This article presents an innovative approach to automating and supporting decision-making processes in maintenance using an intelligent assistant. The proposed solution combines the advantages of modern language models, self-learning mechanisms, and a modular architecture that allows for adaptation to the specific requirements of production environments. The research aimed to verify the system’s effectiveness, flexibility, and resilience in real-world conditions, as well as to assess its potential for integration with automation modules and enterprise IT systems. An in-depth analysis of implementation challenges was conducted, focusing on critical aspects such as the quality and completeness of training data, the reliability of communication between modules, and the system’s self-improvement during ongoing operation. A detailed concept for organizing the knowledge base and validation mechanisms was presented. Practical tests demonstrated that the implemented system effectively supports decision-making processes at production stations, enabling the automation of operational tasks and streamlining maintenance activities. The importance of continuous self-improvement, which leads to systematic efficiency gains, was emphasized. The article concludes with an indication of future development directions, including further optimization of input data quality, expansion of system functionality, and adaptation of the solutions to other industrial sectors.