Many large, widely distributed organizations struggle with the enormous task of providing the right information to the right people at the right time. Organizations facing this task often develop groups of analysts who specialize in supplying information transport and access capabilities to end-users. However, this approach has several drawbacks. Our aim is to address these problems at their source -not by replacing analysts in the information access problem, but by automating the roles assumed by analysts. Toward this end, this paper describes a strategy that combines four techniques to solve such problems: (1) an architecture for coarse-grained agents (CGAS}, (2) a communication protocol that enables CGAS to interact, (3,) an intermediate query language (IQL), designed around user-level concepts, and (4) a query translation mechanism that transforms IQL requests into database-specific queries. A prototype implementation, known as oMIE, is described.