An experimental workflow system is controlled so that its performance may be brought into line with strategic goals. This system is based on a single intelligent agent that manages the workflows. This single component represents both organisational rules and cultural factors, as well as the state of the decision making environment. This experimental system manages the processing of applications received by a university department from potential research students. This workflow involves decision making by the majority of its users. These decisions are recorded together with a confidence factor in that decision. These confidence factors are combined to give an overall confidence to the end product of a workflow that is either "accept the applicant" or "reject the applicant". The performance of the system is brought into line with strategic goals by adjusting the thresholds for the confidence values. Overall performance measures include the quality of processing and the cost of processing. This particular application was selected because it involves decision making in a range of different types of interaction with the players that interact with the system. This experimental system uses no physical documents. Interactive documents are available on the World Wide Web; 'read only' documents are sent by electronic mail. Principles for designing intelligent workflow systems have emerged from this experiment. As a result of this experiment a second version of this intelligent workflow system is being constructed in a distributed, multi-agent environment. Workflow systems admit considerable flexibility in the formation of decision making structures including structures that cannot be realised in traditional systems.
展开▼
