Artificial intelligence is spreading quickly among European firms. Using new surveys from Germany, Italy, and Spain, this column shows that adoption rates of generative AI differ sharply across and within countries. Uptake is strongly skewed towards larger, more productive, service sector firms, while German manufacturing shows substantial adoption too. There are strong complementarities between AI, cloud computing, and robotics, and early experimentation appears to be an important stepping stone towards more intensive use. Most adopters still use AI experimentally and mainly to upgrade automated or business support processes, and tend to see AI as reshaping tasks while hardly affecting overall employment.
The rapid diffusion of advanced digital technologies – most notably, cloud computing and artificial intelligence (AI) – is reshaping business operations across sectors and countries. AI, in particular, is increasingly recognised as a general purpose technology (GPT), characterised as such by its broad applicability, continuous improvement, and its ability to spur innovation across industries (Bresnahan and Trajtenberg 1995, Cockburn et al. 2019). Despite AI’s prominent role in current economic and policy debates, comprehensive empirical analyses of its diffusion patterns across firms remain scarce – especially evidence based on firm-level surveys that are both representative of the full population of firms and harmonised across countries. Understanding these patterns matters, as macro-level assessments of AI’s impact on productivity growth depend critically on the pace and breadth of firm-level adoption (Calvino and Fontanelli 2023, Filippucci et al. 2024, Cullen et al. 2025). Yet, existing survey data on AI adoption differ substantially in both the questions they pose and in the composition of firms, creating significant challenges for cross-country comparability.
Fast-rising AI adoption across European firms
This column draws on harmonised business surveys on AI use conducted by the Deutsche Bundesbank, Banca d’Italia, and Banco de España in 2024 and 2025 (Deutsche Bundesbank 2024, Banca d’Italia 2025, Banco de España 2025, Bencivelli et al. 2025). Using a common questionnaire, the surveys measure the adoption of AI technologies, with a particular focus on generative AI (GenAI), among private non-financial firms in manufacturing and services in Germany, Italy, and Spain. The survey samples are aligned with respect to firm sector and size coverage, considering only firms with at least 20 employees. The results reveal marked cross-country differences: in 2024, only 13% of Italian firms reported using AI (either predictive or generative) compared with 47% in Germany and 31% in Spain. As for GenAI, only 5% of Italian firms reported using it compared with 33% in Germany and 26% in Spain (Figure 1). Over the subsequent 12 months, GenAI adoption almost doubled in Germany (to 58%) and rose even more sharply in Italy (to 24%), pointing to very rapid diffusion. The harmonised surveys also provide insights into how intensively GenAI is used in business activities – an important aspect that most surveys on GenAI adoption do not capture. Cross-country differences in the rate of adoption are mostly related to experimental and limited usage, while only a small share of firms (less than 4% in all countries) makes intensive use of this technology. This suggests that, despite growing interest and widespread pilots, GenAI has so far been integrated into the core business processes of only a narrow group of frontrunner firms.
Figure 1 Adoption of generative AI
Note: The figure covers firms in industry (excluding construction) and in the non-financial private services sector with at least 20 employees. Generative AI is shown by intensity. For Germany and Italy, the total for 2024 corresponds to the share of firms reporting intensive, limited, or experimental AI adoption (excluding firms that report using only predictive AI) in April-June 2024 (Germany) and February-May 2024 (Italy). Data are weighted using firm weights.
Sources: Bundesbank Online Panel – Firms (BOP-F), April-June 2025; Bank of Italy’s Survey of Industrial and Service Firms (INVIND), February-May 2025; Bank of Spain Business Activity Survey (EBAE), November 2024.
Rates of adoption of GenAI are consistently higher in Germany than in the other two countries, regardless of most firm-level characteristics. Yet, the cross-country evidence shows some common patterns. In line with previous findings (Calvino and Fontanelli 2023), we observe that adoption increases with firm size, measured by the number of employees (Figure 2), and that the relative increase for the largest firms is particularly pronounced in Italy and Spain. Similarities are also present for sector-specific adoption rates. GenAI is most prevalent among service-sector firms in all three countries, particularly in logistics, telecommunications, and other services, a category including real estate, professional and support service activities. The manufacturing sector in Germany, however, stands out: adoption rates are only about a quarter lower than in the leading sector. Adoption among Italian and Spanish manufacturing firms is, by contrast, considerably lower than in the service sectors. Finally, with respect to firm productivity (measured by turnover per employee), adoption rates increase for firms with above-medium productivity in all three countries under consideration. Taken together, these patterns suggest that GenAI diffusion is currently skewed towards larger, more productive firms and service activities, with German manufacturing representing an important exception.
Beyond size, sector and productivity, GenAI adoption strongly correlates with the use of other advanced technologies and prior AI experimentation. Firms already leveraging tools like cloud computing or robotics are significantly more likely to adopt GenAI, underscoring that digital maturity is a key driver of innovation readiness. Complementarities with cloud computing and robotics appear particularly relevant, as these technologies often provide the infrastructure and operational flexibility needed for GenAI integration. GenAI usage in 2025, as well as the intensity of use, is also higher among Italian and German firms that experimented with AI technology in 2024 (either predictive or generative), suggesting that early experimentation tends to pave the way for more systematic and intensive adoption rather than remaining at the pilot stage. This is consistent with a gradual, path-dependent diffusion process in which firms build digital and organisational capabilities over time.
Figure 2 Adoption of generative AI by firm size and sector
Note: The figure covers firms in industry (excluding construction) and in the non-financial private services sector with at least 20 employees. The share of firms reporting intensive, limited, or experimental AI adoption is shown by firm class size (left panel) and by sector (right panel). Data are weighted using firm weights. 1 Comprises NACE Section L (Real estate activities), Section M (Professional, scientific and technical activities), and Section N (Administrative support and support service activities).
Sources: Bundesbank Online Panel – Firms (BOP-F), April-June 2025; Bank of Italy’s Survey of Industrial and Service Firms (INVIND), February-May 2025; Bank of Spain Business Activity Survey (EBAE), November 2024.
What drives businesses to embrace AI?
In 2024, most firms identified the enhancement of already automated processes or the improvement of business support processes as a relevant objective for adopting AI (Figure 3). By contrast, expanding the product range or developing new services consistently ranks lower than process-related improvements. Evidence from Spanish firms reinforces this pattern: when asked to identify the single most important objective for AI use, firms most frequently cited upgrading already automated processes, followed by task automation and improvements to support functions. This pattern suggests that, for most firms, AI is currently viewed primarily as a tool for efficiency gains and process optimization rather than for product diversification.
An additional module, administered exclusively to Italian and Spanish firms, sheds light on perceptions regarding the impact of AI on the labour market. In Spain, about 80% of firms in 2024 believed that AI would not affect their overall level of employment, but those that were already using it tended to expect a positive impact on job creation within their organisation. Similarly, in Italy most firms using GenAI expect that it would create new job opportunities and lead to a redistribution of tasks within the firm, while a smaller share foresees a change in employment. This is consistent with the findings of Albanesi et al. (2023), Ilzetzki et al. (2023), and Arias et al. (2025). Taken together, these results suggest that, at the current stage of diffusion, firms mainly view AI as a technology that reshapes tasks and supports employment rather than as a driver of net job cuts.
Figure 3 Objectives for AI use
Note: The figure covers firms in industry (excluding construction) and in the non-financial private services sector with at least 20 employees that reported using generative and/or predictive AI in 2024. The share of these firms is shown that rate each objective for AI use as somewhat or very relevant, not very relevant, or not relevant. Data are weighted using firm weights.
Sources: Bundesbank Online Panel – Firms (BOP-F), April-June 2025; Bank of Italy’s Survey of Industrial and Service Firms (INVIND), February-May 2025; Bank of Spain Business Activity Survey (EBAE), November 2024.
Conclusion
Digital technologies are spreading at an accelerating pace. While cloud computing has become a standard feature of firms’ technological infrastructure, the diffusion of GenAI remains uneven across sectors and countries. Drawing on harmonised business surveys on AI use conducted in Germany, Italy, and Spain, this study finds that GenAI adoption increases with firm size and is more common in services, though in Germany, manufacturing firms also show substantial adoption. Adoption is also positively associated with firms’ productivity. Our evidence also highlights complementarities with other advanced technologies, such as cloud computing and robotics, and shows that early experimentation is a key stepping stone towards more intensive use. At this still early stage, intensive use of GenAI is more concentrated among digitally advanced firms. Most firms rely on AI technologies mainly to upgrade and improve business processes and perceive them as reshaping tasks while supporting, or at least not reducing, overall employment.
Source : VOXeu
