Aplicación de servicios de OpenAI para la generación automática de tareas en Odoo a partir de transcripciones de reuniones

Abstract

Abstract:This work addresses the gap between what is discussed in project meetings and what is recorded in Odoo at GAHEOS S.A. This gap, sustained by manual processes, consumes approximately 10 hours per week, resulting in omissions of around 12%, which negatively impacts and complicates traceability, planning, and client trust. A semantic workflow was designed and implemented to enable the integration of transcripts, supported by Retrieval-Augmented Generation (RAG) and vector search and evidence indexing strategies using pgvector in PostgreSQL. Odoo's ORM plays a key role in converting agreements into tasks. This solution utilizes human validation supported by a verifiable log for decision-making. In a production pilot, the system managed to reduce the time for registering tasks and activities from 15 to 3 minutes, decrease rework by between 3% and 4%, and reduce task duplication to below 3%, thereby strengthening customer confidence.