Table of contents
Key facts
- Poles use AI at roughly the same rate as the rest of Europe. The AI diffusion rate for Poland was 28.5% in the second half of 2025, against a European average of about 27% and 28.3% in the United States.
- Polish companies rank second from last in the European Union. In 2025, 8.36% of enterprises with 10 or more employees used at least one AI technology, against an EU average of 19.95%. Only Romania sits lower.
- The smaller the company, the wider the gap. Among large enterprises it is 45.81% against 55.03% in the EU, among medium ones 14.77% against 30.36%, and among small ones 5.79% against 17.0%.
- 90.2% of Polish companies use no AI technology at all, and the share that has ever even considered using it stands at 3.9% against 11.1% in the EU.
- The labour market tells a completely different story. In 2025, 2.92% of all job postings in Poland required AI skills, more than in the United States (2.56%) and the United Kingdom (1.93%).
- 2026 brought the legal framework. The Polish Act on Artificial Intelligence Systems has been in force since 11 August 2026, with its main part applying from 28 October 2026. The EU AI Omnibus moved the obligations for high-risk systems listed in Annex III to 2 December 2027.
Stanford HAI has published the AI Index Report 2026, the most comprehensive annual collection of hard data on artificial intelligence. Poland appears in it more than a dozen times and comes out respectably almost every time. The catch is that this report measures people above all: users, employees, graduates, job candidates. Add Eurostat data on enterprises to it and the picture splits into two halves that have very little to do with each other.
The world in 2025: adoption stopped being an experiment
According to the McKinsey survey cited in the AI Index, 88% of organisations used AI in at least one business function, against 78% a year earlier. For generative AI the figure was 79% against 71%. More than half of respondents pointed to three or more areas of the company where AI is already at work.
Europe recorded one of the highest year-over-year increases, 11 percentage points, and at 91% it moved ahead of every other region in the world. Two years earlier, in 2023, the figure was 57%. This is not a trend anyone gets to sit out.
On the consumer side it went even faster. Generative AI reached 53% population adoption in three years, faster than the personal computer and faster than the internet at the same stage.
There is one thing in this data worth remembering before planning anything for 2026. AI agents, meaning systems that carry out tasks on their own, are still marginal. Across most business functions a majority of respondents do not use them at all, and scaled deployments sit in single-digit percentages almost everywhere. The highest rates are in the technology sector: 24% in software engineering, 22% in IT, 21% in service operations. 2025 was the year of the chat window, not of process automation.
Poland in the AI Index: people and skills look good
The AI diffusion rate, calculated by the Microsoft AI Economy Institute from telemetry data, reached 28.5% for Poland in the second half of 2025. That is a rise of 2.1 percentage points in six months and 22nd place among the thirty largest economies. For comparison, the European average is about 27%, while the United States, at 28.3%, sits in 24th place. The leaders are the United Arab Emirates (64.0%), Singapore (60.9%) and Norway (46.4%).
The remaining Polish figures in the report add up to a consistent picture of a country that has something to work with:
| Indicator | Poland | Context |
|---|---|---|
| Share of job postings requiring AI skills (2025) | 2.92% | more than the US (2.56%) and the UK (1.93%); leader Singapore 4.69% |
| AI hiring growth relative to the whole market (2025) | 12.32% | 10th place in the ranking |
| AI talent concentration among LinkedIn members | 1.00% | in the top fifteen; leader Israel 2.10% |
| Relative AI skill penetration | 1.14 | above the global average, which stands at 1.00 |
| New ICT bachelor’s graduates (2023) | 14,688 | up from 12,817 in 2022; more than Australia, Canada and France |
| New ICT master’s graduates (2023) | 4,571 | up from 4,164 in 2022 |
| Data centres (2025) | 144 | in the top fifteen worldwide |
| Public AI-related contracts in Europe (2013-2024) | 162 | 4th in Europe, ahead of France (161) |
Attitudes add to that. According to the Ipsos survey quoted in the report, 61% of Poles trust their own government to regulate AI responsibly, against a global average of 54% and 31% in the United States, the lowest score in the whole sample. Asked about priorities, 67% of Poles choose advancing science and medicine through AI over protecting industries through regulation, against 58% globally. That is one of the highest results in the entire survey.
In other words: we have the people, we have the graduates, we have stronger demand for AI skills than the United States, and we have a society leaning towards progress rather than braking. Now let us look at what companies do with all that.
Poland in Eurostat data: companies are far behind
Eurostat measures something different from Stanford: not people, but enterprises. In 2025, 8.36% of Polish companies with 10 or more employees used at least one AI technology. The EU average is 19.95%. A year earlier the figures were 5.9% and 13.48% respectively, so we are growing, but more slowly than the EU, and the gap is widening rather than closing.
Among the twenty-seven member states Poland ranks 26th. Only Romania, at 5.21%, comes lower. Bulgaria has 8.55%, Greece 8.93%. At the other end of the table sit Denmark (42.03%), Finland (37.82%) and Sweden (35.04%). Germany has 25.97%, Czechia 17.6%.
The smaller the company, the wider the gulf
This is the most important table in the whole set, because it shows exactly where the problem sits.
| Company size | Poland | European Union | Difference |
|---|---|---|---|
| 250 employees and more | 45.81% | 55.03% | 9.2 pp |
| 50-249 employees | 14.77% | 30.36% | 15.6 pp |
| 10-49 employees | 5.79% | 17.0% | 11.2 pp |
Large Polish companies are keeping up. They are nine points behind the European average, but they play in the same league. Medium-sized companies use AI more than twice as rarely as their European counterparts. Small companies almost three times as rarely. This is not a problem of the Polish economy as a whole, but specifically of the Polish small and medium enterprise sector.
What Polish companies actually use AI for
| Technology | Poland | EU |
|---|---|---|
| Generating text, speech or code | 5.23% | 8.76% |
| Generating images, video or audio | 2.75% | 9.55% |
| Process automation and decision support | 2.60% | 5.35% |
| Written language analysis (text mining) | 2.10% | 11.75% |
| Image recognition | 1.84% | 3.78% |
| Machine learning for data analysis | 1.83% | 5.10% |
| Speech recognition | 1.59% | 7.22% |
The biggest difference is not in the flashiest technologies, but in the ones most useful in everyday work. Text analysis, meaning the automated reading of contracts, tickets, product descriptions and correspondence, is used by 11.75% of companies in the EU and 2.10% in Poland. More than a fivefold difference on something available off the shelf today.
The way it gets deployed says a lot too. Ready-made commercial software used without any changes accounts for 6.18% of all companies, meaning close to three quarters of those that reached for AI at all. Building or modifying a solution was outsourced to an external provider by 1.91% of companies, while 2.11% built something with their own staff. AI in Polish companies is mostly a subscription, not an implementation.
Broken down by purpose: marketing and sales 4.93% (EU 6.92%), business administration and management 3.58% (EU 6.19%), accounting and finance 2.57% (EU 4.61%), production processes 2.51% (EU 4.14%), ICT security 2.20% (EU 3.94%), research and development 2.04% (EU 3.80%), logistics 0.84% (EU 1.21%).
The barrier is not where people look for it
Eurostat asks companies what holds them back. The Polish answers look calm at first glance: 2.44% of companies point to a lack of relevant expertise, 2.17% to costs being too high, 2.04% to unclear legal consequences, 1.93% to data protection concerns. In the EU the same barriers are reported by 7.76%, 4.24%, 5.92% and 5.82% of enterprises respectively.
This does not mean Polish companies have it easier. It means fewer of them ever got to the point where a barrier appears. One figure confirms it: the share of companies that have ever considered using AI stands at 3.9% in Poland and 11.1% in the EU. Before a problem with cost, skills or legal compliance shows up, someone has to sit down and work out whether it pays off. In Poland that conversation most often never happens.
What is happening in 2026
The Polish AI act is already in force
The Act of 3 July 2026 on Artificial Intelligence Systems was published on 27 July 2026 (Journal of Laws 2026, item 1003). Its core entered into force on 11 August 2026, while articles 8-18 and chapters 3-5, 8 and 9 start applying on 28 October 2026.
The act creates the national framework for applying EU Regulation 2024/1689 and establishes a new body: the Commission for the Development and Security of Artificial Intelligence, known as KRiBSI. It is the market surveillance authority for AI systems in Poland and the single point of contact.
The EU gave more time on high-risk systems
On 27 July 2026 the so-called AI Omnibus entered into force, having been published in the Official Journal of the EU on 24 July 2026. It moves the deadlines for high-risk systems: the rules covering stand-alone systems from Annex III will apply from 2 December 2027 instead of 2 August 2026, and those for AI embedded in products regulated by separate legislation from 2 August 2028.
For most small and medium companies this means that the tools they use day to day, meaning text assistants, customer service chatbots or document automation, do not fall into the high-risk category and there is nothing to wait for. The postponement covers recruitment, credit scoring, education, critical infrastructure and similar uses.
Does it actually pay off
The AI Index gathered the productivity research and the result is uneven, but it is not zero. The gains are largest where the work can be measured: 14-15% in customer support, 26% in software development, 50% in the volume of marketing output. In tasks that require deeper judgement the effects are weaker.
The most interesting one for a European company is the study by Aldasoro and co-authors from 2026, covering 12,000 European enterprises between 2019 and 2024. AI adoption raised labour productivity there by 4%, with no drop in employment in the short run. The second result of that study is the crucial one, though: every 1% of budget spent on training delivered an additional 5.9 percentage points of gain. Simply buying licences makes no difference.
The other side is worth knowing too. A survey of 6,000 executives across four countries showed widespread adoption alongside minimal measured productivity impact so far. One third of organisations expect AI to reduce their headcount within a year, though almost half expect no change at all. In the United States, employment of software developers aged 22 to 25 fell by nearly 20% compared with 2024. The number of documented AI-related incidents rose to 362 from 233 in 2024.
What follows from this
Poland has the graduates, it has a labour market that asks about AI more often than the American one, it has people using these tools at the European average level, and since August 2026 it has a national legal framework. The only thing missing is implementations inside companies. Nine out of ten Polish enterprises use no AI technology at all, and among companies with 10 to 49 employees more than 94% use none. The difference between Poland and Denmark does not come from access to technology, which is identical, but from the fact that over there someone sat down and worked out what specifically in the company can be automated.
At Web Systems we work on exactly that part: we do not sell subscriptions, we build AI-based solutions wired into what a company already runs, meaning its shop, ERP system, document flow or customer service. We have been running the company since 2006, there are six of us and we maintain our own infrastructure across dozens of machines, so an implementation here usually ends with maintenance rather than a handover of files. If you want to check whether there is anything in your company worth automating, see how we handle the development of artificial intelligence applications, or write to us through the contact form. The initial quote is free.
Sources
- Stanford HAI, Artificial Intelligence Index Report 2026 (data from McKinsey, Microsoft AI Economy Institute, Lightcast, LinkedIn, OECD, Ipsos, Cloudscene).
- Eurostat, dataset isoc_eb_ai, use of artificial intelligence by enterprises, 2025 data (updated 15 June 2026).
- Act of 3 July 2026 on Artificial Intelligence Systems (Journal of Laws 2026, item 1003).
- European Commission, AI Omnibus enters into force.
