Given a feature-based CAD model of a part’s geometry, the problem is to select the shape and size of bar stock from which the part can be made. It is assumed that standard stock shapes and sizes available in the workshop varies, so the task is to select the best suited stock geometry and cut-off length from the available inventory. Automatic workpiece selection involves the interrogation of a part database to determine the part’s shape characteristics, and to match the shape with available geometries. The requirements of a system to perform this task are presented along with several methods of determining the appropriate shape — GT coding, shrinking box, silhouette projection, point network, principal machining directions, and face extension. None of these methods is sufficiently robust by itself, so a combination of methods is required. Based on this hybrid technique, a system to perform automatic workpiece selection was designed and implemented. A coding scheme is used to select the workpiece shape, based on primary features and their relationships. Once the shape is selected, it is matched with available workpiece shapes. The workpiece is then sized using a version of the bounding box method.