Artificial intelligence in Polish companies. What is missing in the debate on AI implementations?

Artificial intelligence is increasingly bold to enter Polish companies – not only as a tool, but besides as a origin that changes the structure of employment, decision-making models and regular work practices. From Ernst&Young reports (of 2023 and benchmarking year to year from 2025) and Deloitte emerges a marketplace image at a crossroads: on the 1 hand, companies see real benefits from AI, on the another hand, they face challenges.

The data shows that companies face respective common barriers – both technological and organisational.

Common concerns between sectors

The study, which was commissioned by EY Poland and performed by the agency Cube Research, included 501 companies in the first edition in 2023. His analysis year to year, repeated in the last 4th of 2024, again included 501 companies from the same industries and confirmed the previously recognized trends and trends. Although EY 61% of companies declare good or very good cognition of artificial intelligence regulations, safety – especially data – inactive raises crucial concerns.

In the 3 sectors analysed (services, trade, production) companies consistently point to information safety issues as 1 of the 3 main challenges in implementing AI.

This concern stems, among others, from the uncertainty about how external AI tools process data, who has access to it and whether the organization's IT infrastructure is ready for it. This applies in peculiar to client individual data and interior data which may be unwittingly shared in cloud-based AI tools. There is besides a deficiency of clear practices that regulate the usage of specified tools on their own.

Importantly, although the awareness of threats is comparatively high, the EY study does not state whether real actions are following it – specified as investments in security, access control or the creation of corporate AI standards.

Such a gap between declarations and the strategy may in practice imply an increased hazard of data leakage and non-compliance with the requirements of European regulations specified as the AI Act.

The costs associated with the implementation of the AI stay 1 of the main barriers, especially in the commercial sector, where nearly half of the companies indicate them. advanced licence prices, the deficiency of ready-made models tailored to the company's specifics, as well as the request to grow IT infrastructure make it an investment inactive out of scope for many AI organizations.

In turn, technological difficulties, which indicate about 31% of companies, are mainly about the deficiency of compatibility with existing systems, limited access to appropriate training data and the deficiency of specialists that can effectively integrate fresh solutions into the existing IT ecosystem. AI is not a ready solution – it requires method preparation, which is frequently lacking.

Changes in organisational structure

While the implementation of the AI creates fresh roles and needs for professionals, it besides leads to major changes in the employment structure. The study shows that in the trade sector already 36.6% of companies limited recruitment to junior positions, and more than half of them plan specified restrictions in the future.

This phenomenon is revealed by the paradox: companies urgently request workers with experience in AI, but at the same time they restrict access to the first professional roles in which specified experience could only develop. In the long term, this may lead to a shortage of qualified experts – especially if the juniors do not have a place to practice.

As a result, AI not only changes operating processes, but besides destabilizes previously known career paths – especially for younger workers and graduates entering the market.

Optimism Despite Barriers: Most Companies See Benefits

Despite the challenges, around 80% of companies that have already implemented AI tools declare their intended effects. Among the main benefits most frequently mentioned are the automation of regular processes, faster decision-making through data analysis, improved operational efficiency and relief of teams in areas requiring advanced manual work.

It is worth noting that both the baseline EY study prepared in 2023 and its update published in 2025, are based on respondents' declarations – not on verified operational or financial data. The conclusions on the effectiveness of implementations and future plans so reflect, above all, the moods and subjective assessments, not the actual results or actions already taken.

Moreover, the EY study shows that only 28% of companies associate AI-related activities with the overall organisational strategy. In another cases, decisions are made locally – within circumstantial departments specified as marketing, IT or HR (44% of companies) – or are ad hoc and uncoordinated (24%). specified distribution indicates a deficiency of uniform approach and long-term vision, making it hard to implement AI liable and sustainable – regardless of the level of method readiness.

This is an area that needs further monitoring – besides so that public debate does not miss the everyday reality of utilizing AI at work.

Sources:

Ernst&Young: How do Polish companies implement AI? Analysis of changes year to year [accessed 24.10.2025]
Ernst&Young: How Polish companies implement AI [accessed 24.10.2025]
Deloitte: Europeans are optimal about generic AI but there is more to this close the trust gap [accessed 24.10.2025]Deloitte: The State of Generative AI in the Enterprise. 2024 year-end Generic AI report [accessed 24.10.2025]
The European Parliament and the Council of the European Union, This Regulation shall enter into force on the twentieth day following that of its publication in the authoritative diary of the European Union., OJ L 2024/1689, 12.07.2024