Poland is now home to 19,000 companies associated with artificial intelligence, according to data from Dun & Bradstreet. This means that one in every 12 businesses in the country’s IT sector is officially labelled as operating in AI.
In the first half of 2026 alone, 6,000 new entities offering AI-related solutions were registered.
These figures, however, primarily reflect the supply side of the market—the determination of providers to offer fashionable services. Changing a company’s official business activity classification to include AI costs nothing and does not require it to have a product, a customer or any revenue.
A large proportion of the market also consists of sole proprietorships and programmers working under business-to-business contracts. The real test for the sector therefore lies on the demand side.
Companies want to invest, but few are ready to transform
An analysis of demand reveals a clear gap between corporate declarations and companies’ actual management maturity.
Although 77% of medium-sized and large businesses in Poland intend to increase their AI spending over the next 18 months, only one in 10 is using the technology to redesign its business model. The remaining companies are effectively paying for a faster version of what they were already doing, according to a 2026 EY study.
The proportion of businesses with a clearly defined role for AI in their strategy has increased to 35%. At the same time, their readiness to implement it has fallen to 10.2%.
In other words, more companies understand why they want to use AI than know how to deploy it effectively, according to the February 2026 report The Artificial Intelligence Market in Poland, prepared by PMR Market Experts by Hume’s.
The main driver of investment is not currently a carefully calculated business case, but fear of falling behind and pressure from competitors. Businesses are investing primarily in ready-made tools rather than in their own organisational capacity to use them effectively.
Only 18% of AI investments have generated a profit
This lack of preparation is already producing measurable consequences.
Only 18% of AI investments have generated a profit for the companies implementing them, while 37% have merely covered the cost of deployment, according to the June 2026 Human Capital Balance study conducted by the Polish Agency for Enterprise Development and Jagiellonian University.
This is a classic example of the so-called last-mile problem.
Building a prototype represents only around 10% of the total effort. The remaining 90%—integrating AI into teams’ everyday work—is where projects most often fail.
The problem is not the technology itself. The main causes are inadequate data readiness, low organisational maturity and poorly defined use cases.
There is also a significant human factor. Business leaders often fail to address employees’ entirely understandable resistance, particularly in the context of widely publicised mass redundancies justified by the introduction of new technologies.
As a result, as many as 80% of Polish chief executives identify implementation capabilities as the greatest barrier to further AI development.
Global AI developers are moving directly into implementation
Poland’s experience closely reflects broader global trends.
The world’s largest AI model developers, including Anthropic and OpenAI, have already recognised the implementation gap. They are building their own implementation businesses and placing engineers directly within client organisations.
The fact that the creators of the technology themselves are investing in people whose job is to complete the implementation process is perhaps the clearest indication that customers are often unable to manage it independently.
Providers can still charge prices disconnected from results
Businesses’ lack of implementation capabilities means that the current market—where AI service providers can charge prices that are largely disconnected from measurable outcomes—remains strong.
More than half of the consulting and implementation services market is controlled by the largest global players, including IBM, Accenture and Deloitte. Smaller integrators and freelancers share approximately 25% of the market between them.
Globally, increasing competition is gradually creating pressure for providers to be paid according to the results they deliver. This trend has not yet reached Poland’s AI implementation market.
As long as demand continues to be driven by fear, providers can maintain pricing models that are not directly connected to business outcomes. This market situation, however, is unlikely to last indefinitely.
Companies should treat an AI agent like an employee
To escape this trap and create genuine value, companies must change their approach and begin treating an AI agent like an employee—one that requires context, a clearly defined objective and measurable success criteria.
Practical implementation should begin with narrow, specific processes in which eliminating repetitive work can quickly produce a measurable result.
Companies must also organise their decision-making architecture, responsibilities and data at the source.
Otherwise, the cost of organisational chaos will continue to rise, particularly amid the market trend of “tokenmaxxing”—maximising the use of computational tokens—and growing pressure on major AI companies such as OpenAI and Anthropic to increase revenue ahead of their anticipated stock market debuts.
Ultimately, artificial intelligence does not create an entirely new category of business problems. It merely makes the problems that companies have neglected for years more visible—and more expensive.
Author: Agnieszka Jórczyk, founder of Zenergetic and creator of the “Founder 2.0” series
