This paper presents a methodology based on Artificial Intelligence to enhance top down accountability in the public sector of Bogotá. I use public information on the contract level from the e-procurement platform of Colombia and I train multiple machine learning models to predict which of the public contracts will result in pasive waste. I then quantify the importance of every feature used in the prediction as a new tool to understand the drivers of inefficiency. This approach will be useful for governments, specially at the local level, on the design of cost - efficient audit policies.