The paper studies a contracting problem in which a Principal enters in two-sided moral hazards with N independent agents. There are no technological or informational linkages between the N agency problems: The Principal's costs are additive across agents; there is no common uncertainty in the agent's performance; and the Prinicipal can freely vary his actions across agents. Despite this, optimal incentive schemes essentially eliminate the Prinicipal's incentive problem when team size N is large enough. This implies that it is suboptimal to require each agent's compensation to depend only on his own outcome. The result also implies the existence of purely informational economies of scale to increasing team size. Thus, the concentration of otherwise unrelated transactions in a single 'firm' creates wealth through a more efficient use of information about the Principal's actions. The paper shows that extremely simple statistical contracts are approximately optimal in large teams. The outcome of such contracts is observationally indistinguishable from standard Principal-Agent contracts. This provides a theoretical justification for using standard Principal-Agent contracts in environments that involve two-sided hazard in a fundamental way.