Current trends in automation show that technology has been very successful at the unit control level, and that present methods are very efficient at providing effective local control. Commercial pressure is now on ensuring total optimal integration of all system components. This must lie in improving overall performance - not in bulding a better, more intellectual local controller, but in total, plant-wide control. The key to successful system-wide control must, therefore, lie in improved co-operation and co-ordination between these localised control units. Conventional mathematical control techniques have proven to be impractical to implement on complex control systems due to their inherent distributed nature, and where they have been implemented have proven to be inflexible and unreliable. There is therefore, a need for a structure that supports such control and hence an architecture upon which this structure may be built. In searching for the basis of an architecture, one invariably looks for existing working examples of effective solutions. If a well organised human workforce is considered, certain characteristics emerge, the prime ones being their flexibility and adaptability, in the face of new experiences and situations. From a study of these characteristics, some fundamental structures emerge: that in a human workforce, each worker has localised knowledge about the system as a whole, and that this can be used for effective, non-centralised, truly distributed control. The solution to the problem of increasing overall system-wide performance is sought by modelling the characteristics of a human system and to mimic these characteristics in an automated system. A solution is offered in this thesis in the form of DENIS - a distributed artificial intelligent architecture - which distributes the intelligence of the system by giving each agent accessto a system-wide view of the plant, and also a temporal model of the system. Indeed, this thesis suggests that humans owe much of their flexibility and reliability not only to their ability to reason logically, but also by their ability to reason in terms of time.