Building an Expert System That Works

Although the principles behind expert system design are well understood, the success rate for commercial implementations is still quite low. One of the main risk factors is the difficult and time-consuming task of designing and constructing the problem-solving software itself. Non-Al oriented organizations can increase their likelihood of success by utilizing an existing special-purpose expert system shell, and focusing their efforts on the construction of a domain-specific knowledge base for the shell to use. This paper provides some guidelines for successfully applying special-purpose expert system shells in a commercial environment, and describes how one such shell, built for machine diagnosis, has been successfully applied.