I study a behavioral mechanism design problem involving a principal and a single agent. The principal seeks to implement a function mapping agent types to outcomes and must commit to a mechanism mapping actions to outcomes. In this setting, only a small class of trivial functions are implementable if the agent is fully rational. I introduce a model of bounded rationality where the agent has a limited ability to combine different pieces of information. Specifically, the agent transitions between belief states by combining his current beliefs with up to K pieces of information at a time. By expressing the mechanism as a complex contract---a collection of clauses, each providing some information about the mechanism---the principal manipulates the agent into believing that the mechanism is ambiguous. I assume the agent is averse to this perceived ambiguity and characterize the set of implementable functions. The main result shows that, without loss of generality, the principal selects a robust, maximally complex contract achieving implementation for all abilities below a bound K*. The model of bounded rationality introduced here extends naturally to other environments and provides an intuitive notion of what it means for a problem to be difficult to solve.