For owners and analysts comparing European enterprise AI use, this reference explains the country, size and industry figures and how to use them. It includes the full activity tables, Germany’s breakdown and reproducible downloads.
What you'll learn
- How adoption differs between countries and enterprise sizes
- Which activity categories the full sector tables cover
- How Germany compares with the EU by size and industry
- Which denominators apply to technologies, purposes and barriers
- How to reproduce a figure and interpret missing values and flags
Eurostat's 2025 figures show AI use in 17% of small EU enterprises, 30.36% of medium enterprises and 55.03% of large enterprises.[1] Professional, scientific and technical activities reached 40.43%.[2]
These figures describe broad AI use, not autonomous-agent deployment. They show a clear size and sector gap, but they do not tell a small business which workflow to automate. This analysis maps the gap and explains what it means for workflow readiness, expertise, legal clarity, and privacy.
Which three baselines does European AI adoption have?
European AI adoption is not one number. The answer changes when the question changes from any use, to production use, to an autonomous agent inside a business process.
Eurostat measures reported use of listed AI technologies by enterprises with at least 10 employees and self-employed persons in the covered economic activities. The list includes text analysis, language generation, speech recognition, image recognition, machine learning, workflow automation, autonomous physical movement and generated pictures, video or audio. The survey is a broad AI-use baseline, rather than a count of agents deployed in production.[4]
For the distinction between a model, an agent and a workflow, see what is an AI agent. Use the tables here to compare the European market around your business.
Say a business uses a writing assistant to prepare an email. Another runs a workflow that reads a CRM record, drafts a follow-up, waits for approval and records the action. These are different operating arrangements. The broad AI-use question does not separate them or measure how much work either system completes.[4]
Eurostat's figures describe measured AI technologies and purposes. They do not prove that a business has deployed an autonomous agent, achieved a specific return, or moved a workflow into production.
The survey also defines the size classes that shape the rest of the analysis:
| Eurostat size class | Definition | Why it matters |
|---|---|---|
| Small | 10–49 persons employed, including self-employed people | The closest official category to many lean service businesses, but not to every solo operator |
| Medium | 50–249 persons employed, including self-employed people | Compare with firms in this class; size alone does not describe staffing |
| Large | 250 or more persons employed, including self-employed people | Compare separately from a ten-person firm; size does not establish its resources |
The 2025 survey sampled approximately 157,000 of 1.53 million EU enterprises. About 83% of the population were small, 14% medium and 3% large. Enterprises below 10 persons employed are outside the common population used in these tables.[6]
How does AI adoption differ by enterprise size?
Large EU enterprises reported AI use more often than small ones in 2025: 55.03% against 17%, a difference of 38.03 percentage points. Medium enterprises reached 30.36%. These are size-class observations, not a measured explanation for the gap.[1]
Compare the resources your own workflow needs: available budget, specialist help, responsibility for data and time to connect the tools. If one person already owns the process, approvals and exception handling, make that workload explicit before expanding it. The adoption percentage cannot tell you which constraint matters in your team.
Adoption is not the same as a controlled workflow in production.
Consider two hypothetical setups. A large firm tests a system in one department with separate data and security teams; a small agency runs a client workflow with one operator managing access and exceptions. Both might report AI use. Compare the actual operating arrangement before using either as a model for your business.
The comparison should therefore guide questions rather than produce a verdict:
- Is the business measuring tool experimentation or live workflow use?
- Does one named person own the process and the agent's output quality?
- Can the business define what the agent may read, write, draft, or send?
- Does the workflow have a baseline before automation begins?
- Can the team review exceptions without creating a second full-time job?
Yardwork's service sits in that gap between broad interest and practical operation. The relevant work is not convincing a founder that AI exists. The relevant work is turning one real process into a bounded system with connected tools, approval points, and a way to see whether the output is right.
Choose the first workflow from your own work, with one owner and a measurable outcome. Use a larger firm’s plan only where its staffing, controls and task boundaries also fit your team.
What is the AI adoption rate in every EU country by size?
The country table compares the same 2025 measure across all EU member states: enterprises using at least one listed AI technology. The EU total is 19.95%; Germany is 25.97%. The size columns show why a national total should not substitute for a comparison with firms of your size.[1]
What is the AI adoption rate in the member states that joined before 2004?
| Country | All 10+ | Small 10–49 | Medium 50–249 | Large 250+ |
|---|---|---|---|---|
| EU-27 | 19.95% | 17.00% | 30.36% | 55.03% |
| Belgium | 34.54% | 28.82% | 54.49% | 76.41% |
| Denmark | 42.03% | 37.45% | 58.27% | 74.52% |
| Germany | 25.97% | 23.06% | 35.56% | 56.99% |
| Ireland | 19.64% | 16.75% | 28.02% | 57.27% |
| Greece | 8.93% | 7.33% | 15.18% | 37.58% |
| Spain | 20.27% | 17.17% | 30.10% | 57.46% |
| France | 18.16% | 15.04% | 30.81% | 58.01% |
| Italy | 16.40% | 14.23% | 27.62% | 53.09% |
| Luxembourg | 33.61% | 30.48% | 43.68% | 57.97% |
| Netherlands | 33.21% (b) | 28.76% (b) | 46.62% | 67.64% |
| Austria | 29.95% | 26.18% | 44.56% | 68.26% |
| Portugal | 11.54% | 9.43% | 18.21% | 49.15% |
| Finland | 37.82% | 33.51% | 51.35% | 79.40% |
| Sweden | 35.04% | 30.74% | 49.60% | 71.87% |
What is the AI adoption rate in the member states that joined from 2004?
| Country | All 10+ | Small 10–49 | Medium 50–249 | Large 250+ |
|---|---|---|---|---|
| Bulgaria | 8.55% | 7.17% | 13.30% | 26.18% |
| Czechia | 17.60% | 13.38% | 28.79% | 54.10% |
| Estonia | 23.40% | 20.74% | 33.10% | 52.90% |
| Croatia | 15.19% | 12.77% | 25.50% | 39.57% |
| Cyprus | 9.27% | 7.72% | 15.31% | 35.07% |
| Latvia | 12.21% | 9.77% | 19.78% | 47.09% |
| Lithuania | 21.30% | 18.06% | 30.36% | 51.75% |
| Hungary | 10.37% | 8.70% | 15.34% | 39.99% |
| Malta | 21.51% | 18.04% | 30.66% | 51.75% |
| Poland | 8.36% | 5.79% | 14.77% | 45.81% |
| Romania | 5.21% | 4.13% | 7.79% | 20.75% |
| Slovenia | 21.61% | 18.29% | 30.62% | 71.74% |
| Slovakia | 18.00% | 15.59% | 23.05% | 43.74% |
Values are percentages of enterprises within each column’s population. b = break in time series; u = low reliability; c = confidential; — = no published value. Missing values are never treated as zero. The Netherlands total and small-enterprise values carry b. The EU figure is Eurostat’s aggregate, not the average of the national percentages.[1][4]
For a meeting or offline comparison, download the English country and size table (PDF). For your own calculations, download the complete size dataset (CSV).
How do country and sector patterns change the story?
The country range in 2025 ran from 5.21% in Romania to 42.03% in Denmark. Finland reached 37.82% and Sweden 35.04%. Compare the same population and year when using these national rates.[1]
A national rate describes the covered enterprises in that country. It does not explain why an individual firm uses AI. Compare sector and size separately before drawing a business conclusion. The Netherlands total and small-enterprise series carry a break flag; use care with a year-to-year comparison.[1]
Information and communication reached 62.52% in 2025, professional, scientific and technical activities 40.43%, real estate 24.76% and construction 10.79%. Each percentage uses all covered enterprises in that activity as its denominator.[2]
For a service business, the professional-services category is a relevant comparison. Use the more detailed activity table below if it fits your work better. Then choose a concrete information task, such as a client report or document handoff, and assess that task independently of the sector percentage.
The next question is not “Which country should adopt AI?” The decision question is “Which operating conditions make a workflow easier to measure and control?” A ten-person consultancy in a high-adoption country may still lack a system of record. A small agency in a lower-adoption country may already have a clear workflow and a willing owner.
Country and sector data clarify the decision when they narrow it without replacing it. A founder can use the figures to compare context, then return to the work itself: repeated inputs, stable outputs, clear ownership, acceptable risk, and a human review point.
How does AI use compare across countries in service-related sectors?
The country table compares AI use in six service-related sectors in 2025. Information and communication reached 62.52% across the EU; professional, scientific and technical activities reached 40.43%. Each rate covers enterprises with at least 10 persons employed in that activity.[2]
We selected information and communication; professional, scientific and technical activities; administrative and support services; wholesale and retail; transport and storage; and accommodation and food services to compare the information, professional and operational work of service businesses.
How does AI use compare in the member states that joined before 2004?
| Country | Information and communication | Professional, scientific and technical activities | Administrative and support service activities | Wholesale and retail trade; repair of motor vehicles and motorcycles | Transportation and storage | Accommodation and food service activities |
|---|---|---|---|---|---|---|
| EU-27 | 62.52% | 40.43% | 19.86% | 18.62% | 11.15% | 11.98% |
| Belgium | 78.15% | 57.11% | 28.95% | 32.84% | 22.36% | 18.45% |
| Denmark | 79.08% | 63.71% | 33.06% | 44.80% | 26.90% | 25.20% |
| Germany | 75.38% | 48.30% | 22.08% | 25.46% | 12.22% | 14.63% |
| Ireland | 59.82% | 31.76% | 25.61% | 13.13% | 16.76% | 13.55% |
| Greece | 43.93% | 30.81% | 14.79% | 9.49% | 9.56% | 3.39% |
| Spain | 60.57% | 38.35% | 18.79% | 17.81% | 14.69% | 15.23% |
| France | 58.98% | 32.50% | 19.32% | 12.90% | 8.96% | 11.83% |
| Italy | 51.28% | 35.72% | 15.97% | 18.13% | 11.62% | 11.73% |
| Luxembourg | 73.48% | 51.65% | 31.88% | 27.24% | 27.90% | 16.95% |
| Netherlands | 69.99% | 54.98% | 30.85% | 30.47% (b) | 19.65% | 16.03% |
| Austria | 79.93% | 52.80% | 25.72% | 26.46% | 19.37% | 23.72% |
| Portugal | 52.80% | 25.68% | 11.06% | 8.71% | 14.81% | 8.25% |
| Finland | 79.79% | 74.75% | 38.65% | 34.26% | 12.93% | 23.53% |
| Sweden | 87.87% | 64.88% | 31.61% | 33.13% | 12.17% | 23.74% |
How does AI use compare in the member states that joined from 2004?
| Country | Information and communication | Professional, scientific and technical activities | Administrative and support service activities | Wholesale and retail trade; repair of motor vehicles and motorcycles | Transportation and storage | Accommodation and food service activities |
|---|---|---|---|---|---|---|
| Bulgaria | 42.01% | 26.38% | 7.83% | 7.26% | 5.71% | 3.95% |
| Czechia | 60.49% | 33.31% | 13.01% | 19.17% | 6.06% | 10.14% |
| Estonia | 56.60% | 52.24% | 19.07% | 18.40% | 13.99% | 16.25% |
| Croatia | 68.98% | 35.28% | 16.43% | 16.60% | 4.83% | 6.16% |
| Cyprus | 46.61% | 14.14% | 10.95% | 4.38% | 5.52% | 4.56% |
| Latvia | 49.82% | 29.02% | 12.33% | 10.37% | 5.90% | 4.46% |
| Lithuania | 60.24% | 37.68% | 29.75% | 19.16% | 20.02% | 8.00% |
| Hungary | 46.05% | 22.86% | 8.68% | 9.79% | 5.86% | 5.20% |
| Malta | 60.40% | 35.43% | 23.20% | 14.75% | 18.55% | 10.34% |
| Poland | 38.09% | 22.16% | 10.47% | 8.88% | 2.97% | 5.57% |
| Romania | 32.07% | 15.60% | 5.04% | 4.13% | 3.07% | 1.71% |
| Slovenia | 67.06% | 28.29% | 20.20% | 29.81% | 8.51% | 12.62% |
| Slovakia | 50.25% | 28.48% | 19.79% | 19.26% | 15.78% | 9.82% |
Use the same flag legend as the country-size table. — means no published value, not zero. The full matrix includes all published activity categories and overlapping aggregates; do not add these rates together.[2]
Full table for all sectors: CSV, English PDF.
How does Germany compare by enterprise size and industry?
Germany’s overall rate was 25.97% in 2025, compared with 19.95% across the EU: a difference of 6.02 percentage points. German small enterprises reached 23.06%, medium enterprises 35.56% and large enterprises 56.99%. These size comparisons use the same covered activities and enterprise definition.[1]
| Size | Germany | EU-27 |
|---|---|---|
| All 10+ | 25.97% | 19.95% |
| Small 10–49 | 23.06% | 17.00% |
| Medium 50–249 | 35.56% | 30.36% |
| Large 250+ | 56.99% | 55.03% |
The following detailed German activity table uses all enterprises with 10+ persons employed. It does not tell you the rate for a small German firm in that sector: the published sector dataset has no size split. Use the separate size and sector comparisons as context, and request a suitable national tabulation if the joint estimate is essential.[2]
| Activity | Germany | EU-27 |
|---|---|---|
| Manufacturing, electricity, gas, steam and air conditioning supply; water supply and construction | 19.98% | 14.75% |
| Manufacturing, electricity, gas, steam and air conditioning; water supply, sewerage, waste management and remediation activities | 24.28% | 17.32% |
| Manufacturing | 24.38% | 17.27% |
| All activities (except agriculture, forestry and fishing, and mining and quarrying), without financial sector | 25.97% | 19.95% |
Which German manufacturing sectors use AI?
| Activity | Germany | EU-27 |
|---|---|---|
| Manufacture of products based on: food, beverages, tobacco, textile, leather, wood, pulp and paper; publishing and printing | 19.27% | 13.17% |
| Manufacture of food products; beverages and tobacco products | 15.10% | 12.47% |
| Manufacture of textiles, wearing apparel, leather and related products | 22.60% | 12.01% |
| Manufacture of wood, paper, printing and reproduction | 27.38% | 15.63% |
| Manufacture of coke, refined petroleum, chemical and basic pharmaceutical products, rubber and plastics, other non-metallic mineral products | 25.84% | 20.35% |
| Manufacture of coke and refined petroleum products | 45.65% | 30.41% |
| Manufacture of chemicals and chemical products | 34.14% | 28.45% |
| Manufacture of basic pharmaceutical products and pharmaceutical preparations | — (u) | 41.32% |
| Manufacture of rubber and plastic products and other non-metallic mineral products | 22.05% | 16.77% |
| Manufacture of basic metals and fabricated metal products, except machinery and equipment | 21.54% | 14.33% |
Which German engineering, utility and construction sectors use AI?
| Activity | Germany | EU-27 |
|---|---|---|
| Manufacture of computers, electric and optical products, electrical equipment, machinery and equipment n.e.c, motor vehicles, other transport equipment, furniture, other manufacturing, repair and installation of machinery and equipment | 29.09% | 22.73% |
| Manufacture of computer, electronic and optical products | 42.49% | 37.44% |
| Manufacture of electrical equipment, machinery and equipment n.e.c. | 30.77% | 25.57% |
| Manufacture of electrical equipment | 29.37% | 25.90% |
| Manufacture of machinery and equipment n.e.c. | 31.28% | 25.45% |
| Manufacture of motor vehicles, trailers, semi-trailers and of other transport equipment | 30.92% | 23.86% |
| Manufacture of furniture; jewellery, musical instruments, toys; repair and installation of machinery and equipment | 22.94% | 17.16% |
| Electricity, gas, steam and air conditioning supply; water supply; sewerage, waste management and remediation activities | 22.97% | 18.16% |
| Electricity, gas, steam and air conditioning supply | 37.38% | 33.57% |
| Electricity, gas, steam and air conditioning supply | 37.38% | 33.57% |
| Water supply; sewerage, waste management and remediation activities | 16.88% | 13.12% |
| Construction | 14.28% | 10.79% |
Which German trade, transport and hospitality sectors use AI?
| Activity | Germany | EU-27 |
|---|---|---|
| Wholesale and retail trade; repair of motor vehicles and motorcycles | 25.46% | 18.62% |
| Services, without financial sector | 29.13% | 22.96% |
| Wholesale and retail trade and repair of motor vehicles and motorcycles | 18.28% | 14.35% |
| Wholesale trade, except of motor vehicles and motorcycles | 32.36% | 23.42% |
| Retail trade, except of motor vehicles and motorcycles | 23.42% | 15.46% |
| Transportation and storage | 12.22% | 11.15% |
| Accommodation and food service activities | 14.63% | 11.98% |
| Accommodation | 23.70% | 21.31% |
Which German information, professional and support services use AI?
| Activity | Germany | EU-27 |
|---|---|---|
| Information and communication | 75.38% | 62.52% |
| Publishing, motion picture, video, television programme production; sound recording, programming and broadcasting activities | 67.34% | 56.20% |
| Telecommunications | 47.59% | 39.76% |
| Computer programming, consultancy, and information service activities | 78.22% | 65.70% |
| Real estate activities; professional, scientific and technical activities | 45.60% | 38.16% |
| Real estate activities | 29.06% | 24.76% |
| Real estate activities | 29.06% | 24.76% |
| Professional, scientific and technical activities | 48.30% | 40.43% |
| Legal and accounting activities; activities of head offices; management consultancy activities; architectural and engineering activities; technical testing and analysis | 46.11% | 37.92% |
| Scientific research and development | 59.14% | 52.08% |
| Advertising and market research; other professional, scientific and technical activities; veterinary activities | 57.24% | 48.02% |
| Activities for rental and leasing, employment, security and investigation, services to buildings and landscape, office administrative, office support and other business support | 21.71% | 19.05% |
| Administrative and support service activities | 22.08% | 19.86% |
| Travel agency, tour operator and other reservation service and related activities | 32.32% | 34.96% |
| Repair of computers and communication equipment | — (u) | 24.50% |
| Information and Communication Technology - total | 71.25% | 60.13% |
Which operating purposes do businesses use AI for?
Among EU enterprises already using AI in 2025, 34.70% used it for marketing or sales and 31.04% for business administration or management. The denominator here is AI users, not all enterprises.[3]
Turn these broad purpose labels into questions about the work. For marketing, ask whether the system drafts material, analyses leads or supports campaign decisions. For administration, ask whether it prepares reports, handles documents or coordinates a handoff. Those are possible workflow examples; the purpose percentage does not identify the task performed or the approval arrangement.[4]
Among AI users, administration and management reached 28.49% in small enterprises and 43.38% in large ones; production processes reached 19.02% and 33.46%; ICT security reached 14.51% and 47.51%. These comparisons keep the AI-user denominator within each size class.[3]
| Purpose | All AI users | Small | Medium | Large |
|---|---|---|---|---|
| Marketing or sales | 34.70% | 34.06% | 35.67% | 37.89% |
| Administration or management | 31.04% | 28.49% | 34.97% | 43.38% |
| Production processes | 20.76% | 19.02% | 21.82% | 33.46% |
| Logistics | 6.08% | 4.92% | 6.62% | 14.95% |
| ICT security | 19.77% | 14.51% | 26.96% | 47.51% |
| Accounting, controlling or finance | 23.11% | 22.20% | 24.21% | 28.26% |
| R&D or innovation | 19.05% | 17.33% | 20.84% | 29.67% |
More than one purpose may be reported, so these rows do not add to 100%. Do not compare a purpose share among AI users directly with an adoption share among all enterprises.[3]
Purpose data redirects attention from tool names to work. A founder does not need to start with OpenClaw, Hermes, or a model provider. The first question is whether a recurring process produces a measurable output that a system can prepare, route, or complete under defined controls.
That process includes:
- collecting information from a client intake;
- drafting a weekly account report;
- routing an inbound request to the right owner;
- preparing a proposal from approved information;
- checking a record for missing fields;
- preparing a follow-up that waits for human approval.
The agent is not the business outcome. The workflow is the business outcome. An agent that sends more messages but creates more corrections has not improved the process. An agent that reduces forgotten handoffs while keeping external actions under approval may create value even if it never acts autonomously.
Use which workflows to automate first to choose the first job. Start with a recurring task whose inputs, expected output and review point you can name.
Which AI technologies does the European survey count?
The survey counts different technologies, from language tools to autonomous machines. Each row below is a share of all covered EU enterprises in 2025, not only AI users. A business can use several technologies, so the rows must not be added to estimate total adoption.[3]
| Technology | All 10+ | Small | Medium | Large |
|---|---|---|---|---|
| Text analysis | 11.75% | 9.87% | 18.22% | 35.04% |
| Speech recognition | 7.22% | 5.89% | 11.27% | 26.13% |
| Language or code generation | 8.76% | 6.96% | 14.79% | 31.68% |
| Image recognition or processing | 3.78% | 3.00% | 5.78% | 16.66% |
| Machine learning for data analysis | 5.10% | 3.81% | 8.73% | 25.05% |
| Workflow automation or decision support | 5.35% | 4.13% | 8.67% | 24.42% |
| Autonomous physical movement | 1.39% | 1.00% | 2.23% | 8.55% |
| Generating pictures, video or audio | 9.55% | 8.10% | 14.45% | 27.92% |
Language generation can occur in a writing assistant, a larger workflow or another product. Workflow automation can involve human review. None of these technology labels tells you whether the system runs without approval, how often it succeeds or what a business saved.[4]
What do the reported barriers point to?
Among EU enterprises that had considered AI but did not use it in 2025, lack of relevant expertise was reported by 70.31%, unclear legal consequences by 53.61% and data-protection or privacy concerns by 52.72%. These reasons use the considered-AI group as the denominator.[3]
| Reason | All | Small | Medium | Large |
|---|---|---|---|---|
| Lack of expertise | 70.31% | 70.87% | 69.18% | 65.11% |
| Unclear legal consequences | 53.61% | 54.79% | 49.41% | 51.22% |
| Privacy and data protection | 52.72% | 52.49% | 52.32% | 58.76% |
| Data availability or quality | 43.51% | 43.66% | 42.83% | 43.81% |
| Incompatibility with existing systems | 41.60% | 41.69% | 41.98% | 37.99% |
| Costs too high | 38.37% | 38.82% | 37.32% | 34.78% |
| Ethical considerations | 24.64% | 25.31% | 22.70% | 21.20% |
| Not useful for the enterprise | 17.79% | 19.22% | 13.76% | 9.77% |
Reasons can overlap. These are self-reported reasons from the considered-AI group, not causal effects, not a ranking of every non-user, and not evidence that paying for a particular tool will remove a barrier.[3]
What does each reported barrier to AI adoption mean in practice?
Each of the three headline barriers becomes a different implementation question.
A reported lack of expertise does not tell you which skill is missing. Ask whether the difficulty lies in choosing a task, connecting tools, checking outputs or maintaining the workflow. A team may know the process well and still need help with implementation. That is a question to investigate locally, not a conclusion from the survey.
For a legal-clarity concern, write down what the system will do, whose information it will use and who will receive the result. Have the relevant obligations checked for that use before launch. The reported percentage cannot decide which duties apply to your business.
The privacy figure also changes how an implementation should begin. An agent should not receive broad access because the team has not mapped the data. The team should identify the system of record, the fields required for the workflow, the actions that need approval, and the records that need logging.
What implementation sequence do the reported AI barriers support?
The reported barriers support a practical sequence of six steps, from naming one workflow to reviewing results against a baseline:
- Name one workflow and its measurable outcome.
- Identify the systems, data, owner, and failure modes.
- Decide which outputs remain drafts and which actions may run.
- Connect only the permissions needed for the approved scope.
- Test normal inputs, missing data, edge cases, and unsafe requests.
- Review the results against a baseline before expanding.
Say you want to prepare a weekly client report. Start with approved source records, require review before the report leaves the business and record corrections. This makes the implementation question specific without treating a survey response as evidence of savings.
The OECD discussion paper on SME adoption describes lower AI uptake among SMEs than larger firms and distinguishes pathways by digital maturity, complexity and application scope. Use those dimensions to describe your planned task; they do not select a workflow for you.[5]
Where is the implementation boundary in European adoption data?
European adoption statistics provide context. They do not answer the deployment question by themselves.
Eurostat does not report the percentage of European businesses running autonomous AI agents. It does not measure whether an agent has approval gates, whether a person checks the output, or whether the system writes to a CRM correctly. Eurostat also does not establish whether an enterprise achieved a return, recovered a specific number of hours, or reduced headcount.
The dataset has other boundaries:
- Enterprises with fewer than 10 persons employed, including self-employed people are outside the reported population.
- “AI technology” includes several categories that do not involve agents.
- Survey responses capture reported use, not an audited deployment.
- Country and sector comparisons can reflect different economic mixes.
- The Netherlands has a 2025 time-series break.
- Reported barriers are reasons given by enterprises that considered AI, not causal estimates for all non-users.
Those limits do not make the data weak. They make the data usable. A founder can use the figures to understand context, then ask a more specific question about the workflow they own.
The practical dividing line is not whether a business belongs to a high-adoption country or sector. The dividing line is whether the business can define a process clearly enough to test. The process needs an owner, a repeatable trigger, a measurable output, and an acceptable boundary for mistakes.
Yardwork's role begins after that question becomes concrete. Yardwork scopes the workflow, connects the agreed tools, defines approval and escalation points, launches against real data, and supports the system after go-live. The AI agent adoption statistics page covers broader adoption context, and the tables here let you compare the European figures for your business.
Method, calculations, and refresh policy
Use the dated downloads to reproduce a figure with the same population, year and denominator. Country and size values come from isoc_eb_ai; sector values come from isoc_eb_ain2. The tables use the frozen 2025 observations, annual frequency and Eurostat’s published percentages rather than estimates from chart pixels.[1][2][3]
| Field | Selection or meaning |
|---|---|
| freq | A |
| time | 2025 |
| geo | EU27_2020; BE; BG; CZ; DK; DE; EE; IE; EL; ES; FR; HR; IT; CY; LV; LT; LU; HU; MT; NL; AT; PL; PT; RO; SI; SK; FI; SE |
| indic_is | E_AI_TANY |
| unit | PC_ENT |
| size_emp | GE10, 10-49, 50-249, GE250 (isoc_eb_ai); GE10 (isoc_eb_ain2) |
| Activity scope | Covered non-financial activities; complete activity list in the download |
| value | Published percent; blank means missing |
| flag | Original Eurostat cell status retained |
Activity codes: the size aggregate uses NACE C10-S951_X_K; the full coded activity mapping is retained in the CSV.[1][2]
For technology rows, select the relevant E_AI_T code with PC_ENT. Purpose rows use the E_AI_P code and PC_ENT_AI_TANY. Barrier rows use the E_AI_B code and PC_ENT_AI_EC. The purpose, technology and barrier extract (CSV) preserves these separate denominators and size classes.[3]
How are percentage-point gaps calculated?
Subtract the two published percentages for a percentage-point difference. The large–small EU gap is 55.03 − 17.00 = 38.03 percentage points. Germany’s total minus the EU total is 25.97 − 19.95 = 6.02 percentage points. These are descriptive differences, not estimates of the effect of company size or location.[1]
Keep the published precision and flags. Percentages displayed to two decimals remain rounded survey estimates. Do not calculate a joint German-small-professional-services rate by multiplying the German size rate and the German sector rate: the two populations overlap in an unknown way. Do not average national percentages to reconstruct the EU aggregate.[4]
Which businesses are outside this comparison?
The common tables cover enterprises with at least 10 persons employed in the specified non-financial activities. Microbusinesses below that threshold, the financial sector and activities outside the survey coverage cannot be represented by these rates. Optional microbusiness observations elsewhere in the dataset do not change the population of these tables.[4]
When should the reference be refreshed?
Compare a new Eurostat release with the same dimension selections, then check definitions, flags and missing values before replacing a table. A new survey year and a revised observation are different changes. Review this reference annually or after a material data release; use the dated downloads when you need the figures behind this version.[4]
For your own workflow decision, pair the market comparison with an owner, representative inputs, an expected output and a review point. The country table can show your context; the trial should show whether the proposed work is useful.
Frequently asked questions
What percentage of European small businesses use AI?
Eurostat reported that 17% of small EU enterprises used at least one measured AI technology in 2025. The figure covers enterprises with 10 to 49 persons employed, including self-employed people and does not measure autonomous AI agents specifically.[1]
Which European countries use AI most often?
In Eurostat's 2025 comparison, Denmark had the highest share of enterprises using at least one measured AI technology at 42.03%, followed by Finland at 37.82% and Sweden at 35.04%. Country comparisons need the survey definitions and the Netherlands time-series break.[1]
Does Eurostat's AI adoption data measure AI agents?
No. Eurostat measures several AI technologies and purposes, including text mining, language generation, workflow automation, and decision support. The dataset can inform the market context for AI-agent implementation, but it cannot be reported as an AI-agent adoption rate without additional evidence.[4]
Why are European businesses not using AI?
Among EU enterprises with 10 or more persons employed that had ever considered using AI technologies, the most common reported reasons for not using it in 2025 were lack of relevant expertise at 70.31%, unclear legal consequences at 53.61%, and data-protection or privacy concerns at 52.72%.[3]
How does AI adoption differ between small and large European firms?
Eurostat measured 17% for small, 30.36% for medium and 55.03% for large EU enterprises in 2025. The large–small difference is 38.03 percentage points. These descriptive rates do not explain the cause or predict an individual company’s outcome.[1]
Which European sectors use AI most often?
Among the broad activities in the 2025 comparison, information and communication reached 62.52%, followed by professional, scientific and technical activities at 40.43%. Detailed activity categories and overlapping aggregates are listed separately in the full sector tables.[2]
Notes
- Eurostat size observations, accessed October 14, 2026. Dataset updated June 15, 2026; 2025 observations extracted October 14, 2026.
- Eurostat activity observations, accessed October 14, 2026. Dataset updated June 15, 2026; 2025 observations extracted October 14, 2026.
- Eurostat purposes, technologies and barriers, accessed October 14, 2026. Dataset updated June 15, 2026; 2025 observations extracted October 14, 2026.
- ICT enterprise survey methodology, accessed October 14, 2026.
- OECD, AI adoption by small and medium-sized enterprises, accessed October 14, 2026.
- Eurostat, Use of artificial intelligence in enterprises, accessed October 14, 2026.
