Corporate AI play and short term skill-biased AI change

We develop a task-based model that illustrates how skill-bias emerges in firms, either because AI competes with lower skills and/or augments more complex jobs, while the extent of bias depends both on the corporate focus on efficiency/innovation and on AI performance scope across the whole range of firm tasks. Based on those predictions, we build an empirical model of skill change across 12 categories to assess whether, and how large, short term changes in skill labor demand correlates with firms use of AI technologies in their business, in the context of AI before genAI development. While AI is skill-biased in favor of more advanced skills, the effect of AI on skill demand is usually positive for most skills, reflecting that early AI adopters tend to gain business against non adopting rivals. A key mediator however is how firms leverage AI; when AI is used mostly for automation, basic skills prospects become negative. • Studies regarding kill biased change associated with a detailed portfolio of AI technologies are lacking. • We model a firm choice of skill allocation which depends on the relative performance of labor and AI; and how firms adopt a substitution or augmentation strategy to AI.Empirically, we assess model prediction of skill changes for a list of 13 skill categories. • We first find evidence that AI increases probability of employment for most skills, The distribution of change however favors higher end skills, confirming the SBTC theory applied to any AI technologies. • The extent of skill bias is mediated by how firms plan to use AI with a bias towards automation pressing a decline in lower skills.