Based on evolutionist theories and project management knowledge, a connectionist model based on genetic algorithm is built to simulate innovation process in organizations. Transformation and selection produce micro dynamics to create macro behavior in the agents' population. Results clarify Darwinian Lamarckian adaptation mechanisms as their relation to environment and their interactions. Simulations show the existence of an optimal level of experimentation and selection of projects upstream the innovation process, demonstrate that the efficiency of evolutionist processes is contingent to environment complexity and allow exploring interdependencies and coexistence between two paths of evolution. The model validity is approached through similarity to admitted theory and through a comparative study of the innovation processes of two car makers (Renault and Ford).