Task-Interdependencies between Generative AI and Workers

Our paper formalizes a production function to give microeconomic foundations for the adoption of Generative AI (GAI) within workplaces. The production function accounts for task-interdependencies, the worker-GAI interaction and indistinguishability between human-created and AI-generated outputs. We show that workers and GAI represent two distinct but interdependent sides of the production, that jointly generate a network externality in learning that drives productivity. We find that in open learning organizations favoring the worker-GAI interaction, GAI should be matched to workers based on their ability to detect errors. We analyze configurations where the worker-GAI interaction is limited, referred as closed learning organizations, including firms banning the use of GAI, technological superclusters and emergence of small entrepreneurs innovating with GAI.