Knowledge-based engineering (KBE) approaches are designed to reduce the time and cost of product development by capturing, retaining and re-using design knowledge. They currently focus on repetitive design tasks where knowledge is considered as a static resource. However, knowledge is intrinsically linked to the organizations and people who use it. Thus, to be efficient, these knowledge-based systems (KBS) have to be able to take into account all the mechanisms of knowledge creation, sharing and evaluation made by the users. Using the agent paradigm, new knowledge-based systems can be designed in order to address this research issue. Indeed, the agents have social abilities and are able to achieve very complex tasks. These two features are necessary for making a knowledge-based system efficient. However, there still exists today a lack of approaches and methodologies to help design such applications. This paper presents DOCK, a methodology to design an intelligent knowledge-based system that aims to support the knowledge management process. In order to take into account all the mechanisms of knowledge generation, sharing and re-use, DOCK is based on the hypothesis that efficient modelling of human organizations, by highlighting their roles, collaborations, skills, goals and knowledge, will help the KBS designer to specify an adapted knowledge-based system. Finally, DOCK is implemented to design the SMA SNOTRA that is dedicated to supporting a decision-making process for design projects.