Expert systems, rule-based knowledge systems, have been widely heralded as an important tool in management and accounting. Expert system shells have become available for personal computers, and accountants are investing in systems which are supposed to be capable of intelligent decisions. The limitations of rule-based knowledge systems are discussed and illustrated by means of an expert system built to calculate and diagnose standard cost variances. It is argued that, because of the nature of human intelligence as intuitive and reasoning rather than rule-based, such a system is, at best, a proficient or competent system. The limitations and applications of expert systems, particularly in relation to training, are discussed. (10 references) (Author/GL)