Organizations use combinations of explicit performance incentives and rules prescribing behavior to motivate workers. However, monitoring of adherence to rules creates a second set of agents subject to their own agency problems. In this paper we provide a formal model of the effectiveness of performance incentives and worker autonomy in improving organizational performance. The model highlights the importance of the relative alignment of frontline workers and their monitors with organizational goals. We implemented a large-scale randomized control trial with the government of Punjab, Pakistan to provide the first experimental evidence on the effects of incentives and autonomy in a bureaucracy. We find that increasing procurement officers’ autonomy vis-à-vis their auditors improves prices paid by 7% throughout the fiscal year without reducing quality. Performance pay incentives reduce prices early in the year, but increase them at the end of the year when auditors have greater hold-up power, making the average effect of incentives zero. The results suggest auditors are less concerned with saving public money than procurement officers are. The results have important implications for the design of monitoring and anti-corruption policies.