The development of expert systems (ES) for manufacturing problems indicates a need to interact with potentially large amounts of data, much of which resides elsewhere in the ES user's organization. A large amount of information required for planning, design, and control operations can be made available through an existing database management system (DBMS). The need for an ES to access that data is critical. This paper presents two approaches to the development of ES-DBMS interfaces, both query-language based. One approach uses a procedural attachment to the ES language to obtain the required data via the DBMS query language, while the other one uses a separate interface program between the ES and the query language of the DBMS. The procedural attachment is able to acquire data from a DBMS at a faster rate than the interface program, however, the procedural attachment lacks knowledge of the DBMS schema. On the other hand, the interface program sacrifices speed but promotes flexibility, as it has the capability of selecting which DBMS to extract the required data from and allowing augmentation of schema knowledge outside of the ES. A disadvantage of the interface approach is the amount of time involved in data retrieval. The process of writing information to disk files is I/O intensive. This can be quite slow, particularly in PROLOG, the language used to implement the ES. Thus the use of such an interface is only suitable in applications such as design, where extremely fast I/O is not required