We will discuss the design of a novel adaptive, inferentially seamless phase 2/3 study for an experimental drug to treat diabetes. The design employs a Bayesian Analytical approach to allocate patients to a set of doses of the experimental drug and to determine if there are 1-2 doses that could be continued to be studied for the purposes of "confirming" safety and efficacy of those doses. The preference for a dose is determined by a Clinical Utility Index, which balances the select efficacy and safety measures. The algorithm is completely pre-specified and the operating characteristics were assessed via simulation. This design was developed through iterative simulations that involved key decision makers who would normally have input as to what doses were to be selected for confirmatory trials. It involved much more apriori planning than would be required for a typical fixed design.