The primary objective of effective productivity improvement is to reduce unnecessary and wasteful effort,notsimply to speed things up. Constructing an organizational model along these lines might assist a manager in assessing the organization's current state and in moving to a more desirable position. Such models may be construed as managers’ mental models. A mental model is not a static set of knowledge, but rather, a dynamic memory that integrates the new information resulting from environmental scanning. Thus, a mental model consists of some internal representation that reflects the essential features and relationships in a corresponding real‐world system, for example, the organization. This article proposes an architecture and a representation scheme for implementing computational models that correspond with the mental models in managers’ minds. A four‐level architecture is outlined that consists of a linguistic layer, an epistemological layer, an object‐management layer, and an implementation layer. An iterative, triarchic research method was used that simultaneously developed a theoretical framework, synthesized evidence from an American Express case study, and specified a computational representation. A prototype management support system developed as part of this research, called SPRINT (Strategic Plan and Resource INTegration), is implemented as a frame‐based semantic network using a hypertext interface and is programmed in Smalltalk/V286.