Deviation management is essential for the stability of the shop floor. When production deviates from the target, it is necessary to propose a systematic response to ensure continuous improvement. In practice, deviation management decision making requires experience and process understanding, often driven by limited expert resources. Large Language Models (LLMs) have recently excelled in many knowledge-processing tasks and can provide various opportunities for knowledge-intensive processes, where the implementation in such domain-specific areas still needs to be explored. This paper presents the concept of implementing LLMs in deviation management. Through applications incorporating various domain knowledge from the learning factory, this study also synthesizes the potentials and challenges and provides insights for future development.