Polish Companies Spend Millions on AI Without a Strategy, Forbes.pl Analysis Shows

Polish companies are investing more boldly in artificial intelligence, but increasingly without a clear goal. An analysis published on September 3, 2026 by Forbes.pl shows that even AI implementation spending running into hundreds of thousands, sometimes millions, of zloty fails to translate into real efficiency gains, because companies should first get their data and processes in order before turning to algorithms.

Chaos Instead of Savings

The report's author, Dorota Kaczyńska, describes a pattern that repeats across many Polish organizations: management boards buy access to language models or commission dedicated implementations, expecting an immediate boost in competitiveness. In practice, the tool lands on disorganized data and chaotic processes, so instead of fixing them, it only speeds them up.

If it receives chaotic data and disorganized processes, it will only generate a better organized chaos - Piotr Kawecki, CEO of exeAI

Kawecki adds that implementation failures rarely stem from the quality of the models themselves. In his view, companies lose out because they lack a coherent data and process architecture for AI to even connect to, an investment in the tool without that foundation turns into a cost with no return.

The Work Slop Phenomenon

The article introduces the concept of work slop, a situation in which AI generates professional-looking documents, reports, or content that still require tedious human verification before they can actually be used. Instead of saving employees' time, this work consumes it, because someone has to check, correct, and refine the algorithm's output.

Joanna Adamiak, CEO of PR agency Face it!, points to the limitations of AI in creative work. In her view, text prepared by a person remains more natural and resonates better with audiences than automatically generated content, which limits the scope for AI in communication and marketing, despite growing pressure to automate these processes.

Efficiency That Fades Over Time

Jaromir Sroga, CEO of media company re58, describes a typical implementation curve: at first teams see a clear jump in productivity, but over time the momentum of those gains weakens by 10-20 percent. In his view, AI speeds up many processes, but it is not a universal solution and sometimes leads teams into a dead end they have to retreat from.

This isn't a solution that will answer all problems. It speeds up processes, but sometimes leads into a dead end - Jaromir Sroga, CEO of re58

Andrzej Miron, head of IT at Nationale-Nederlanden, believes organizations still face plenty of trial and error before they develop a mature approach to AI. After the first phase of enthusiasm and euphoria comes a harder, more critical stage of verifying whether the tools actually deliver business value, or merely generate new operational and training costs.

The Five Percent Paradox and Shadow AI

The analysis also describes the so-called 5 percent paradox: if even a small fragment of a business process is poorly automated, the entire transformation can fail, because the error propagates through subsequent stages of work. The problem is compounded by the quality of source data, when AI receives faulty input, it returns false information with the same confidence as correct output.

A separate risk is the so-called shadow AI economy. According to the figures cited, nine out of ten employees use tools such as ChatGPT without the knowledge of their IT department, creating a risk of confidential company data leaking into public models and making it harder for companies to control what information leaves the organization.

Adam Pawluć, vice president of the Polish Biogas Group (Polska Grupa Biogazowa), sums up the state of readiness of Polish business bluntly, few companies today are truly prepared to work effectively with artificial intelligence. What's missing isn't just infrastructure and organized data, but also the skills to manage AI projects at an organizational level, not just a technical one.

What This Means for Polish Companies

For managers planning budgets for 2027, the conclusions of the analysis are concrete: spending several hundred thousand zloty on an AI implementation without first getting data and processes in order has, according to the cited MIT and BCG research, a significantly higher chance of ending in failure than success. Instead of starting with the purchase of a tool, companies should first assess the quality of their own data and the maturity of the processes they intend to automate.