Artificial intelligence research has given rise to a kind of computer program often (and erroneously) called an 'expert' system. Expert systems do much less than their name would lead one to believe. Nevertheless, such programs can be very useful, provided one knows what to expect from them, and when to use them. This contribution to the round table discussion will analyse the contruction of expert systems, will discuss the possible uses of expert systems created with current artificial intelligence technology, and will discuss future research directions.