More Than 5 Million Polish Jobs Are Exposed to Generative AI

In 2025, Poland received its most detailed assessment yet of how generative artificial intelligence may affect employment. NASK – Poland’s National Research Institute – together with the International Labour Organization (ILO), estimated how many jobs in Poland are exposed to GenAI, using the ILO–NASK index based on task-level analysis of 427 occupations under the ISCO-08 classification. At the same time, humanoid robots in China are moving from demonstrations to early production and warehouse deployments, while the EU AI Act is introducing phased obligations for transparency, governance and human oversight of high-risk systems. Below is a breakdown of the numbers, timelines and mechanisms – based on published data and sourced analysis.

1. How many jobs in Poland could GenAI change?

According to the report „Generative Artificial Intelligence and the Polish Labor Market” (NASK, ILO, June 2025), based on Labor Force Survey (BAEL) microdata for Q4 2023 and the first three quarters of 2024, about 30.3 percent of all jobs in Poland – roughly 5.08 million positions – show at least some exposure to automation or task transformation by GenAI.

The highest-exposure group is far narrower: about 817,500 people (4.9 percent of the employed) work in occupations where most tasks could potentially be performed or substantially transformed by generative AI. The report’s authors explicitly caution that task exposure does not mean the automatic elimination of an occupation: more often, it points to changes in job content rather than disappearance from the labor market.

Clerical support workers remain the most exposed group – more than 71 percent of them perform tasks susceptible to GenAI, and nearly half (47.2 percent) fall into the very-high-exposure category. Specific occupations cited in the report include data-entry clerks, bookkeeping staff, administrative secretaries, proofreaders, telemarketers, financial analysts, mobile app developers, and web designers and editors. The ILO notes that while clerical occupations remain the most exposed, exposure is also rising among many specialist and highly digitized professions.

2. Women vs. men – who is more exposed?

GenAI’s impact is not evenly distributed. According to NASK–ILO, about 39.1 percent of employed women work in occupations exposed to automation or transformation, versus 22.8 percent of men – equivalent to roughly 3.02 million positions held by women and 2.05 million held by men.

The gap is especially pronounced among the youngest workers – in the 15–24 age group, the report finds that 47.8 percent of employed women work in GenAI-exposed occupations, versus 22.7 percent of men. The disparity holds across all age groups, which the authors attribute to employment structure: women work more often in clerical, administrative and service roles, and less often in physical occupations.

3. How often Polish companies actually use AI

Potential task exposure to automation is far greater than the current scale of deployment. In the NASK–ILO survey (December 2024), only 9.4 percent of workers said their employer had officially rolled out GenAI tools; 19 percent cited plans to do so, while 44.5 percent said their company had no such plans. Guidelines for safe use are also scarce – 68.1 percent of workers said they had received no guidance on GenAI use, and only 8.1 percent confirmed such rules existed at their company.

More recent, harder data comes from Eurostat: in 2025, 8.4 percent of Polish companies with at least 10 employees used AI technology, up from 5.9 percent a year earlier. The EU average reached 20 percent in the same period, with leaders Denmark and Finland at around 40 percent. Poland remains among the EU countries with the lowest level of corporate AI adoption.

4. How quickly could the changes happen?

None of the cited studies points to a single date when AI would trigger mass job losses. NASK–ILO does, however, show how workers themselves assess the outlook over the next five years: among those in highly exposed occupations, 29.8 percent expect significant or complete transformation of their job, versus 17.1 percent among those in occupations with low GenAI exposure. A third of workers in highly exposed occupations believe they will need to adapt their skills, and 58.4 percent of all working Poles say they are willing to learn AI-related skills.

5. A separate front of automation: robots in factories and warehouses

GenAI and physical robotics are two distinct automation mechanisms. NASK–ILO data concerns primarily software capable of generating and processing text, images, code and information. Humanoid robots, by contrast, affect physical work in manufacturing and logistics – figures from the two areas should not be directly combined. Here, Asia – and China in particular – currently leads.

CES 2026 in Las Vegas was described as a turning point: humanoid robots stopped merely demonstrating walking and acrobatics and began performing real production and warehouse tasks.

According to Andreas Brauchle of consulting firm Horváth, cited by CNBC, China is currently ahead of the United States in the early commercialization of humanoid robots – both countries will eventually build comparably large markets, but China is scaling production much faster at the outset. Behind this is a political decision in Beijing: developing humanoids is meant to address an aging population and a shrinking labor force, while giving China a technological edge.

Market estimates vary considerably by methodology, but business interest in physical AI is already substantial. According to the Capgemini Research Institute report „Physical AI: Taking Human-Robot Collaboration to the Next Level,” two-thirds of organizations regard physical AI as a high priority for the next three to five years, while 79 percent are already exploring, piloting or deploying it. Humanoid robots remain a longer-term proposition: respondents estimated an average of about seven years before they can be scaled broadly.

An industry caveat is essential: despite the media attention, humanoid robots still operate mainly in pilot programs or limited deployments, perform a narrow range of tasks and generally move more slowly than established fixed-arm industrial robots. Reliability, dexterity, cost and workforce training remain major barriers. Capgemini’s survey suggests that scaling humanoid robots broadly will take an average of around seven years, meaning widespread deployment is more likely in the early 2030s than within the next two or three years.

6. Does the law actually ban replacing humans with AI?

The widespread belief that European law bans replacing humans with artificial intelligence is imprecise. The EU AI Act (Regulation 2024/1689) does not prohibit job automation as such. Instead, it introduces phased obligations concerning transparency, governance, risk management and genuine human oversight, particularly for high-risk AI systems used in areas such as employment. The timetable for the application of the high-risk rules was adjusted in 2026 following a political agreement on the EU’s Digital Omnibus package.

Article 26 of the AI Act will be central once the relevant high-risk rules become applicable. A company deploying such a system must assign oversight to people with sufficient competence, training and authority. Human oversight cannot be reduced to rubber-stamping an algorithmic output: the person responsible must be able to understand the system’s limitations, monitor its operation and, where necessary, disregard, override or stop it. Employers must also inform workers’ representatives and affected employees before putting certain high-risk workplace systems into use. Under Article 86, a person affected by a decision based on the output of an Annex III high-risk system that produces legal or similarly significant effects may request a clear and meaningful explanation of the role of the AI system and the main elements of the decision.

Poland’s parliament completed work on the national AI systems law on July 3, 2026, after considering amendments introduced by the Senate. The final parliamentary text establishes the Commission for AI Development and Security (KRiBSI) as the national supervisory authority. Until the national oversight structure becomes operational, the provisions of the EU AI Act that have already entered into application remain directly binding in Poland, while the high-risk employment rules will apply according to the revised EU timetable.

7. What new occupations will emerge

Alongside the transformation of existing jobs, the market is generating new roles:

The practical challenge is therefore not only to create new job titles, but also to retrain people whose existing roles will change. Employers will need clearer rules for safe AI use, training focused on real workplace tasks rather than generic tool demonstrations, and procedures for documenting meaningful human oversight. Without those measures, the gap between occupational exposure and employees’ ability to work effectively with AI may widen.

8. Unemployment: what the hard GUS data shows

The registered unemployment rate, after a record low of 4.9 percent in June and October 2024, rose to 6.1 percent in March 2026, before a cyclical, seasonal spring decline: 6.0 percent in April and 5.9 percent in May 2026 (915,900 registered unemployed, up 17 percent year on year).

The internationally comparable LFS (BAEL) unemployment rate stood at 3.3 percent in Q1 2026 – still one of the lowest in the European Union (Eurostat put it at 3.1 percent in January 2026, tying Poland with Bulgaria for first place in the EU). In the same quarter, however, the number of employed people fell by 109,000 versus the end of 2025, while the economically inactive population grew by 82,000, to 12,538,000. The toughest situation remains among young people: the unemployment rate for the 15–24 age group reached 12.5 percent.

A structural factor is at work in the background: according to the GUS’s main demographic forecast, Poland’s population will shrink from 37.8 million in 2022 to 30.9 million by 2060 – nearly 7 million fewer people over four decades. A shrinking labor supply could, in practice, partly offset job losses caused by automation – one of the arguments in the debate over whether AI may turn out to be an answer to labor shortages rather than solely their cause.

Takeaways for business and investors

Poland’s labor-market exposure to GenAI is currently concentrated in clerical work and information-processing tasks rather than simple physical labor – reversing the traditional automation pattern of previous decades. At the same time, actual corporate adoption in Poland, at 8.4 percent, remains well below the EU average of 20 percent. For employers, the main near-term issue is not a legal ban on automation, but the need to introduce transparent procedures, effective human oversight and documented accountability when AI systems influence employment decisions. For investors tracking physical automation, the key reference point remains the pace of humanoid robot commercialization in China, where production scale – not just technological capability – is increasingly shaping competitive advantage.