While it is a truism to underline that modern industrial development in high technology environment strongly relies on innovation, little is known about the innovation success factors. We argue that studying innovation process – defined as a knowledge creation process – from a complexity perspective enables us to better understand this emergent process and its characteristics. This research deals with the knowledge creation trajectory within innovation processes. The framework of the paper is built on a review of a knowledge based view of innovation and the theory of complex adaptive system (CAS). Departing from real-life case studies of innovation processes, the research aims to explore and to understand the knowledge creation trajectory with a focus on the dynamics of this process. The empirical studies refer to IT-based innovations. We use a case study methodology based on a combination of data collection methods: interviews with key actors, non-participant observations and analysis of internal documents explaining the project stakes. The results show that the four processes of innovation evolve in seven, separate but interdependent, knowledge creation stages. The interweaving of these stages allows us to analyze the trajectory dynamics. The research also provides some evidence about the internal and external parameters impacting the trajectory. In addition, we observe a complex dynamic of evolution – specific to each process – characterized mainly by adaptation loops and feedback processes. The results of case studies allow developing an agent-based model of knowledge creation within innovation process and offering a new view of innovation, based on an interactionnist social approach.