Artificial intelligence (AI) can make important contributions to scientific research by performing functional tasks such as reviewing prior literature, classifying digital data, or developing new drug compounds. There is less evidence, however, on the potential of AI as a mechanism to manage human workers who perform such research tasks. We explore algorithmic management in the context of projects that push the limits of human managers’ capabilities: Crowd science projects that involve large numbers of individuals that vary greatly with respect to factors such as their skills, interests, as well as geographic location. Our analysis of archival data as well as interviews with project organizers and crowd members demonstrates the use of AI for five management functions: task division and task allocation, direction, coordination, motivation, and supporting learning. We also provide a deeper understanding of the underlying capabilities that enable AI to perform these functions, of challenges when using algorithmic management in science, and of contingency factors that may shape the effectiveness of AI. We conclude with a discussion of opportunities for future research on the organization of science and on algorithmic management.