Abstract
The objective of this research was to investigate the planning of information assurance in agent-based workflow systems. Two objectives were set: (1) defining information assurance in distributed information systems, such as ERP, and (2) planning agents for effective execution of assurance tasks in workflow systems. To achieve the first objective, a TQM-based definition of information assurance was developed after literature review, MICSS lab experiments, and an industry survey. A list of requirements for information assurance was created. The result of the lab experiments show the difference of impact of information assurance failtures in an ERP system on the profit and due-date-performance of a company. The fact that certain information failures are more critical than others motivated the implementation of variable assurance in workflow systems. The responses to the industry survey proved that comanies are facing Information Significance problems in their ERP systems, and that decision-makers are willing and able to wait for better assured information. To achieve the second objective of planning effective execution of assurance tasks in agent-based workflow systems, two research tasks were performed: development of a model for variable assurance implementation, and investigation if different variale assurance protocols and agent modes. For implementing variable assurance in agent-based workflow system, the concept of request analysis and context analysis for risk assessment were introduced. These features were added to a recently developed multi-agent framework for process monitoring called Agent-based Integration Model of Information Systems (AIMIS). For investigating assurance protocols and models, experiments were performed with AutoMod to simulate assurance and production tasks execution by agents. The main findings are: (1) Compared to systematic total information assurance, significant requestprocessing time can be saved by executing assurance tasks on an assurance-needs basis. The needs-based variable assurance protocol reduces the mean processing time of requests without decreasing the proportion of trusted requests; (2) Polyvalent agents are recommended solution to assure production requests when assurance tasks are serialized.