This paper presents findings from a longitudinal field study examining the problems of managing and transferring local knowledge beyond the specific context and workgroup in which the context and rules for that knowledge are understood. The ways in which a collaborative group managed and communicated local knowledge, translated across organizational boundaries are described. The findings demonstrate a fundamental contradiction between the situated, distributed nature of collaborative knowledge processes and the expectation that software systems will provide codified knowledge. A model is presented that demonstrates a shift from individual knowledge to design knowledge, focusing on the dominant modes of knowledge deployment at different stages of design emergence. The role of specific representational genres, in mobilizing a move from one mode of knowledge-manipulation to another may be significant in boundary-spanning design.