We investigate the scope for supervisory activities in organizations in which information is non-verifiable and opportunism severe. A principal-supervisor-agent hierarchy is considered. Side-contracts between supervisor and agent may be reached both before and after the agent has chosen his hidden action. We find that the supervisor is useful if and only if appointed before the agent has chosen his action. We also show that delegation of payroll authority is suboptimal. Finally, some insights concerning the optimal design of verification activities are provided: when information is non-verifiable, the supervisor should be employed as a monitor rather than as an auditor.JEL Classification: D82, D86, L23