This article is concerned with job design, that is, the grouping of tasks into jobs, in teams of risk-neutral homogeneous agents. It shows that when only some tasks are observable by the agents or monitorable by the principal, job design can restrict the set of sequential equilibria to the Pareto optimal one, by making incomplete information from the agents or the principal effective in overcoming coordination failures and conflict among coworkers. Job design is shown to be a constituent part of the overall incentive system, just as efficient compensation rules are. Some criteria for optimal task assignment are derived.