Recently, researchers have begun investigating an emerging, technology-enabled innovation that involves the use of intelligent software agents in enterprise supply chains. Software agents combine and integrate capabilities of several information technology classes in a novel manner that enables supply chain management and decision making in modes not supported previously by IT and not reported previously in the information systems literature. Indeed, federations and swarms of software agents today are moving the boundaries of computer-aided decision making more generally. Such moving boundaries highlight promising new opportunities for competitive advantage in business, in addition to novel theoretical insights. But they also call for shifting research thrusts in information systems. The stream of research associated with this article is taking some first steps to address such issues by examining experimentally the capabilities, limitations, and boundaries of agent technology for computer-based decision support and automation in the procurement domain. Procurement represents an area of particular potential for agent-based process innovation, as well as reflecting some of the greatest technological advances in terms of agents emerging from the laboratory. Procurement is imbued with considerable ambiguity in its task environment, ambiguity that presents a fundamental limitation to IT-based automation of decision making and knowledge work. By investigating the comparative performance of human and software agents across varying levels of ambiguity in the procurement domain, the experimentation described in this article helps to elucidate some new boundaries of computer-based decision making quite broadly. We seek in particular to learn from this domain and to help inform computer-based decision making, agent technological design, and IS research more generally.
Over the last three decades, a significant stream of research in organizational behavior has established the importance of goals in regulating human behavior. The precise degree of association between goals and action, however, remains an empirical question since people may, for example, make errors and/or lack the ability to attain their goals. This may be particularly true in dynamically complex task environments, such as the management of software development. To date, goal setting research in the software engineering field has emphasized the development of tools to identify, structure, and measure software development goals. In contrast, there has been little microempirical analysis of how goals affect managerial decision behavior. The current study attempts to address this research problem. It investigated the impact of different project goals on software project planning and resource allocation decisions and, in turn, on project performance. The research question was explored through a role-playing project simulation game in which subjects played the role of software project managers. Two multigoal structures were tested, one for cost/schedule and the other quality/schedule. The cost/schedule group opted for smaller cost adjustments and was more willing to extend the project completion time. The quality/schedule group, on the other hand, acquired a larger staff level in the later stages of the project and allocated a higher percentage of the larger staff level to quality assurance. A cost/schedule goal led to lower cost, while a quality/schedule goal led to higher quality. These findings suggest that given specific software project goals, managers do make planning and resource allocation choices in such a way that will meet those goals. The implications of the results for project management practice and research are discussed.