The Negotiated Retrieval and the FPL-based cooperative CBR represent two attempts at exploiting previous experience, distributed within a corporation, for solving new problems. Each of the two approaches brings with it a set of conditions and problem features for which it is most appropriate.
Negotiated Retrieval is based on a model where the response to a query is derived from composing partial responses from distributed case bases. Tasks like assembling cross-functional teams or assembling a set of documents for various aspects of a project from the documentation set available from previously executed projects (to let a project leader derive leads from them for the present project) or accessing a set of relevant components from manufacturer-specified online component catalogues for various design assembly stages represent a few examples where NRA provides a useful tool for the users. On the other hand, DistCBR and ColCBR represent attempts to benefit from the collective experience of peers in a corporation. Each of the peer agents can solve the task on hand by itself. However, their experiences are different and diverse and thus each of them can potentially bring some unique experience to the task. Exploiting this diversity and richness is what these two modes of cooperation attempt to do. Tasks like a project manager's agent exploiting the experience of its peers for project cost estimates or an agent exploiting the experience of a group of expert agents for a new marketing initiative are examples where these modes of cooperation can come in handy.
DistCBR and ColCBR rely on knowledge modeling to flesh out a domain to capture the recursive structure of task-method-subtasks relationships and manage the conflicts and preferences among subtasks or submethods. This represents a knowledge intensive approach to cooperative CBR. On the other hand, Negotiated Retrieval is a search-intensive approach to cooperative CBR where preferences and harmful interactions are managed by augmenting the retrieval of individual subcases with search during integration.
Lastly, DistCBR and ColCBR have certain specific representational requirements like agent control being organized to be able to work with descriptions in Noos language. Thus resources in a corporate memory setting have to be augmented with such capabilities to be able to exploit these two modes of cooperation. However, this is not a serious limitation because Noos is a very general and powerful representation language. Negotiated Retrieval is designed to work with heterogeneous agents that can be equipped with ``wrappers'' to achieve cooperative communication without having to change the representations of their internal problem solving control.