There are few tasks that managers dread as much as employee discipline. Except for cases in which the offense is severe enough to warrant termination, a manager must continue to interact daily with the employee after a punishment has been imposed. When the employee must be fired, the manager often fears that a lawsuit may not be far behind. Thus, when faced with the need to discipline an employee, supervisors might welcome help in figuring out what to do. That is the purpose of the computer expert system called the Supervisor Assistance System, created by the Florida Department of Highway Safety and Motor Vehicles (hereafter referred to as the Florida Department of Motor Vehicles). With the Supervisor Assistance System, a supervisor can first consult with the computer to determine if the employee's offense is actionable under departmental rules. If it is, the supervisor responds to a series of questions about the employee and the details of the offense. The computer then generates one or more recommendations for suggested actions to discipline the employee, along with supportive information from personnel rules. An expert system--one application of artificial intelligence--is a computerized decision-making software that structures expertise in a specific area and emulates human decision-making. The software helps users make a logical decision given its knowledge base drawn from laws, regulations, and human expertise.(1) Public agencies have used expert systems for over a decade, and they are becoming more common for a wide variety of purposes (Leonard-Barton and Sviokla, 1988; Martin, 1991). Social service departments use expert systems to determine eligibility for food stamps or refugee assistance, law enforcement units access solved and unsolved burglary cases through a computerized network to get a profile of possible perpetrators of an unsolved crime, and water testing laboratories apply for state license through an expert system. Although expert systems began as stand-alone computer programs, large public agencies are integrating their expert systems into their information data bases, and thus users may not even recognize their expert system as such (Hedberg, 1995). Most studies of public sector expert systems conclude that their promise is great, as they offer users significant potential benefits: increased productivity, improved service delivery effectiveness, and better problem anticipation (Hadden, 1989; McGowan and Lombardo, 1986; Shangraw, 1987). However, such systems have proven difficult to implement and institutionalize (Hadden, 1989, 204; Hauser and Hebert, 1992; Rubin, 1986). Few empirical studies of expert system implementation have been undertaken to assess what determines the extent to which public-sector decision makers will use an expert system initiated within their organization.(2) This article develops a model of the determinants of the degree to which managers utilize an expert system, and tests the model by examining how managers in the Florida Department of Motor Vehicles react to their expert system. The results offer lessons on how to implement an expert system in an agency so that the organization can achieve the benefits of computer technology. The Florida Department of Highway Safety and Motor Vehicles Supervisor Assistance System The Florida Department of Motor Vehicles developed the Supervisor Assistance System to provide consultations for its Supervisors on handling some 45 different employee offenses that can result in disciplinary actions.(3) (See Appendix 1 for a list of all offenses.) For any type of offense, the Supervisor Assistance System asks questions about the incident under investigation as well as other related information, such as the employee's prior disciplinary record and work history. (As an illustration, Appendix 2 presents the specific questions that the system asks a supervisor for the offense absent without authorized leave. …