One topical challenge in knowledge engineering is to build systems that can ease the sharing and re-using of experience between a large community of users. This requires a certain plasticity of systems and associated interfaces. Yet, the lack of flexibility is a limitation of current case-based reasoning (CBR) systems. They are often developed in order to solve one particular problem and are not designed to be changed by users depending on the use they want to make of them. In this paper we propose to use interaction traces as a knowledge source for CBR systems and we show how it allows us to drive back the current limits of CBR.