directly with leading global developers.
For businesses, the priority is to identify applications that solve operational problems and generate measurable returns, rather than treating AI adoption as an objective in itself.
The report’s central warning is that the region faces a greater risk of adopting too little AI than too much. Without improvements in business capabilities, infrastructure and skills, the technology’s potential to offset labour shortages and revive productivity growth may remain largely unrealised.
