Now that I’m once again working on my PhD, it felt good to attend the TASKS VII Conference, held on September 11 and 12, 2025, in Belval, Luxembourg, and to reconnect with fellow academics. On top of the chance to attend a conference indirectly related to my research (I’m more of an organizational economist interested in personnel economics than an actual labor economist), I relished the opportunity to represent Standard Error at the new home of the prestigious labor economics research network IZA, formerly based in Bonn, at the Luxembourg Institute of Social and Economic Research (LISER). And Luxembourg being only 100 kilometers away from my home in Nancy, I had every reason to drop in.

This seventh edition of the TASKS conference, the first I attended, focused on the economic impacts of AI on work and labor markets. The recent growth of AI in general, and generative AI in particular, has the potential to profoundly reshape skill demands, education and training, labor institutions, occupational mobility, productivity, and labor demand—all topics, and more, that participants discussed during the event.

To start, Terry Gregory (LISER, @terrygregory.bsky.social) presented a paper on the effects of AI on skill demands in Europe. Using data from France, Germany, Luxembourg, and Belgium, Gregory’s paper shows that AI diffusion leads to more diverse skill requirements across occupations and more concentration of skill requirements within occupations. Michael Stops (Institut für Arbeitsmarkt- und Berufsforschung) presented a related paper, this one focusing on establishment-level demand for AI skills. The paper shows that AI has a negative total employment effect, with the effect arising from skilled jobs. For nonskilled jobs, on the other hand, there are no effects, which I find interesting. Eduard Storm (Institut für Höhere Studien, @edthestorm.bsky.social) presented a paper measuring the effects of nongenerative AI (such as machine learning applications) on employment stability and earnings. The paper finds that expertise is key for extracting earnings gains and employment stability from AI. Average workers, however, see no impacts of AI on their outcomes.

On education and training, David Marguerit (LISER) presented a paper studying how occupational AI exposure shapes college major selection in the United States. The paper finds that occupational automation AI decreases the number of graduates while occupational augmentation AI increases the number of graduates, especially at the bachelor’s level.

The session on labor institutions featured presentations by Emilie Rademakers (Utrecht School of Economics and Boston University), Daniela Arlia (European Central Bank), and Oliver Schlenker (University of Konstanz, @schlenkeroli.bsky.social). Rademakers presented a fascinating paper documenting the dynamics of declining collective agreement coverage in the Netherlands from 2006 to 2024. Arlia presented a paper that shows that innovation leads to higher wages, especially for workers at the lower end of the wage distribution. Collective bargaining agreements seem to drive wage gains for manual and routine workers. As a network enthusiast, I enjoyed that Arlia’s paper used an instrumental variable based on a network of patents. Schlenker presented a paper contributing to the literature that shows that different workers are differently exposed to the effects of technological advances, labor market institutions being a potential explanation as to why. The paper shows that, in Germany, work councils shield incumbent workers from the negative consequences of automation.

Thijs Bol (University of Amsterdam, @thijsbol.bsky.social) gave a keynote on whether occupations are “bundles of skills.” The paper he presented uses topic modeling to study skill profiles in job postings. Notably, despite being a sociologist, Bol survived his encounter with the rest of the room.

The second day began with David Autor (MIT, @davidautor.bsky.social) giving an illuminating (and fun!) keynote on expertise. Autor’s key argument is that to understand whether automation puts occupations, jobs, and wages at risk, we need to reframe the question around tasks and expertise. Each job consists of a collection of tasks. Some of these tasks require expertise; others do not. The paper shows differentiated effects of automation on wages and employment: For a given occupation, wages rise and employment is reduced when automation eliminates inexpert tasks. The effects are opposite in occupations where automation eliminates expert tasks.

On occupational mobility, Sara Signorelli (CREST, @sarasgnr.bsky.social) presented a paper that uses a gravity model and US and French data to measure the effects of digitalization on skill distance between occupations and occupational mobility. I particularly enjoyed discovering that gravity models can be diverted from their original use in spatial economics and international economics to other subfields of economics. Matias Cortes (York University, @cortesecon.bsky.social) presented a paper that uses US data to measure whether occupational mobility accounts for wage growth. The paper shows that the occupational ladder plays a role in wage growth, especially early in workers’ careers, and has differentiated effects for women and nonwhite people.

The final session on productivity and labor demand featured, among others, two French economists. I focus on them not out of some misplaced chauvinism but because both papers use natural language processing on a large corpus of documents—in this case, patent filings. The first paper was presented by Fabien Petit (University of Barcelona and University College London, @fabienpetit.bsky.social). Using patent and regional European data, Petit’s paper shows that digital technologies polarize the workforce, increasing low-skilled and high-skilled jobs while decreasing middle-skilled jobs. The second paper, presented by Aurélien Saussay (London School of Economics), was one of the conference papers also studying the environment. The paper shows that green innovation and automation both lead to the creation of high-skilled jobs.

I very much enjoyed the TASKS VII conference. The presentations were diverse and showed the breadth and complexity of how AI has already started to impact the labor market. As the economic role of AI will likely continue to deepen, economists and other social scientists have an important role to play in studying its effects rigorously with the best methods available. The ultimate goal is to develop policies that ensure that the benefits of AI are shared equitably across society. The conference was enjoyable not just because of its scientific and academic content itself but also because of how kind and welcoming the LISER team are. And despite my living close by, my first visit to Belval made clear how interesting a place it is.

Having attended the launch of the LdA-CCA Migration Observatory Report on “Immigrant Integration in Europe” in Paris and the EAYE Meeting in London in May, followed by the 15th Conference of the ASFEE in Nancy and the PSE CEPR Policy Forum in Paris in June, I am looking forward to participating in other conferences and scientific events in Europe, both to enrich my own research and to represent Standard Error.