AI is transforming job profiles – not employment

The research team from Denmark, Portugal, Sweden, and Germany has developed a novel measurement approach that breaks down advances in AI technologies into specific subfields and occupational groups. These are then used to analyze employment effects based on anonymized employer-employee data from Denmark, Portugal, and Sweden for the period 2010 to 2023.

The authors capture progress in artificial intelligence across nine subfields—from language processing and text generation to image recognition and decision- and recommendation systems. Technological advances are mapped to occupations whose tasks are closely linked to these capabilities.

The key findings are:

• There has been tremendous progress in AI over the last two decades, though there have been marked differences across different AI technologies. AI used in video and strategy has improved most, while AI related to translation records the slowest progress. 
• Overall progress in AI over the last decade has had little to no impact on overall employment levels. But different AI technologies have different implications. 
• Not all jobs are equally exposed to AI. Data analysts, software developers, and translators are among the most exposed, in highly cognitive occupations with fairly low social interaction. Construction workers and nursing staff, on the other hand, are in occupations requiring physical strength or that have intensive inter-personal contact, which means they are among the least exposed.  
• But exposure does not necessarily imply that jobs are at risk. AI can support jobs—or it can replace them. 
• AI related to translation and text editing primarily affects negatively middle-skill, and to a lesser extent high-skill, administrative and office occupations. Clerical work, assistance roles, or call centers may partly disappear as writing, documentation, and information retrieval are increasingly automated.
• More general purpose AI based on language modelling—such as ChatGPT or Gemini—and reading comprehension—such as AI tutors or chatbots—have generally positive effects on employment across all occupations, but particularly on production workers as well as managers. AI takes care of tedious “paperwork” and allows employees to focus on the social or physical component of their work. Hiring expands, as well as the required skills. Middle skill administrative and office occupations can be replaced by this type of AI. 
• In general, manual labor in production and crafts is barely affected by AI—if anything, there are positive employment effects. However, job profiles may change with increased use of AI in process monitoring or machine operation.

“Higher AI exposure has no measurable effect on overall employment, but it systematically increases the demand for higher qualifications,” says Holger Görg, Research Director at the Kiel Institute and co-author of the study. “Routine tasks disappear, while new roles emerge that require analytical skills, communication, and technical understanding. In particular, we observe that knowledge-intensive firms increase their share of highly skilled employees over time as they adopt artificial intelligence.”

According to the authors, this means the following for policymakers and businesses: training and reskilling must become more targeted. “By breaking AI into its subfields, we can see which technologies affect which occupations—language AI affects administration, visual AI affects technical work, decision AI affects management,” Görg says. “This provides a basis for targeted education and labor-market policies.”