Small EU firms nearly tripled AI use, but the gap with large firms grew
In 2025, 17.0% of small EU enterprises used AI, up from 6.4% in 2023, against 55.0% of large ones. The EU size gap widened to 38 points and grew in 25 of 27 member states. Lack of expertise, not cost, is the main barrier.
Small EU firms using AI
Enterprises with 10 to 49 persons employed; Eurostat, 2023 to 2025.
Gap between large and small firms
55.0% of large against 17.0% of small enterprises in 2025, up from 24.1 points in 2023.
Gap between countries for small firms
Denmark (37.5%) against Romania (4.1%); Eurostat, 2025.
Cite lack of expertise
Small enterprises that considered AI but did not use it; EU-27, 2025.
Findings for newsrooms
- 17.0% of small EU enterprises (10 to 49 persons employed) used AI in 2025, up from 6.4% in 2023, while 55.0% of large enterprises did.
- The gap between large and small EU enterprises using AI widened from 24.1 percentage points in 2023 to 38.0 points in 2025, even though small firms grew faster in relative terms.
- Small enterprises account for about 70% of the EU enterprises that use AI, because they make up 83% of the firms surveyed.
- Among small EU enterprises using AI, 38% use three or more AI technologies, against 61% of large adopters.
- In Denmark, 37.5% of small enterprises used AI in 2025; in Romania, 4.1%. The size gap widened in 25 of 27 member states between 2023 and 2025.
- Lack of expertise (70.9%), unclear legal consequences (54.8%) and data protection concerns (52.5%) were the main reasons small EU firms that considered AI did not use it; cost was cited by 38.8%, down from 48.2% in 2023.
Cite Edworking Research and the original sources when quoting these findings. Source material retains its original terms. See the study’s disclosure and limitations before reuse.
Abstract
In 2025, 17.0% of small enterprises in the European Union (10 to 49 persons employed) used at least one artificial intelligence technology, up from 6.4% in 2023. Large enterprises (250 or more) moved from 30.5% to 55.0% over the same period. Small firms grew faster in relative terms, but the gap between the two widened from 24.1 to 38.0 percentage points.
This study analyses the Eurostat survey on ICT usage in enterprises for 2021, 2023, 2024 and 2025 across all 27 member states, by size class, technology, reason for not adopting, country and sector. It adds calculations that the official release does not publish: the size gap over time and by country, the depth of use among adopters, and how barriers changed for small firms since 2023. No Edworking customer, team or usage data was used.
Key findings
- 17.0% of small EU enterprises used AI in 2025, against 6.4% in 2023 and 6.1% in 2021. Medium-sized enterprises reached 30.4% and large enterprises 55.0%. Across all enterprises with 10 or more persons employed, the share was 20.0%.
- The gap widened in points. Large enterprises were 4.8 times as likely as small ones to use AI in 2023 and 3.2 times in 2025, but the difference grew from 24.1 to 38.0 percentage points. Between 2024 and 2025 alone, large firms added 13.9 points and small firms 5.8.
- Small firms make up most AI users but use less of it. Because 83% of surveyed enterprises are small, they account for about 70% of EU enterprises using AI. Yet only 38% of small adopters use three or more AI technologies, against 61% of large adopters.
- A ninefold gap between countries. 37.5% of small enterprises in Denmark used AI in 2025, against 4.1% in Romania. The size gap widened between 2023 and 2025 in 25 of 27 member states.
- Expertise, not cost, is the main barrier. Among small enterprises that considered AI but did not use it, 70.9% cited a lack of relevant expertise, 54.8% unclear legal consequences and 52.5% data protection concerns. Cost fell from 48.2% in 2023 to 38.8%.
Introduction
Most announcements about AI at work come from large companies. Most European businesses are small. The question that matters for the typical team of 10 to 50 people is not whether AI is being adopted somewhere, but whether firms like theirs are adopting it, how deeply, and what stops them.
The Eurostat survey on ICT usage in enterprises asks the same questions in every member state and breaks the results down by size class and sector. It covers enterprises with 10 or more persons employed in industry, construction and most non-financial services. It does not cover businesses with fewer than 10 people.
Eurostat publishes the headline numbers each December. This study follows the small-firm series over time and across countries to see whether the period in which generative AI tools became widely available narrowed the distance between small and large firms or widened it.
Research questions
- How did AI use change between 2021 and 2025 among small, medium and large EU enterprises?
- Did the gap between small and large enterprises narrow or widen?
- How deeply do small adopters use AI, compared with large adopters?
- How large are the differences between member states for small firms?
- Why do small enterprises that considered AI not use it, and did that change since 2023?
- How much does AI use differ by sector?
Results
Small firms nearly tripled their use of AI
AI use among small enterprises was flat at about 6% in 2021 and 2023 (6.1% and 6.4%), then rose to 11.2% in 2024 and 17.0% in 2025: 2.7 times the 2023 level. Medium-sized enterprises went from 13.1% in 2023 to 30.4%, and large enterprises from 30.5% to 55.0%.
For all enterprises with 10 or more persons employed, the share rose from 7.7% in 2021 and 8.1% in 2023 to 13.5% in 2024 and 20.0% in 2025, matching the figure in Eurostat's December 2025 release. In 2025, about one in six small enterprises, three in ten medium-sized ones and more than half of large ones used at least one AI technology.
EU enterprises using AI by size, 2021 to 2025
Share of enterprises using at least one AI technology, by number of persons employed.
Eurostat, ICT usage in enterprises survey, isoc_eb_ai, EU-27, enterprises with 10 or more persons employed, excluding the financial sector. Extracted 12 November 2026 (dataset updated 15 June 2026). Eurostat, isoc_eb_ai ↗
The datasets have no AI values for 2022. Small, medium and large refer to 10 to 49, 50 to 249 and 250 or more persons employed.
View data table
| Item | % of enterprises | Notes |
|---|---|---|
| 2021 | 6.1 % of enterprises | — |
| 2023 | 6.4 % of enterprises | — |
| 2024 | 11.2 % of enterprises | — |
| 2025 | 17.0 % of enterprises | — |
EU enterprises using AI by size, 2021 to 2025 — Small (10 to 49). 2021: 6.1 % of enterprises. Eurostat, ICT usage in enterprises survey, isoc_eb_ai, EU-27, enterprises with 10 or more persons employed, excluding the financial sector. Extracted 12 November 2026 (dataset updated 15 June 2026). The datasets have no AI values for 2022. Small, medium and large refer to 10 to 49, 50 to 249 and 250 or more persons employed.
The gap with large firms widened in points
Small firms grew faster in relative terms. Their share multiplied by 2.7 between 2023 and 2025, against 1.8 for large firms, so the ratio between the two fell from 4.8 to 3.2.
In absolute terms, the gap moved the other way. Large enterprises added 24.6 percentage points between 2023 and 2025 and small enterprises 10.6. The difference between them grew from 22.3 points in 2021 and 24.1 in 2023 to 30.0 in 2024 and 38.0 in 2025. The widening accelerated in the last year: from 2024 to 2025, large firms added 13.9 points and small firms 5.8.
Gap between large and small enterprises
Large enterprises (250+) compared with small enterprises (10 to 49), EU-27.
Edworking calculation from Eurostat isoc_eb_ai: large minus small (points) and large divided by small (ratio). Eurostat, isoc_eb_ai ↗
Computed from unrounded Eurostat values. The ratio view is a multiple, not percentage points.
View data table
| Item | percentage points | Notes |
|---|---|---|
| 2021 | 22.3 percentage points | — |
| 2023 | 24.1 percentage points | — |
| 2024 | 30.0 percentage points | — |
| 2025 | 38.0 percentage points | — |
Gap between large and small enterprises — Gap in points. 2021: 22.3 percentage points. Edworking calculation from Eurostat isoc_eb_ai: large minus small (points) and large divided by small (ratio). Computed from unrounded Eurostat values. The ratio view is a multiple, not percentage points.
Small firms are most AI users, but use less of it
Small enterprises make up about 83% of the 1.53 million enterprises in the survey population, medium-sized ones 14% and large ones 3%. Weighting each size class's adoption rate by those shares, small firms account for roughly 70% of all EU enterprises using AI. In number of firms, AI adoption in Europe is mainly a small-business story.
Depth of use is different. In 2025, 6.5% of small enterprises used three or more AI technologies, which is 38% of small adopters. Among large adopters the figure was 61%. Large enterprises were also far more likely to use each technology: 31.7% generated written or spoken language or code, against 7.0% of small enterprises, and 25.1% used machine learning for data analysis, against 3.8%.
Small adopters concentrate on language tools. Text mining was the most common technology, used by 9.9% of all small enterprises (58% of small adopters), up from 5.8% in 2024. By purpose, 34% of small adopters used AI for marketing or sales and 28% for business administration or management, while only 15% used it for ICT security, against 48% of large adopters.
AI technologies used, small and large enterprises, 2025
Share of all enterprises in each size class using each technology.
Eurostat, ICT usage in enterprises survey, isoc_eb_ai, EU-27, enterprises with 10 or more persons employed, excluding the financial sector. Extracted 12 November 2026 (dataset updated 15 June 2026). Eurostat, isoc_eb_ai ↗
An enterprise can use several technologies. Shares are of all enterprises in the size class, not of AI users.
View data table
| Item | % of enterprises | Notes |
|---|---|---|
| Text mining (analysing written language) | 9.9 % of enterprises | — |
| Generating language or code | 7.0 % of enterprises | — |
| Speech recognition | 5.9 % of enterprises | — |
| Workflow automation or decision support | 4.1 % of enterprises | — |
| Machine learning for data analysis | 3.8 % of enterprises | — |
| Image recognition | 3.0 % of enterprises | — |
| Autonomous robots, vehicles or drones | 1.0 % of enterprises | — |
| Three or more AI technologies | 6.5 % of enterprises | — |
AI technologies used, small and large enterprises, 2025 — Small (10 to 49). Text mining (analysing written language): 9.9 % of enterprises. Eurostat, ICT usage in enterprises survey, isoc_eb_ai, EU-27, enterprises with 10 or more persons employed, excluding the financial sector. Extracted 12 November 2026 (dataset updated 15 June 2026). An enterprise can use several technologies. Shares are of all enterprises in the size class, not of AI users.
A ninefold gap between countries
Small-firm adoption ranges from 37.5% in Denmark, 33.5% in Finland and 30.7% in Sweden to 7.2% in Bulgaria, 5.8% in Poland and 4.1% in Romania. Seven member states were at 25% or more and eight were below 10%. The median member state was Ireland, at 16.8%. Among the larger economies, small-firm adoption was 23.1% in Germany, 17.2% in Spain, 15.0% in France and 14.2% in Italy.
Small-firm use rose between 2023 and 2025 in all 27 member states and at least doubled in 24. It fell slightly between 2024 and 2025 in Greece (8.2% to 7.3%) and Slovenia (19.3% to 18.3%).
The size gap widened between 2023 and 2025 in 25 of 27 countries; only Denmark and Luxembourg narrowed it. In 2025 it was largest in Slovenia (53.4 points), Belgium (47.6) and Finland (45.9), and smallest in Romania (16.6) and Bulgaria (19.0), where large-firm adoption is also low. In Poland, large enterprises were 7.9 times as likely as small ones to use AI (45.8% against 5.8%).
Small enterprises using AI by country
Share of enterprises with 10 to 49 persons employed using at least one AI technology.
Eurostat, ICT usage in enterprises survey, isoc_eb_ai, enterprises with 10 to 49 persons employed, excluding the financial sector. Extracted 12 November 2026 (dataset updated 15 June 2026). Eurostat, isoc_eb_ai ↗
National samples differ in size; values with Eurostat flags are noted in the rows.
View data table
| Item | % of small enterprises | Notes |
|---|---|---|
| Denmark | 37.5 % of small enterprises | — |
| Finland | 33.5 % of small enterprises | — |
| Sweden | 30.7 % of small enterprises | — |
| Luxembourg | 30.5 % of small enterprises | — |
| Belgium | 28.8 % of small enterprises | — |
| Netherlands | 28.8 % of small enterprises | Eurostat flag: break in series. |
| Austria | 26.2 % of small enterprises | — |
| Germany | 23.1 % of small enterprises | — |
| Estonia | 20.7 % of small enterprises | — |
| Slovenia | 18.3 % of small enterprises | — |
| Lithuania | 18.1 % of small enterprises | — |
| Malta | 18.0 % of small enterprises | — |
| Spain | 17.2 % of small enterprises | — |
| Ireland | 16.8 % of small enterprises | — |
| Slovakia | 15.6 % of small enterprises | — |
| France | 15.0 % of small enterprises | — |
| Italy | 14.2 % of small enterprises | — |
| Czechia | 13.4 % of small enterprises | — |
| Croatia | 12.8 % of small enterprises | — |
| Latvia | 9.8 % of small enterprises | — |
| Portugal | 9.4 % of small enterprises | — |
| Hungary | 8.7 % of small enterprises | — |
| Cyprus | 7.7 % of small enterprises | — |
| Greece | 7.3 % of small enterprises | — |
| Bulgaria | 7.2 % of small enterprises | — |
| Poland | 5.8 % of small enterprises | — |
| Romania | 4.1 % of small enterprises | — |
| EU-27 | 17.0 % of small enterprises | — |
Small enterprises using AI by country — 2025. Denmark: 37.5 % of small enterprises. Eurostat, ICT usage in enterprises survey, isoc_eb_ai, enterprises with 10 to 49 persons employed, excluding the financial sector. Extracted 12 November 2026 (dataset updated 15 June 2026). National samples differ in size; values with Eurostat flags are noted in the rows.
Size gap by country
Share of large enterprises (250+) using AI minus share of small enterprises (10 to 49), in percentage points.
Edworking calculation from Eurostat isoc_eb_ai, unrounded values. Eurostat, isoc_eb_ai ↗
Large-enterprise samples are small in some countries, so national gaps carry more uncertainty than the EU figure. 2023 values for France and Sweden are flagged as breaks in series.
View data table
| Item | percentage points | Notes |
|---|---|---|
| Slovenia | 53.4 percentage points | — |
| Belgium | 47.6 percentage points | — |
| Finland | 45.9 percentage points | — |
| France | 43.0 percentage points | — |
| Austria | 42.1 percentage points | — |
| Sweden | 41.1 percentage points | — |
| Czechia | 40.7 percentage points | — |
| Ireland | 40.5 percentage points | — |
| Spain | 40.3 percentage points | — |
| Poland | 40.0 percentage points | — |
| Portugal | 39.7 percentage points | — |
| Netherlands | 38.9 percentage points | — |
| Italy | 38.9 percentage points | — |
| Latvia | 37.3 percentage points | — |
| Denmark | 37.1 percentage points | — |
| Germany | 33.9 percentage points | — |
| Malta | 33.7 percentage points | — |
| Lithuania | 33.7 percentage points | — |
| Estonia | 32.2 percentage points | — |
| Hungary | 31.3 percentage points | — |
| Greece | 30.2 percentage points | — |
| Slovakia | 28.2 percentage points | — |
| Luxembourg | 27.5 percentage points | — |
| Cyprus | 27.4 percentage points | — |
| Croatia | 26.8 percentage points | — |
| Bulgaria | 19.0 percentage points | — |
| Romania | 16.6 percentage points | — |
| EU-27 | 38.0 percentage points | — |
Size gap by country — 2025. Slovenia: 53.4 percentage points. Edworking calculation from Eurostat isoc_eb_ai, unrounded values. Large-enterprise samples are small in some countries, so national gaps carry more uncertainty than the EU figure. 2023 values for France and Sweden are flagged as breaks in series.
Expertise, not cost, holds back small firms
In 2025, 10.2% of small enterprises had considered using AI without adopting it, against 17.0% that used it. For every ten small firms using AI, about six more had looked at it and stopped. Among large enterprises the ratio was about three to ten.
Those small enterprises that considered AI but did not use it gave several reasons. The most common were a lack of relevant expertise (70.9%), unclear legal consequences (54.8%), data protection and privacy concerns (52.5%), difficulties with the availability or quality of data (43.7%) and incompatibility with existing systems (41.7%). Costs that seem too high were cited by 38.8%, and 19.2% said AI was not useful for their business.
The mix changed since 2023. Concerns about legal consequences rose by 7.4 points and data protection concerns by 8.6 points, while cost fell by 9.4 points and incompatibility by 6.3 points. Lack of expertise stayed the top reason in every year, cited by about seven in ten.
Why enterprises that considered AI did not use it
Reasons given by enterprises that did not use AI but had considered using it. Several reasons could be chosen.
Eurostat isoc_eb_ai, unit: percentage of the enterprises which ever considered using any AI technology, EU-27. Extracted 12 November 2026 (dataset updated 15 June 2026). Eurostat, isoc_eb_ai ↗ Eurostat, Use of artificial intelligence in enterprises ↗
Asked only of enterprises not using AI that had considered it: about 10% of small and 16% of large enterprises in 2025. Eurostat's December 2025 article quotes values that match the dataset's 2024 figures; this study uses the dataset values as published in June 2026.
View data table
| Item | % of enterprises that considered AI | Notes |
|---|---|---|
| Lack of relevant expertise | 70.9 % of enterprises that considered AI | — |
| Unclear legal consequences | 54.8 % of enterprises that considered AI | — |
| Data protection and privacy concerns | 52.5 % of enterprises that considered AI | — |
| Data availability or quality | 43.7 % of enterprises that considered AI | — |
| Incompatible with existing systems | 41.7 % of enterprises that considered AI | — |
| Costs seem too high | 38.8 % of enterprises that considered AI | — |
| Ethical considerations | 25.3 % of enterprises that considered AI | — |
| AI not useful for the enterprise | 19.2 % of enterprises that considered AI | — |
Why enterprises that considered AI did not use it — Small, 2025. Lack of relevant expertise: 70.9 % of enterprises that considered AI. Eurostat isoc_eb_ai, unit: percentage of the enterprises which ever considered using any AI technology, EU-27. Extracted 12 November 2026 (dataset updated 15 June 2026). Asked only of enterprises not using AI that had considered it: about 10% of small and 16% of large enterprises in 2025. Eurostat's December 2025 article quotes values that match the dataset's 2024 figures; this study uses the dataset values as published in June 2026.
Sector gaps are wider than size gaps
Sector differences are larger than size differences. In 2025, 62.5% of enterprises with 10 or more persons employed in information and communication used AI, and 65.7% in computer programming and consultancy. Professional, scientific and technical activities reached 40.4%, with advertising and market research at 48.0%, up from 12.5% in 2021. At the other end were construction (10.8%), transportation and storage (11.2%) and accommodation and food services (12.0%).
The 51.7-point distance between information and communication and construction is wider than the 38.0-point size gap. Every sector in the table rose between 2021 and 2025. The largest gain was in information and communication (37.2 points), which already led in 2021; the smallest were in transportation and storage (5.9) and construction (6.2).
Enterprises using AI by sector
Enterprises with 10 or more persons employed, EU-27, NACE Rev. 2 sections.
Eurostat, ICT usage in enterprises survey, isoc_eb_ain2, EU-27, enterprises with 10 or more persons employed. Extracted 12 November 2026 (dataset updated 15 June 2026). Eurostat, isoc_eb_ain2 ↗
Eurostat does not publish sector by size at EU level in this table, so sector and size effects cannot be separated here.
View data table
| Item | % of enterprises | Notes |
|---|---|---|
| Information and communication | 62.5 % of enterprises | — |
| Professional, scientific and technical | 40.4 % of enterprises | — |
| Electricity and gas supply | 33.6 % of enterprises | — |
| Real estate | 24.8 % of enterprises | — |
| Wholesale trade | 23.4 % of enterprises | — |
| Administrative and support services | 19.9 % of enterprises | — |
| Manufacturing | 17.3 % of enterprises | — |
| Retail trade | 15.5 % of enterprises | — |
| Water supply and waste | 13.1 % of enterprises | — |
| Accommodation and food services | 12.0 % of enterprises | — |
| Transportation and storage | 11.2 % of enterprises | — |
| Construction | 10.8 % of enterprises | — |
| All covered sectors | 19.9 % of enterprises | — |
Enterprises using AI by sector — 2025. Information and communication: 62.5 % of enterprises. Eurostat, ICT usage in enterprises survey, isoc_eb_ain2, EU-27, enterprises with 10 or more persons employed. Extracted 12 November 2026 (dataset updated 15 June 2026). Eurostat does not publish sector by size at EU level in this table, so sector and size effects cannot be separated here.
Discussion
The data support two readings at once. AI use among small European firms is no longer marginal: it nearly tripled in two years and rose in every member state. But large firms moved faster in absolute terms and adopt more technologies for more purposes, so the distance between them grew.
The most likely explanation is that large firms combined the new language tools with the data, security and automation systems they already had, while small firms mostly added one or two ready-to-use tools. The survey shows that pattern in the technology and purpose breakdowns, but it does not measure the intensity of use, spending or results, so it cannot show whether small adopters get less value from AI.
The barriers point the same way. Small firms that hold back cite skills and rules more than money, and cost became less important between 2023 and 2025. The rise in legal and privacy concerns over the same period coincides with the adoption of the EU AI Act in 2024, although the survey does not ask about it directly.
Country differences are large and persistent. They likely reflect the sector mix, digital skills, the use of other digital technologies and access to advice. This study does not test those explanations.
Practical implications
For managers of small teams, the data suggest that the useful question is not whether to try AI, but how to make it part of how the team works. Concrete steps:
- start where small adopters already see use, such as drafting, summarising and searching documents, and write down which tasks AI may and may not do;
- treat expertise as the main constraint: give one person time to test tools, document prompts and share examples with the rest of the team;
- answer the legal and privacy questions before rollout, with a short policy on what data may be shared with AI tools and who reviews outputs;
- keep documents, decisions and project data in one shared place, since AI tools are only as useful as the information they can reach;
- compare yourself with your sector and country, not with the EU average: a 20-person agency and a 20-person construction firm face very different norms.
Edworking's guide on how to use AI to automate tasks covers practical first uses, and our earlier study on UK business AI adoption looks at depth of use in the UK, with a different survey and definitions.
Conclusion
Between 2023 and 2025, the share of small EU enterprises using AI rose from 6.4% to 17.0%, and small firms now make up about seven in ten of the businesses that use it. Large enterprises still moved faster in points and use more technologies, so the size gap widened to 38 points and grew in 25 of 27 member states. For small firms that considered AI and held back, the main obstacle is a lack of expertise, followed by legal and privacy questions, while cost has become less important.
Data and methodology
Source
All figures come from the Eurostat survey on ICT usage in enterprises, datasets Artificial intelligence by size class of enterprise (isoc_eb_ai) and Artificial intelligence by NACE Rev. 2 activity (isoc_eb_ain2), both last updated 15 June 2026 and extracted on 12 November 2026 through the Eurostat dissemination API. Values are for the EU-27 (2020 composition) and the 27 member states.
The survey covers enterprises with 10 or more persons employed in manufacturing, utilities, construction, trade, transport, accommodation and food, information and communication, real estate, professional and administrative services and computer repair. The financial sector, agriculture and mining are excluded. National statistical offices run it in the first months of each year; in 2025, 157,000 of the 1.53 million enterprises in scope were surveyed, according to Eurostat's Statistics Explained article. The datasets have no AI values for 2022.
Definitions
Size classes follow the number of persons employed: small (10 to 49), medium (50 to 249) and large (250 or more). An enterprise "uses AI" if it uses at least one of the technologies listed in the survey: text mining, speech recognition, natural language generation (including code), image recognition, machine learning for data analysis, AI-based process automation or decision support, and autonomous robots, vehicles or drones. Reasons for not using AI were asked only of enterprises that did not use AI but had considered it.
Calculations
All figures were computed in Python from the API response, using unrounded values and rounding to one decimal at the end:
- size gap = share of large enterprises minus share of small enterprises (2025: 55.03 - 17.00 = 38.0 points; 2023: 30.48 - 6.41 = 24.1);
- ratio = share of large divided by share of small (2025: 3.2; 2023: 4.8);
- depth of use = enterprises using three or more technologies divided by enterprises using any (small 2025: 6.50 / 17.00 = 38%; large: 33.58 / 55.03 = 61%); purposes are computed the same way;
- share of AI users that are small = 0.83 x 17.00 / (0.83 x 17.00 + 0.14 x 30.36 + 0.03 x 55.03) = 70%, using Eurostat's rounded population shares, which reproduce the published 20.0% total;
- country counts compare the 27 member states only; the median is the 14th of 27 values.
Headline values match the Eurostat release of 11 December 2025: 20.0% of EU enterprises used AI in 2025, against 13.5% in 2024, 8.1% in 2023 and 7.7% in 2021, and 17.0%, 30.36% and 55.03% of small, medium and large enterprises.
Limitations
The survey measures whether an enterprise uses a technology, not how much, how well or by how many employees. A firm where one person uses a writing assistant counts the same as one that runs AI across its operations. The depth indicator used here (number of technologies) is a rough proxy.
Some values carry Eurostat flags: 2023 values for France and Sweden and 2025 values for small enterprises in the Netherlands are marked as breaks in series and are noted in the figures. Large-enterprise samples are small in some countries, so national size gaps are less precise than EU figures. Eurostat's December 2025 article quotes reasons for not using AI that match the dataset's 2024 values; this study uses the 2025 values as published in the dataset in June 2026.
Enterprises with fewer than 10 people are not covered, so the results do not describe micro businesses or freelancers. Sector results are only available for all enterprises with 10 or more persons employed, so sector and size effects cannot be separated. Differences between countries, sizes and sectors are descriptive; this study does not estimate causes or effects on productivity. It has not been peer reviewed.
References
- Eurostat, Artificial intelligence by size class of enterprise (isoc_eb_ai), updated 15 June 2026.
- Eurostat, Artificial intelligence by NACE Rev. 2 activity (isoc_eb_ain2), updated 15 June 2026.
- Eurostat, Use of artificial intelligence in enterprises, Statistics Explained, data extracted December 2025.
- Eurostat, 20% of EU enterprises use AI technologies, news release, 11 December 2025.
- Eurostat, ICT usage in enterprises: reference metadata (isoc_e_esms).
- European Union, Regulation (EU) 2024/1689 laying down harmonised rules on artificial intelligence (AI Act).
Disclosure
This study was produced by Edworking with AI-assisted data processing and writing, using public Eurostat data only. Edworking sells collaboration software for small teams that includes AI features and has a commercial interest in AI adoption. No Edworking customer, team or usage data was used, and the analysis does not evaluate any product. It is editorial research, not a peer-reviewed paper.
Suggested citation
Edworking Team (2026). "Small EU firms nearly tripled AI use, but the gap with large firms grew." Edworking Research. https://edworking.com/research/eu-small-business-ai-adoption-gap-2025
Cite this study
Edworking Team (2026). Small EU firms nearly tripled AI use, but the gap with large firms grew. Edworking Research. https://edworking.com/research/eu-small-business-ai-adoption-gap-2025
Cite Edworking Research and the original sources when quoting these findings. Source material retains its original terms. See the study’s disclosure and limitations before reuse.