AI agents take over offices: Which jobs in Europe are most at risk?

Unlike previous technological revolutions that automated physical labor, the wave of artificial intelligence development directly targets high-paying, highly educated professionals, which is the main finding from an extensive study on the impact of so-called ‘agent-based artificial intelligence’ on the future of work, conducted by the risk management company Coface and the Observatory of Threatened and Emerging Jobs (OEM).

As explained by Aurélien Duthoit, an analyst at Coface, and Axelle Arquié, a researcher and co-founder of OEM, during an online media conference, artificial intelligence is slowly becoming an independent participant in the global labor market, which has its advantages but also disadvantages.

The safest physical jobs

By analyzing over 900 occupations in 12 key global economies, researchers mapped the potential for AI automation. One of the most important conclusions is that jobs requiring physical presence and interaction with the material world are currently the most resistant to the changes brought by automation. This connection to the tangible environment somehow protects employees, as artificial intelligence, despite its advanced logic, still cannot replace human dexterity and resourcefulness.

Similarly, jobs in construction, hospitality, or manufacturing have an exposure rate to automation of less than 10 percent, as today’s artificial intelligence models still struggle to navigate physical spaces.

On the other hand, jobs that involve processing information and data (so-called processing jobs) are on the front lines of impact. Financial analysts, lawyers, administrative staff, and IT engineers are no longer protected by their formal knowledge.

As experts from Coface explained, ‘agent-based’ artificial intelligence represents the second phase of the revolution that will be able to independently plan, manage files, and execute complex tasks without constant human supervision in the next five years. This means that artificial intelligence will soon completely take over some human tasks, but not all.

Croatia safe for now

Moreover, how resistant each country will be to these changes primarily depends on the structure of its economy, so researchers clearly defined European regions according to their risk level.

Countries with high exposure, such as Denmark, Sweden, the United Kingdom, the Netherlands and Switzerland, lead the list as their economies are based on knowledge, finance, and the IT sector, making them the most vulnerable to the automation of office jobs.

Central Europe and France record moderate exposure. Although industrial powers like Germany, Poland, Austria, Slovakia, and Slovenia largely rely on manufacturing where assembly line workers are safe, these industries employ a huge number of engineers and legal staff who are at high risk.

Southern Europe has the lowest exposure rate, e.g., countries like Spain, Greece, Turkey, and even Croatia, as their economies are based on services in tourism, agriculture, and physical construction.

Fiscal challenges

One of the most intriguing and simultaneously alarming conclusions from the press conference concerned public finances. Since AI targets the highest-paid workers, countries could face a sharp decline in income tax revenue alongside a simultaneous rise in unemployment costs.

An additional risk is ‘fiscal leakage’; for example, if local labor is replaced by AI infrastructure owned by American tech giants, the tax base shifts from local wages to international capital, which could drain national treasuries.

Although speakers emphasized that ‘technical exposure’ does not necessarily mean the automatic disappearance of the mentioned jobs, early signals of this trend are already visible in the market. Data shows that employment in the IT sector in France has noticeably stagnated since the introduction of advanced generative models at the end of 2022, marking a sharp departure from the growth trend that lasted since the 1990s.

However, the authors emphasized that the goal of the study is not to spread panic but to understand where the greatest opportunities for automating work tasks are opening up. The transition period will be challenging, Duthoit and Arquié admit, especially for the highly educated workforce, but new roles and industries that we cannot even imagine today will emerge in the long term.