Summary

• In the ECB's late-2025 survey, 64% of businesses reported experimental or moderate AI use, compared with 7% reporting significant use.
• The proportion of workers using AI rose from 26% in 2024 to 52% in 2026, with users reporting median savings of three hours weekly.
• Companies expect spending on software, training, and data infrastructure to increase, although financing and implementation remain uneven.

European businesses are adopting artificial intelligence more widely, but research presented by the European Central Bank shows that intensive use remains concentrated among a relatively small group of companies, even as workplace adoption and spending on digital technology increase.

The findings, presented by ECB Executive Board member Philip R. Lane on 30 October, combine business surveys, employee research, and investment data to examine how AI is affecting the euro area economy. They show considerable interest in the technology alongside substantial differences between experimentation and regular integration into business operations.

In the ECB’s Survey on the Access to Finance of Enterprises, conducted during the fourth quarter of 2025, 64% of participating businesses reported experimental or moderate AI usage. Only 7% described their use as significant, indicating that the spread of AI tools has outpaced their adoption as a substantial part of everyday operations.

Improving existing business processes was the leading motivation among companies using AI, cited by 62% of respondents. Innovation, research and development, or extending product ranges was identified by 23%, while 15% cited labour-cost reductions.

Workplace adoption rises across Europe

The ECB’s Consumer Expectations Survey found that 52% of employed respondents were using AI at work in 2026, compared with 26% in 2024. Younger and more highly educated workers reported higher adoption, while usage also varied according to the activities undertaken within different occupations.

Among workers using AI, the median respondent estimated saving three hours each week, equivalent to approximately 7.7% of median working time. Those results describe perceived individual savings rather than independently measured increases in organisational output.

Software development and debugging produced some of the largest reported benefits, averaging six hours weekly for workers using AI on those activities. However, approximately 15% of workers reported using AI for coding, while more common applications included researching information, preparing written material, and editing documents.

These differences influence how productivity gains appear within organisations. A worker may complete a particular task more quickly without reducing the total time spent on a wider process, especially where colleagues must still review the result or complete subsequent activities.

Businesses also need to account for the quality of work produced. Generating a document more quickly may create a benefit when the result is accurate and useful, while additional checking or correction can absorb some of the time initially saved.

The ECB’s survey indicates that workplace use is expanding, but the reported benefits remain uneven. Comparisons with other European research also suggest that estimated time savings depend partly on how usage is measured and which workers are included.

Investment increasingly includes supporting capabilities

Digital investment in the euro area has expanded by approximately 73% in real terms since 2014, according to ECB calculations. Its share of GDP rose from around 2.3% to an estimated 3.3% by the first half of 2026, with recent figures partly inferred from digital services production data.

Much of that investment has involved intangible assets, including software, databases, and research and development, while data centre construction accounts for a comparatively small part of the aggregate measure.

The ECB’s Corporate Telephone Survey found that businesses expected AI to add around 10% to their capital expenditure plans during 2026. The figure reflects intended expenditure rather than an independently measured increase in completed investments.

Among companies planning AI-related investment over the next 12 months, 49% expected to spend on technology and tools, 46% on employee training, and 40% on data and infrastructure. Only 12% planned to recruit AI specialists.

Respondents could select several investment categories, so these percentages overlap. They nevertheless indicate that companies increasingly recognise the need to support AI tools through training and improvements to the technology environments in which they operate.

That combination can be expensive because enterprise deployment often requires more than purchasing access to a model. Businesses may need to prepare data, connect software systems, establish access controls, and change procedures before employees can use an AI application consistently.

Financing constraints influence deployment

The ECB’s analysis distinguishes investment in physical infrastructure from expenditure on capabilities that cannot readily be used as loan collateral. Data centres and computing equipment provide tangible assets, while employee training, software integration, and organisational knowledge are harder for lenders to value independently.

Businesses may therefore find it easier to finance certain forms of infrastructure than the operational improvements required to obtain commercial value from them.

The distinction is particularly relevant to smaller organisations, where investment capacity may be limited and specialist technology teams are less common. A company can have access to general-purpose AI software while still lacking the skills or data infrastructure needed to incorporate it into routine operations.

The ECB also examined the financing environment surrounding the international AI infrastructure boom. European investors are participating in technology-related equity and debt markets, while banks are responding to demand for data centre and computing investment.

This creates several channels through which AI influences the European economy. Businesses purchasing software contribute to demand for digital services, infrastructure developers require financing for physical installations, and investors gain exposure to companies building computing capacity in other markets.

Although these activities are connected, they do not necessarily produce equivalent benefits for European companies adopting the technology. The location of infrastructure investment and the ownership of software platforms can influence where economic value is captured.

Competition affects investment decisions

In a field experiment using the late-2025 business survey, the ECB examined whether companies changed their AI investment plans after receiving information about the behaviour of competitors.

Many businesses underestimated the proportion of domestic and foreign competitors investing in AI. When researchers provided more accurate information, participating companies revised their expectations, although their subsequent investment intentions differed according to where the competitors were located.

Information about domestic competitors had a measurable effect. The ECB estimated that a one-percentage-point increase in the expected share of domestic businesses investing in AI raised a company’s expected AI investment rate by approximately 0.6 percentage points.

Comparable information about foreign competitors did not produce the same response, suggesting that businesses may be more influenced by technological changes occurring within their immediate competitive environment.

These results concern expectations and planned investment rather than completed purchases or demonstrated improvements in productivity. They nevertheless indicate that companies’ decisions can depend partly on perceptions of competitor activity, alongside their own operational requirements.

Training remains central to deeper adoption

Employee research identified several reasons for not using AI. Around one-third of non-users considered it irrelevant to their current tasks, while others preferred established methods, questioned reliability, or lacked access through their employer.

Interest also varied considerably. The ECB found that 41% of workers expressed no interest in using AI, indicating that wider availability of tools will not necessarily produce universal adoption.

Training and clearer guidance could influence usage where employees have relevant tasks but lack the knowledge or confidence to apply the technology. Companies investing in those capabilities will still need to identify suitable applications and assess whether the benefits justify the operational changes required.

The ECB’s evidence therefore describes an economy moving from widespread experimentation towards selective implementation. Growth in software usage and investment is already visible, while the effects on organisational productivity, employment, and commercial returns will become clearer as businesses integrate the technology more extensively.