Résumé : How does Artificial Intelligence (AI) affect the organization of work and the structure of wages? We study this question in a model where heterogeneous agents in terms of knowledge--humans and machines--endogenously sort into hierarchical teams: Less knowledgeable agents become "workers" (i.e., execute routine tasks), while more knowledgeable agents become "managers" (i.e., specialize in problem solving). When AI's knowledge is equivalent to that of a pre-AI worker, AI displaces humans from routine work into managerial work compared to the pre-AI outcome. In contrast, when AI's knowledge is that of a pre-AI manager, it shifts humans from managerial work to routine work. AI increases total human labor income, but it necessarily creates winners and losers: When AI's knowledge is low, only the most knowledgeable humans experience income gains. In contrast, when AI's knowledge is high, both extremes of the knowledge distribution benefit. In any case, the introduction of AI harms the middle class.