The purpose of this paper is to argue for rule extraction as an integral part of data mining. The paper contains two case studies where rule extraction is used for typical data mining tasks. More specifically, rule extraction is used both to explain existing opaque models and to produce concept description via an opaque model. The main result is that the rule extraction approach generally yields comprehensible models with higher accuracy than transparent models created directly from the data set by CART and See5. The implication is that rule extraction can enhance the capability of data mining for decision support.