ChatBanker: la solución a la búsqueda manual para especialistas de procesos bancarios

Abstract

Abstract: The work presents the development of an artificial intelligence-powered chatbot, focused on offering an efficient document management system and fast and accurate responses. It emerged as a response to the growing need of financial institutions to digitize and have immediate access to their processes, manuals, and guides. An agile approach was adopted for its development through the SCRUM methodology. Programming used Python as the coding language and SQLite3 as the database, integrating technologies such as HTML, CSS, and Bootstrap. Its main feature is the use of embeddings for character storage and cosine similarity to identify semantic similarities, in conjunction with the OpenAI API, which allows generating contextual responses. The results show that the chatbot achieved a 90% accuracy rate in responses, with an average response time of 4 seconds. Regarding usability, an adaptation of the SUS questionnaire was used, which gave a score of 91/100, indicating a satisfactory level.