Pawel Gmyrek, Senior Researcher at the ILO and co-author of Generative AI and Jobs: A Refined Global Index of Occupational Exposure, discusses how generative AI could shape jobs across countries and sectors.

Based on your research, what are the key trends in how Generative AI is influencing jobs and occupations? 

For the time being, we are still mostly discussing exposure to generative AI. Employment statistics usually react slowly, while exposure – measured through the automation potential of tasks across occupations – gives us a clearer sense of the transformations likely to occur in the mid-term. The picture that emerges is one of job transformation, not a “job apocalypse.” 

In May 2025, the ILO and NASK – Poland’s national research institute under the Ministry of Digital Affairs – released a joint report updating our global index of occupational exposure to GenAI. This project unique is because we could refine the global estimates while also carrying out in-depth studies on Poland, including focus group discussions with workers and managers, and a nationally representative survey on GenAI adoption at the workplace. 

The results confirm many of our earlier findings, but with some important updates. Clerical occupations remain the most exposed, although we observed that automation of many clerical tasks is more difficult in practice than what theory might suggest. Still, the most exposed jobs continue to include data entry clerks, typists, accounting and bookkeeping clerks, and administrative secretaries. At the same time, compared to 2023, exposure has grown in professional and technical roles, such as financial analysts, web and multimedia developers, application programmers, and investment advisers – reflecting GenAI’s expanding ability to take on more specialized and highly digitized tasks. 

Globally, about one in four jobs (24%) show some degree of exposure, and this varies strongly with countries’ income levels: one in three jobs in high-income countries, but only one in ten in low-income economies. Persistent gender differences in exposure are striking. In high-income countries, nearly 10% of women’s employment are now in occupations with the highest automation potential, compared with 3.5% of men’s. These figures are also increasing, up from 7.8% for women and 2.9% for men in 2023. 

Such exposure does not imply the immediate automation of an entire occupation, but rather the potential for a large share of its current tasks to be performed using this technology. Whether this leads to the disappearance of an occupation or workforce replacement is a more complex question – one that will depend on the initial decision to adopt the technology, but also the extent to which individuals in these occupations are given opportunities to learn to work with these technologies and adapt to the evolving nature of their tasks.  

How is GenAI transforming workplaces?    

Despite the intensity of the public debate, the actual use of GenAI in workplaces remains limited, especially when it comes to integration that goes beyond basic tasks. When we surveyed Polish workers in late 2024, only 9.4% said their employer had officially introduced GenAI tools. At the same time, almost half of the workers reported that their companies had no plans to do so. Yet this contrasts with individual behaviour: 16.7% of workers reported using GenAI in the past week, which suggest that a large share of use happens through private accounts or devices. In other words, the technology is entering the workplace “through the side door,” faster than company strategies or regulations can keep up. 

Current adoption dynamics create two distinct groups. Experienced users tend to focus on practical applications and see verifying AI outputs as a normal part of their work. Occasional users, by contrast, are more concerned with broader fears – from loss of creativity to the erosion of core skills. What makes the difference is direct professional experience with the tools. 

The workplace context is critical here. Where GenAI was officially introduced in consultation with employees, most workers reported using it, and two-thirds said they wanted to use it even more. Where no such dialogue exists, uncertainty dominates: 41% of employees in these firms were firmly opposed to adoption. A lack of communication deepens this divide – over two-thirds of workers surveyed in Poland said they had received no guidance on how to use AI tools, and only 2.1% reported that their workplace had set clear boundaries on what uses should be avoided. This is surprising, given the potential risks of misuse for companies themselves. 

What we see, then, is that the main transformation is not only technological, but organizational. How workplaces choose to introduce GenAI – whether in partnership with workers or imposed without guidance – will determine whether the technology enhances job quality and productivity, or undermines them. 

How can we make sure that the transition to AI will lead to the creation quality employment opportunities for workers without increasing inequalities?  

We should remember that digital technologies are never neutral – they reshape labour markets and societies in ways that are as much social and structural as they are technological. Generative AI is no exception. 

There is growing agreement among economists that the real productivity gains from GenAI will not come from cutting jobs, but from the extent to which human expertise can be complemented by new technological capacities. If adopted in a human-centred way, AI can reshape occupations, add value, and improve job quality. But if it is used only to cut costs, increase monitoring, or reduce worker autonomy, the risks of exclusion and inequality grow. 

That is why public policy is crucial. Institutional frameworks, including effective use of social dialogue structures, worker support systems, and clear guidance on prohibited uses are needed to minimize risks and ensure equal access to benefits. Our research shows that some groups are more vulnerable – women, clerical workers, young people, and older workers without digital skills. Managing the transition fairly requires special attention to these groups: systematic reskilling and upskilling, building digital competencies, protecting incomes of those who may lose their employment, and supporting their reintegration into the labour market. 

Ultimately, whether generative AI becomes a driver of development or a source of exclusion depends on our ability to manage the transition with existing political and social mechanisms, with workers having a real say in how these tools are implemented through dialogue with employers.