This work considers the problem of designing a control law to yield a desired closed-loop response subject to input constraints and plant-model mismatch. To this end, a two-tiered model predictive controller is designed that first computes the best response of a certain desired form, and then computes the control action to achieve the prescribed closed Poop response. The key idea is first illustrated in the absence of plant model mismatch for a prescribed first- and second-order response of the controlled variables, and shown using a simulation example. Subsequently, a formulation is presented to handle plant model mismatch. Finally, the proposed formulation is applied to a chemical process example.