An Intelligent Agent for Web-based Process Redesign

Process redesign is an expensive, time consuming and labor-intensive activity. Analysis of the redesign process indicates the "first generation" computer-based tools are inadequate for redesign today. Knowledge-based systems and intelligent agents have the ability to address the key, intellectual activities required for effective process redesign. An intelligent redesign agent called KOPeR is developed and employed, in an "industrial strength" reengineering engagement, to redesign processes from the procurement domain. This paper describes the KOPeR design and implementation and highlights its use as a redesign agent in the field. The field results reveal insights into the use, utility and potential of this agent technology, and the paper closes with a number of promising future directions for related research. 1. Redesigning process redesign With nearly all major corporations—and many other enterprises in the military, government, universities and elsewhere—actively engaged in business process reengineering (BPR; see [2]), the reengineering phenomenon continues to be very important in business and management. However, reengineering practice to date reflects a questionable record of success, and as currently practiced, process redesign is an expensive, time consuming and labor-intensive activity. This makes the process of process redesign an attractive target for process redesign. Analysis of the redesign process (i.e., meta-redesign) indicates the "first generation" computerbased tools are inadequate for redesign today [6]. Although a plethora of tools exist for the modeling and simulation of enterprise processes [5], such tools fail to support the deep reengineering knowledge and specialized expertise required for effective redesign. Rather, the key, intellectual redesign activities must be performed manually at present, or provided through expensive BPR consulting services. The reengineering domain offers good opportunity for applied AI such as knowledge-based systems (KBS) and 0-7695-0001-3/99 $1 intelligent agents [12, 29]. For example, knowledgebased systems have the ability to address intellectual redesign activities directly, and an intelligent redesign agent can augment first-generation tools to streamline the redesign process itself. Two primary approaches to automated redesign problem solving have been proposed: 1) case-based reasoning [24, 28, 29], and 2) measurement-driven inference [19]. Case-based reasoning (CBR) mirrors the kind of human problem solving accomplished by most reengineering consultants, and a relatively large number of redesign cases now exist for knowledge representation and indexing. However, organizational processes represent very complex systems, each of which has a great many idiosyncrasies and details that can be critical to redesign efficacy. But rarely are such factors expressly recorded in terms that can be incorporated into a casebase; hence reengineering CBR systems have considerable difficulty adapting previous redesign cases to the needs of novel and dissimilar process instances and pathologies. This represents a textbook problem with CBR [23]. A more serious difficulty with this CBR approach, perhaps, involves automation of an analytical method that fails more than half the time in practice [4, 11]. Alternatively, a data-driven method can address such idiosyncrasies and details directly [27], and measurement-driven inference represents one of the most powerful data-driven methods. For example, some classic diagnostic systems such as MYCIN [25] and SOPHIE [3] have long been implemented around measurement-driven inference, and they reason with performance matching or exceeding that of experts and professionals in their respective medical and electronics domains. Specifically, MYCIN uses counts of white blood cells to drive inference oriented toward the diagnosis of blood disease in people, and SOPHIE uses electrical measurements such as voltage to guide inference oriented toward diagnosing faults in electronic circuits. Both systems perform heuristic classification [13] and suggest measurement-driven redesign problem solving may be feasible as well. In this paper, we describe an analogous KBS developed to provide intelligent, Web-based redesign 0.00 (c) 1999 IEEE 1 Proceedings of the 32nd Hawaii International Conference on System Sciences 1999 Proceedings of the 32nd Hawaii International Conference on System Sciences 1999 automation. The KBS performs its measurement-driven problem solving to diagnose process pathologies and recommend appropriate redesign transformations for improving performance. Further, as a software agent (e.g., see [8, 10, 22]), this second-generation redesign tool called KOPeR is designed to distribute reengineering expertise across the Web to support the redesign of enterprise processes without the geographical or simultaneity constraints that limit human redesign agents or consultants. In the sections that follow, we first outline the key background information pertaining to this intelligent system and then highlight its "industrial strength" application in the field to redesign a number of operational processes from the procurement domain. This field application shows KOPeR-supported process redesign is not only feasible, but effective as well, and it illuminates a number of promising applications of the technology.