Case-based reasoning (CBR) has a great deal to offer in supporting creative design, particularly processes that rely heavily on previous design experience, such as framing the problem and evaluating design alternatives. However, most existing CBR systems are not living up to their potential. They tend to adapt and reuse old solutions in routine ways, producing robust but uninspired results. Little research effort has been directed towards the kinds of situation assessment, evaluation, and assimilation processes that facilitate the exploration of ideas and the elaboration and redefinition of problems that are crucial to creative design. Also, their typically rigid control structures do not facilitate the kinds of strategic control and opportunism inherent in creative reasoning. In this paper, we describe the types of behavior we would like case-based design systems to support, based on a study of designers working on a mechanical engineering problem. We show how the standard CBR framework should be extended and we describe an architecture we are developing to experiment with these ideas.