Some important tasks in a production system are not easy to automate completely. Examples are the visual inspection and defect diagnosis tasks, which determine the quality of the system. The production system of a glass bottle plant was taken as a case study. Two experimentns were conducted. The first dealt with the inspection task. The results implied that the computer assistance should focus on simplifying the human decision process in inspection. The second experiment was a computer-simulated diagnosis task. It was found that the degree of computer assistance had significant effect on the improvement of human defect diagnosis performance.