Background: For several years adaptive designs became more and more popular in the pharmaceutical industry and in particular much attention was brought on adaptive seamless designs. Those designs combine the phase II dose finding trial and the phase III confirmatory trial in a single protocol (with a fixed total sample size). The objective of this paper is to propose some utility-based tools to optimize those designs: first in terms of ratio between phase II and phase III sample sizes, and, second, in patient allocation to doses at the beginning of phase II. Methods: Design optimization methods are generally based either on Fisher information matrix (D-optimality) or on the variance of some statistics of interest (C-optimality). Instead, we propose to define utility functions associated to sponsors' decision related to choice of dose for the phase III and we propose design optimization metrics based on the expected value of this utility. Results and Conclusions: After reviewing and discussing several kinds of utility functions, we focused on two of them, that we have assessed through simulations. We concluded that in most of the scenarios simulated, the expected utility was in a sense more sensitive to the timing of the interim analysis (ratio between phase II over total sample size) than on the patients allocation between the doses. This result points out the fact that it might be necessary to enroll a larger number of patients in phase II to allow an accurate identification of the optimal dose.