Although more and more companies are using artificial intelligence-based tools, Poland still remains below the EU average in terms of AI implementation in business. Experts point out that the biggest barrier is not access to technology, but a lack of knowledge about how to use it to solve specific business problems. At the same time, they warn about the growing concentration of the AI market in the hands of a few global companies and Europe’s weak position in developing its own models.
“Indicators relating to the implementation of artificial intelligence in Polish companies are several years behind what is happening in Western Europe, not to mention the United States. This is despite the fact that, as private users, Poles make very extensive use of language models,” said Dr. Piotr Sankowski, professor at the University of Warsaw and Director of the IDEAS Research Institute, in an interview with Newseria.
According to Eurostat data, in 2025 artificial intelligence was used by 20 percent of enterprises in the European Union employing at least 10 people. Among EU leaders such as Denmark and Finland, the figure was around 40 percent, while in Poland it stood at 8.4 percent, placing the country among those with the lowest levels of AI use. A year earlier, the share was 5.9 percent, showing that interest in the technology is growing, although the pace of implementation remains clearly below the EU average. The European Commission assumes that by 2030 at least 75 percent of enterprises will be using AI, cloud computing or data analytics.
“Before implementing certain solutions, companies are primarily held back by a lack of trust that these tools can actually be useful. There is still a lack of education and understanding, but also a lack of a partnership-based approach from companies implementing artificial intelligence,” Prof. Piotr Sankowski stressed. “This is not about selling licences, but about ensuring that the company where the solution is implemented actually benefits in business terms.”
According to McKinsey & Company’s global report, “The State of AI 2025,” although 88 percent of surveyed companies use artificial intelligence in at least one business function, two-thirds are still at the experimentation or pilot stage. Only around one-third of companies say they have implemented AI at greater scale in their operations. The authors of the report indicate that most organisations are not using these tools in a sufficiently advanced way to achieve measurable benefits at enterprise level.
“There is a lack of understanding on the industry side of how to use artificial intelligence. We should start with the problems and then consider which solutions are best suited to solving them. Not every problem requires large language models. Most can be solved with sovereign solutions developed in Poland. Further education in this area is needed,” said the Director of the IDEAS Research Institute.
The artificial intelligence market is moving increasingly beyond chatbots and content generation. So-called agentic systems, or agentic AI, which can independently plan and perform complex tasks using multiple tools at the same time, are playing a growing role. According to OpenAI’s study published in June, “The Shift to Agentic AI: Evidence from Codex,” the number of active users of such solutions increased more than fivefold in the first half of 2026. They are being used increasingly by businesses, not only by programmers. McKinsey & Company’s research also points to strong interest in AI agents: 23 percent of companies say they are scaling this technology in at least one business function, while a further 39 percent say they have begun experimenting with it.
“How will the artificial intelligence market change? There is, of course, a risk that it may be controlled by a few players. Already now, the largest models come from only a few producers. There are not many companies capable of investing such large amounts of money,” Prof. Piotr Sankowski noted. “In Europe, there is a lack of companies developing large language models. There is Mistral, but unfortunately it operates on a smaller scale due to limitations in the financial market and the availability of venture capital. It is much harder for it to compete on equal terms with large American corporations.”
According to the “AI Index 2026” report prepared by the Stanford Institute for Human-Centered AI, private investment in artificial intelligence in the United States reached around USD 285.9 billion in 2025. This was 24 times more than in China, where investment amounted to USD 12.4 billion, although these figures do not include government support for the technology. In Europe, investment totalled less than USD 21 billion. The United States also clearly led in terms of the number of companies developing AI. The report indicates that although most models originate mainly from the US and China, other countries are seeking to increase control over their AI ecosystems.
Access to the most advanced AI models increasingly depends not only on financial capacity, but also on decisions made by producers and even by national authorities. In the spring, Anthropic restricted access to its most advanced cybersecurity model, Claude Mythos Preview, to selected organisations participating in the Project Glasswing programme. As the company explained, the model was not yet ready for broad release due to the risk of misuse and the need to develop more effective safeguards. In June, OpenAI announced that, at the request of the US administration, it was restricting access to its latest GPT-5.6 model, which was initially made available only to a narrow group of trusted partners as part of a safety testing programme. The restrictions were lifted on 9 July after approval was obtained.
“Just a year ago, everyone had access to the latest model — they could simply pay for it and use it for their own purposes,” said the Director of the IDEAS Research Institute.
