A recent study of the Swedish labor market, analyzing data from 1880 to 2019, sheds light on the origins and evolution of jobs over more than a century. Researchers used artificial intelligence to evaluate every occupation with two distinct scores. The first, termed the “Smithian” score, measures how much a job depends on specialization, with occupations like wholesale traders and radiologists rating highly. The second, the “Schumpeterian” score, reflects the extent to which a job is linked to new technology, exemplified by roles such as computer programmers and electricians.

The study finds that while both types of jobs have grown as a share of total employment over time, the majority of employment growth since 1990 has come from highly specialized, or Smithian, jobs. These accounted for 60 percent of employment between 1990 and 2019, compared to roughly 33 percent for technology-driven, Schumpeterian jobs. Furthermore, Smithian roles tend to be more durable, whereas occupations closely tied to specific technologies appear more vulnerable to obsolescence. For example, punch-card operators were once at the forefront of technological change but eventually became redundant.

The researchers note that although there is widespread concern about artificial intelligence disrupting jobs, it may be more meaningful to consider how AI affects the organization of specialized tasks rather than solely focusing on the displacement of entire occupations. Approximately 70 percent of employment during the period studied took place in jobs that existed at the end of the 19th century, highlighting the persistence of certain occupational structures despite technological advancement.

Luis Garicano, co-author of the book *Messy Jobs*, offered insight into how AI could influence the division of labor. He explained that some task bundles within jobs are only loosely connected, making them susceptible to automation and reconfiguration by AI. However, he suggested that an opposite trend might emerge if AI enables individuals and firms to internalize specialized skills more easily, reducing reliance on external experts. A scenario he illustrated involved an economics columnist choosing to build and maintain her own website instead of contracting with a freelancer.

Early indications of this shift are evident. Data provided by OpenAI revealed that about 17 percent of work-related AI interactions involve tasks traditionally belonging to different occupations. Garicano noted that if AI leads to a reversal in the division of labor, it would mark an unusual departure from the patterns observed in previous technological revolutions. Overall, the study emphasizes that while new technologies continue to reshape work, the nature and outcomes of such transformations are diverse and do not necessarily follow a uniform trajectory.