Which skills protect you from automation, and which make you more vulnerable? These questions were examined by Tomáš Oleš from the Department of Economic Policy at NHF EUBA in a study published in the Journal for Labour Market Research. In his research, he analyzed nearly 350,000 job postings published on the Slovak job portal Profesia.sk in 2022 and explored how required skills relate to wages and exposure to different types of automation technologies. We spoke with the author about how the research was conducted, what lies ahead in the era of artificial intelligence and robotics in the labor market, and what employees themselves can do to maintain their position.
Tomáš, first tell us about your motivation to study skills from job advertisements. What questions does your research answer?
Job ads are a window into what companies actually look for and what they are willing to pay for specific skills. Most existing research has focused on changes between occupations, but differences within the same occupation can be enormous. For example, two accountants in two different firms may have the same job on paper, but if one knows machine learning and the other does not, their market value differs significantly. I was therefore interested in how specific skills affect wages and how they relate to how exposed a job is to automation. I examined three main questions: which skills are most in demand in Slovak firms, which are associated with higher or lower wages, and which can act as a “shield” protecting workers from automation.
Your results show that some skills are associated with higher wages while others come with wage penalties. Could you explain this further?
In simple terms: skills that are rare and hard to replace are better paid, while skills that almost everyone has lose their value. The highest wage premium is associated with skills in machine learning and artificial intelligence—about 4% more compared to similar positions. This reflects their real scarcity in the labor market, as only a very small share of job postings mention them. Similarly, people with managerial abilities, financial skills, project management experience, and social skills such as negotiation or teamwork are paid above average. These findings are consistent with research from the United States and the United Kingdom.
On the other hand, general computer skills—basic work with a computer, spreadsheets, or the internet—which almost everyone possesses today, are linked to a slight wage penalty. This aligns with the observation that some skills have become trivial. For example, IT specialists are not required to list the ability to use a web browser, because it is considered a basic skill in that profession. The labor market is essentially saying: what everyone can do has no special value. Interestingly, skills such as customer service, writing, or physical abilities are not associated with either a wage premium or penalty, suggesting they are seen as a given.
The results suggest that the relationship between the level of automation in firms and demand for skills is not linear, but has a kind of “hump.” What does this mean in practice for employees and firms?
This was the most interesting finding for me. First, I should explain how we measure the extent to which firms adopt technology. Imagine a company that uses several robotic arms in production. It will likely seek workers with skills related to operating or maintaining such equipment. Its job postings indirectly reveal the technologies it uses. By ranking firms based on how closely their job ads align with specific technologies, we can observe which skills they still demand—that is, which skills machines complement rather than replace.
And here comes the “hump.” Firms in the early or middle stages of adopting automation technologies demand more skills from employees, because automation complements rather than replaces them. Only firms at the very frontier of adoption begin to reduce their demand for skills, as machines take over more tasks. This pattern appears across all three technologies studied: artificial intelligence, software, and robotics. For employees, this means automation is not an immediate threat and initially increases the value of human work. However, firms that fully automate may gradually change the structure of required skills quite significantly, which requires active preparation and reskilling.
You found that language skills are paradoxically associated with higher exposure to automation—how do you explain this? Does your research suggest that learning languages is no longer necessary?
Definitely not, and this result needs to be interpreted very carefully. This is one of those cases where a statistical association can be misleading if interpreted causally. When a job ad requires German, Ukrainian, or Hungarian, it does not always mean the company is looking for someone with advanced language proficiency. Often, it signals that the position is targeted at migrants who are willing to work in routine and physically demanding jobs. These occupations are also more exposed to automation.
This is an important finding in itself, as it suggests that some groups of workers face a double pressure: they have limited access to better jobs and at the same time work in segments where automation is progressing fastest. Foreign languages therefore remain valuable; in this context, however, we are capturing something slightly different than language proficiency itself.
What practical recommendations follow from your research—what should employees, firms, or policymakers do to respond to the rise of automation technologies?
For employees, the message is clear: invest in skills that machines still struggle to replicate. These include the ability to work effectively with people, lead teams, solve unstructured problems, as well as specialized digital skills such as working with AI or advanced data analytics. Interestingly, manual skills requiring physical coordination in unstructured environments—such as repair, maintenance, or care—also remain relatively protected from automation.
By contrast, routine cognitive tasks, customer service, and simple administrative work are most at risk, and workers in these areas should actively consider expanding their skill sets.
Firms should view automation not as a simple replacement for people, but as a tool for reallocating work toward more valuable and engaging tasks. The research suggests that companies that do this well actually increase their demand for skills, which can motivate employees.
For policymakers, the key message is that automation does not arrive evenly and does not affect everyone equally. Workers in routine jobs, including migrants concentrated in automatable segments of the labor market, are most at risk. Targeted reskilling programs, support for lifelong learning, and attention to vulnerable groups are therefore essential.
I would summarize it in one sentence: technology is changing the rules of the game, but human adaptability remains our strongest asset.
This research was supported by the project Scholarships for Excellent Researchers (R2–R4) — Stagnant Wages in the Face of Rising Labour Productivity: The Role of Automation Technologies — and by the APVV project Towards a Shared Prosperity: Technology, Inequality and Labor Market.
