A commenter on a previous entry asked for some background on why expert systems have "failed." I'm not exactly sure of his context, but there are two schools of thought here. One is that expert systems haven't really failed, business has found the appropriate place for rules-based systems that capture the heuristics experts use to problem solve or answer questions. The other school is that the traditional expert system could never fully capture the expertise used in difficult situations, as this required understanding of the context in a much deeper way than could be embedded in traditional rules. This suggests expert systems have "failed" because they could never capture this deeper, tacit, knowledge.
A quick Google for expert system shortcomings produced a number of useful results. I particularly appreciate the first link, as it is a researcher's blog on their expert system project: The Use of Expert Systems in Conservation, and the author provides links to her research and some bibliographical references.