This paper introduces a new interactive multi-agent approach to automate the assembly modeling problem in order to generate the parts precedence relations of a product starting from its design data. Assembly parts are considered as self-interested intelligent agents with local information and goals. Centralized approaches based on graphical representation resulted so far in both combinatorial and information complexity, especially when dealing with actual complex industrial products. The proposed decentralized approach overcomes exhaustive combinatorial search, while the ability to store and exchange dynamic information among agents reduces the oversimplification of graphical representation. Furthermore, the incorporation of human agents helps the system to overcome the limitations of artificial intelligence. This approach offers a promising platform for assembly planning automation with minimum human intervention especially for all-new products.