A Design Space for Enterprises

The performance of an enterprise depends on the capabilities of the agents that control it and on how these agents are organized. The division and coordination of labor are two of the principal functions of an organization. This paper assembles a five-dimensional space from which organization designers can select coordination policies for a wide class of distributed enterprises. The assembly is based on the extension of model-predictive control to distributed agents, and on the conjecture that distributed enterprises are reducible (meaning that their control problems can be broken into much smaller sub-problems.) This conjecture has not been verified, nor has the space of coordination policies been explored. But samples from the space show promise. They allow agents to work asynchronously--in parallel, each at its own speed--and generate solutions that dominate Nash equilibria.