Abstract: Using Artificial Intelligence Generated Content (AIGC) as the technical approach, a leather bag design method that can match user expectations was proposed by combining the Analytic Hierarchy Process. This work first established a sample library of leather bag styles through online collection and offline photography. Secondly, a semantic vocabulary library of leather bags was obtained through user interviews to describe users' design expectations, and then the Analytic Hierarchy Process was used to couple user expectations with leather bag design factors. Next, based on the leather bag style factors, all image samples were manually labeled, and the Low-Rank Adaptation (LoRA) method of large language models was trained. The obtained model was then tested and evaluated, and the optimal model was selected based on the criteria of "accuracy" and "generalization". Finally, the specific role of the LoRA model in the leather bag design process was illustrated through leather bag design practices. The conclusion shows that combining the Analytic Hierarchy Process with LoRA model training can effectively improve the output efficiency of Artificial Intelligence (AI) in leather bag styling and make the output results meet user expectations.