Recent advances in AI represent improvements in prediction. However, AI should take place in activities that are currently performed by humans. We examine what type of human labor will be a substitute versus a complement to emerging technology. If the AI may provide cheap predictions, the human may mitigate the risks that can be implied by the prediction of the AI. We find that there is an optimal way to manage the AI and the human together. However, I suppose that there is a principal-agent scheme. A principal asks an agent for recommendations in exchange for a wage. The agent can use his own human judgment or an AI in order to provide accurate recommendations. However, the use of the AI implies a rent for the agent. The presence of a rent implies distortions, due to conflicting interests between the principal and the agent.

The objective is to study the distortion toward the first best.