Knowledge Management (KM) initiatives are driven by the need to preserve and share knowledge, in particular tacit knowledge that experts have built up in the course of doing their jobs. Such initiatives require key experts to be identified and their knowledge elicited. However, knowledge elicitation generally runs into a number of communication and motivational problems. These are well known in domains such as expert systems but it is only more recently that KM practitioners have become aware of them. Standard KM approaches separate the elicitation and, possibly, encoding of knowledge from its subsequent sharing. This paper outlines an approach where elicitation and transfer, and possibly also creation, are carried out in one process. This involves identifying key experts and stakeholders. These two groups then work together to develop a representation of the experts' domain knowledge. The role of the KM specialist thus becomes one of facilitation rather than elicitation. This approach has a number of advantages. It is more likely to engage the interest of experts and so avoid some of the motivational problems that are commonly encountered in knowledge elicitation. It does not rely on knowledge management specialists who do not share the experts' language, to capture and record their expertise. In particular the approach helps overcome the perceptual biases of domain experts. It is well known that perception is often selective and that judgements can be anchored on false premises. Experts are not immune from these biases but they are more likely to be eliminated as a result of the critical dialogue that occurs between experts and stakeholders using our approach. Our approach has been developed in the course of an action research project with a major engineering company. Staff who worked on a help desk had particular expertise which was of interest to other departments, such as design and production. The research data gathered was necessarily qualitative since the focus of concern was on the richness of transfer achieved. Early results suggest that communication or motivation problems encountered by conventional approaches are avoided and that a richer transfer of knowledge results. In particular it helps to identify and capture relevant tacit knowledge. The resulting representation may also form the starting point for a knowledge base which will be available to a wider community.