DEVELOPMENT OF A GENERATIVE ARTIFICIAL INTELLIGENCE ASSISTANT FOR THE RESOLUTION OF INCIDENTS IN INDUSTRIAL PROCESSES

Abstract

The application of Generative Artificial Intelligence (GAI) in industrial environments is opening new possibilities for automating the consultation of technical information and supporting operational decision-making. In processes where failures may have multiple causes and technical documentation is often unstructured or inaccessible, having a system capable of identifying the root causes of errors, classifying and optimizing responses, and providing clear explanations can offer a significant operational advantage. This paper focuses on the integration of technical data into a GAI-based virtual assistant, with the goal of enabling it to narrow down the origin of failures in a real industrial printing process, as part of a technological demonstration at Telefónica’s La Cabina. The core of the work lies in the collection, organization, and delivery of manuals, error reports, and operational documents, structured in a way that allows the assistant to interpret and reason over them accurately. The methodology followed includes the identification of critical points in the process, the selection and preparation of relevant documentation, and the training of a generative model adapted to the industrial domain. Finally, functional tests have been conducted to validate the usefulness of the assistant as an explanatory diagnostic tool, with potential to be scaled to more complex industrial processes.