The Uneven Future of Work: Task-Based Evidence on AI Adoption and Strategic Workforce Reconfiguration in Global Industries

Abstract

This thesis investigates the labour market impact of Artificial Intelligence (AI) through the lens of task-based technological change. As AI technologies become increasingly embedded across industries, their effect on employment structures, wage dynamics, and occupational inequality is not uniform. Instead, these outcomes depend on the specific tasks that comprise different occupations— particularly the distinction between routine and non-routine work. This raises important questions about how AI adoption reshapes work in practice and how such changes can be systematically measured. To address these questions, the thesis builds on task-based economic theory and formulates five hypotheses concerning employment effects, wage trajectories, inequality, polarization, and sectoral variation. These are empirically tested using a longitudinal panel dataset covering U.S. occupations and industries from 2014 to 2023. The analysis integrates occupational task measures from O*NET, wage and employment data from the Occupational Employment and Wage Statistics (OEWS), and proxies for AI adoption derived from patent intensity and industry-level surveys. Methodologically, the study employs fixed-effects regression and difference-in-differences models to capture both average treatment effects and temporal dynamics. The findings suggest that AI has heterogeneous and temporally dynamic effects. Routine-intensive occupations experienced employment declines following early AI diffusion but also displayed relatively greater wage growth. A complex pattern of wage inequality emerged: initial disparities widened but later moderated as tasks were reallocated and organizational structures adapted. Sectoral analysis further revealed pronounced polarization and task restructuring in high-AI industries such as finance, while low-AI sectors like healthcare remained comparatively stable. The thesis concludes by discussing the theoretical and policy implications of these findings, emphasizing the need for task-oriented strategies to facilitate adaptive responses to AI-driven transformation.