<p>Generative artificial intelligence (GenAI) holds immense potential in various fields of manufacturing. However, GenAI applications to assist job scheduling in supply chains are still rare. In addition, user queries in natural language can have multiple interpretations that are indistinguishable unless the relevant technical terms are clearly articulated, which undermines the necessity of large language models (LLMs). For these reasons, a domain-specific automated modelling (DSAM) system is built in this study to facilitate schedulers to formulate and solve scheduling problems in three-echelon supply chains, which have been recognized as NP-hard. In the proposed methodology, users first enter their scheduling requirements in natural-language prompts through the system interface. A deep neural network (DNN), instead of LLM, is used to parse users’ queries to build the extended five field notation (EFFN) for the scheduling problem. Based on the EFFN, a customized genetic algorithm (GA) is automatically generated to help solve the job scheduling problem. The DSAM system has been applied to an experimental three-echelon supply chain. According to the experimental results, the DNN demonstrated high query parsing accuracy in training data and were effective at classifying scheduling requirements that have not been previously learned. Therefore, a viable strategy is to feed a large number of queries into the DNN for learning, since users scheduling requests remain limited due to the dedicated nature of the DSAM system.</p>