Meeting customer demands for higher quality is a major business focus in the development of steel products. The quality of steel can be increased in many ways, for example, by improving its toughness or its resistance to atmospheric corrosion or by making thinner sections without sacrificing strength. As the largest steel supplier in the world, Nippon Steel Corporation (NSC) receives many kinds of customer requests for new steel products with increased quality, sections of different size, and so on. Customer requests have become more complex, and their quality requirements have increased. At the same time, quality-design experts are expected to decrease the time to judge whether production of a desired product is possible and to design the product. Furthermore, fewer quality-design experts are available to design the new products. Problems in the quality design of steel products have become important. To deal with these problems, NSC developed a design support system with a production database and a retrieval mechanism. Unfortunately, this system was efficient only for the design of products that were similar to products produced in the past. To overcome the limitations of the design support system, NSC undertook the challenge of developing the quality-design expert system (qdes). qdes is fully operational and provides valid designs for shaped-steel products. The application of ai to design problems is generally considered difficult in terms of knowledge acquisition and system modeling because of the combinatorial explosion that is inherent in a problem with a huge solution space. The achievement of qdes is a milestone in the application of ai to design. This paper analyzes the experts’ quality-design process and presents the technical points required to create the expert system model.