Abstract
Generative artificial intelligence (Gen AI) is transforming work, yet how it reshapes the skill portfolio of occupations remains poorly understood. Prior research has focused on tasks and productivity, overlooking changes in the underlying structure of skills. Grounding on the task-based view, we develop a skill portfolio theory that conceptualizes occupations as dynamic configurations of hard and soft skills characterized by both size and diversity. We test this framework using a near-universe panel of 724 million U.S. job postings from 2020 to 2025 and exploit the public release of ChatGPT in November 2022 as an exogenous shock. Using a difference-in-differences design with occupation and time fixed effects and AI exposure measures, we estimate the causal impact of Gen AI on occupational skill portfolios. We find a pronounced asymmetry whereas hard-skill portfolio size contracts, while soft-skill portfolio diversity increases following Gen AI diffusion. These changes unfold gradually and persist over time. Importantly, occupations experiencing this reconfiguration earn a 23.6% wage premium, indicating that labor markets reward effective skill re-composition rather than skill accumulation alone. Our findings advance theory on Gen AI and work by shifting attention from tasks to skill portfolios and demonstrating that different dimensions of skills evolve independently under technological change.
Recommended Citation
Wang, Paul and Jiao, Wenyu, "From Tasks to Skills: How Generative AI Transforms Occupational Skill Portfolios" (2026). ASAC 2026. 15.
https://aisel.aisnet.org/asac2026/15