While successful knowledge management depends on numerous organisational and psychological aspects, the effective documentation, dissemination and utilisation of knowledge recommends the introduction of computerized systems to manage knowledge. The design of such a system requires a notion of knowledge that allows to distinguish it from information as it is handled by traditional information systems. In this paper, a pragmatic notion of knowledge is suggested. On the one hand, it is inspired by some characteristics of knowledge stressed in philosophy. On the other hand, it reflects ideas about knowledge as a corporate asset and as subject of organisational learning. Against this background, a number of requirements which should be fulfilled by a system that manages knowledge are developed. They result in suggestions for the content as well as for the architecture of a Knowledge Management System (KMS). Different from organisational memory systems, the proposed KMS features a high level of formal semantics. Different from expert systems or decision support systems, a KMS does not only help with individual problem solving. In addition to that, it provides a medium to foster discourses between people with different perspectives. Both, content and architecture, are inspired by languages that are part of a method for enterprise modelling. To give an idea of the content a KMS provides, some of its various views are illustrated from a user’s perspective. After that, the object-oriented software architecture is described in more detail by various excerpts from object models on different levels of abstraction. It emphasizes the reuse of existing, state of the art knowledge and allows for individual revisions and enhancements as well. The architecture also includes an interface level layer that helps with the semantic integration of KMS and traditional IS. Both, content and architecture of the suggested KMS are one out of many possible solutions. For this reason, we will also briefly discuss the pivotal challenges that have to be faced by research in knowledge management if it includes the design of specialised software systems.