Soft Systems methodologies provide a means by which systems researchers can understand and work with complex systems. However they rarely attempt to model the cognitive processes that often occur within their models. Artificial Intelligence researchers, on the other hand, do try to develop models of cognitive processes. This paper examines the possibility of combining the two approaches in such a manner as to maintain the strengths of each. To do so it combines Checkland and Holwell's POM (Processes for Organisational Meanings) model with the BDI (Beliefs, Desires and Intentions) architecture from AI. It offers three possibilities for such a mapping, and then briefly explores the possibilities inherent in doing so. Finally, it offers some directions for future research.