Delegation is a critical tool for busy managers. Early delegation research suggests that managers are reluctant to delegate beyond a few highly competent employees or those with whom they have a strong relationship. Extending this line of research, we integrate signaling theory with a view of social networks as "prisms," to demonstrate the relevance of employees' network ties in the work unit for delegation. Signaling theory argues that when direct data about employee competence are mixed or ambiguous, decision makers will look for more indirect signals with which to make inferences about quality and reputation. One such signal is suggested by the networks as "prisms" perspective, which argues that network ties can operate as reputational signals in the absence of more direct quality data. Combining these insights and data from a field study and two follow-up laboratory studies, we find that in situations of moderate employee competence, managers will draw positive reputational inferences and be more willing to delegate to employees when such employees are more central in the friendship network of the work unit as well as when they share common memberships in friendship cliques with the managers. These network ties, however, do not matter for delegation decisions when there are direct data to indicate that employees are highly competent or clearly poor performers. Our lab studies elucidate the mediating role of perceived reputation. Implications for theory and practice are discussed.