Computer-aided manufacturing systems are considered as ‘what-if’ processors, whereby the user suggests manufacturing features and for each one of them a tool and a machining strategy, whereas the system computes tool paths and generates a computer numerical control program. Given the multitude of factors for which decisions are needed, complexity is high. In addition, given the executive nature of computer-aided manufacturing systems, the user relies on experience in exploring parameter value combinations in order to obtain good results. In this work, structured generation of a representative subset of the parameter combination space is advocated using the Taguchi approach. This is applied to parts that are simpler than the average part, since they consist of a single, repeating, yet geometrically complex, machining feature involving sculptured surfaces, such as helical bevel gears. The lack of multiple features reduces the solution space, enabling a manageable number of experiment runs from which good results for roughing and finishing are attained. Statistical processing of these results also point to relative significance of machining strategy parameters within the particular space explored and for the particular weighted criteria employed, that is, rest material, machining time and gouge volume. Excellent results are obtained in this way, based on the machining simulation capabilities of modern computer-aided manufacturing systems and bypassing the inherent lack of intelligence of the latter.