Automation, AI, and the Future of Work: A Unified Framework With Evidence From Agriculture Literature

This paper surveys the evolving literature on the labor market impacts of digital technologies, contrasting the distinct mechanisms of automation and Artificial Intelligence (AI). While the dominant “task approach” successfully explains automation‐induced labor polarization through the substitution of routine tasks, the recent emergence of Large Language Models (LLMs) challenges this framework by demonstrating “general‐purpose” capabilities in non‐routine cognitive domains. Using agriculture as a comparative case study, the paper highlights that automation primarily displaces medium‐skilled labor, whereas AI increasingly substitutes for junior‐level analytical work while complementing senior decision‐making. To bridge these disparate findings, the paper proposes a unified analytical framework grounded in the bidirectional transformation “from atoms to bits and from bits to atoms.” It argues that digitalization “front‐loads” non‐routine cognitive tasks into the digital realm—making them amenable to AI optimization—while automation executes the resulting codified instructions in the physical realm. The paper concludes by discussing the imminent convergence of these forces into “embodied intelligence,” which threatens to compress the residual human role to final oversight and accountability tasks, and outlines the need for differentiated policy responses regarding education and labor regulation.