We conduct a principal-agent experiment and demonstrate that a monopolist agent protects the interest of a (robotized) principal better than when three agents compete to provide a service (project recommendation). The agents have perfect information regarding project types and can provide either positive or negative recommendations (continue/stop). An agent can increase revenues only via a positive project recommendation to the principal, who always trusts the agent. The insight from our experiment can be applied to a number of financial environments (e.g. money-managers and rating agencies) and provides additional evidence on how market structure shapes individual incentives and affects equilibrium outcomes.