Artificial intelligence is one of the hottest topics in Polish business today. It features at conferences, in strategies, presentations, LinkedIn posts and discussions about productivity. The problem is that there is still a clear gap between market rhetoric and corporate practice. Poland currently has no trouble generating interest in AI. It does, however, struggle to translate that interest into processes, accountability and results.
According to Eurostat data, in 2025, 20 per cent of businesses in the European Union were using AI technology. In Poland, the figure was 8.4 per cent. The Central Statistical Office (GUS), using its own national methodology, paints a similar picture: 8.7 per cent of firms reported using AI. This represents an increase on the previous year, but it does not alter the key conclusion. Poland still finds itself closer to the bottom of the European rankings than to the middle.
It would be easy to interpret this result as a simple story of technological backwardness. However, that would be too convenient and not entirely true. Poland has a strong IT sector, a large pool of engineering expertise, well-developed service centres, a dynamic e-commerce sector, digital banking and companies capable of building complex technological systems. The paradox is that a country capable of creating technology does not necessarily implement it widely across the economy at the same pace.
This is an organisational gap.
The biggest difference is not between companies that have heard of AI and those that haven’t. It lies between those that are experimenting and those that are able to turn an experiment into a repeatable operating model. Many organisations have already carried out their first trials: AI-generated text, the automation of simple analyses, chatbots, and support for sales, customer service, marketing or HR. However, a pilot is not yet a roll-out. And a roll-out is not yet a transformation.
Strategy only begins when AI is assigned to a specific process, an owner and a metric. Does it reduce customer service times? Does it reduce the number of errors? Does it improve conversion rates? Does it reduce operating costs? Does it enable faster decision-making? If the answer is simply ‘employees are using the tool’, the company does not yet have an AI strategy. It has a new layer of productivity, often beyond the organisation’s full control.
This leads to one of the most interesting phenomena of recent months: AI is spreading through companies from the bottom up faster than through formal channels. In their Work Trend Index survey, Microsoft and LinkedIn showed that a large proportion of employees using AI bring their own tools to work. This is not a marginal technological issue, but a cultural signal. Employees have already recognised its usefulness. Organisations are only just trying to define the rules.
This is how ‘shadow AI’ emerges. On the one hand, it is evidence of people’s willingness to experiment. On the other, it highlights the weakness of formal processes. Someone summarises a document using a public tool. Someone analyses customer data. Someone generates a response for a business partner. Someone drafts part of a strategy without revealing that a language model helped them. Each of these instances can boost productivity. Each can also create a risk: a data leak, a wrong decision, a breach of compliance rules, or a loss of control over the company’s knowledge.
That is why Poland’s AI gap is not solely down to the fact that too few companies use artificial intelligence. It also stems from the fact that some use it without a management framework. AI within a company can be both too little and too much: too little in processes, too much in informal practices.
The difference between large companies and SMEs further exacerbates this picture. Eurostat data show that, across the EU, AI is implemented much more frequently by large enterprises than by small ones. A similar pattern can be seen in Poland. According to data from the Central Statistical Office (GUS), in 2025, 42 per cent of large firms, 15.6 per cent of medium-sized firms and only 6.1 per cent of small firms were using AI technologies. This means that the real adoption gap is primarily a gap among smaller organisations.
Large companies have the budgets, teams, data and suppliers, but they often struggle with complexity. Decision-making is protracted, responsibility is fragmented, processes are cumbersome, and every integration requires the approval of several departments. Smaller firms have speed on their side, but they lack time, expertise, straightforward use cases and the confidence that the investment will pay for itself quickly. As a result, some get bogged down in governance, whilst others struggle with a lack of resources.
Expertise is one of the main constraints here. PARP points out that only around one in four companies report having access to specialists capable of implementing and maintaining AI solutions, whilst almost half are not taking any action to address these shortcomings. But this is not solely a problem of a shortage of data scientists either. In many companies, there is rather a shortage of people who can integrate technology with business processes. AI does not need only model experts. It needs process owners who know where costs, delays, errors or missed opportunities arise.
Global reports point in a similar direction. McKinsey notes that companies deriving greater value from AI do not merely roll out tools, but also redesign their workflows. PwC points out that the lion’s share of the benefits from AI is concentrated in a relatively small group of companies. It is therefore not ambition alone, nor the number of pilot projects, that makes the difference. What makes the difference is the ability to translate AI into the way work is done.
This is important because in Poland, the discussion about AI often remains at the level of mere declarations. The focus is on the tool, rather than the process. It centres on training, rather than a shift in responsibilities. It revolves around automation, rather than who takes responsibility for the outcome of a model-supported decision. Meanwhile, the AI Act, rising customer expectations and pressure for efficiency will shift this discussion towards governance, auditability and measurable business impact.
Successful implementations do not necessarily have to start with a major transformation. On the contrary, they often begin with a single, well-chosen process. Handling customer enquiries. Document analysis. Searching for internal information. Preparing quotations. Call centre support. Automation of repetitive tasks in finance or HR. The difference is that the process has an owner, the data is organised, the risks are identified, and the results are measured after a few months.
Poland has the potential to utilise AI much more widely than current statistics suggest. But potential is not yet a competitive advantage. A competitive advantage only emerges when technology begins to change the pace of an organisation. When it shortens the path from data to decision. When it relieves people of low-value work. When it enables faster customer service, better operational planning or a reduction in the cost of errors.
The greatest risk today is not that Polish companies aren’t talking about AI. They’re talking about it a great deal. The risk is that talk will replace implementation, pilot schemes will replace strategy, and the bottom-up use of tools will be mistaken for transformation.
