This paper offers a rationale for limiting the delegation of (real) authority, which neither relies on insurance arguments nor depends on ownership structure. We analyse a repeated hidden action model in which the actions of a risk neutral agent determine his future outside option. Consequently, the agent can improve his future bargaining position, which gives the principal an incentive to retain sufficient control over the agent’s actions. Using respective one-period contracts, the principal can implement the efficient outcome while “selling the shop” to the agent is sub-optimal. This provides an argument for integration if the boundary of the firm is defined by control rights rather than the entitlement to revenues.