Diseño de asistente virtual para optimizar cumplimiento tributario en empresa fabricante de calderas y equipos

Abstract

The present study examines in detail the existing complexities within the Ecuadorian tax system, specifically as they apply to a company whose core business is manufacturing. This project explains the development of a solution based on cutting-edge technology that improves tax administration and reduces the risk of penalties arising from tax obligations that are not fulfilled within the deadlines established by the regulatory authority, the Internal Revenue Service (SRI). The primary objective is the creation of a virtual tax assistant that, through the use of digital tools, is capable of timely interpreting current tax regulations and automating pending tax obligations. Throughout this project, a mixed methodology was employed, with an exploratory–experimental approach aligned with the DCOR study model. To implement each of the three virtual tax assistants, two types of language models are used—OpenAI and Gemini—along with a vector database, the n8n platform, and the regulations of the Income Tax Law (LRTI). The results obtained show a significant reduction in response time, with an average response time of 5.63 seconds, and each batch of 100 electronic documents being processed in 28 minutes. Likewise, a 100% accuracy rate was achieved in identifying tax obligations that are overdue or approaching their due date through the documentary upload of the RUC certificate. Implementing automation (RPA) and artificial intelligence is a key strategy, as it not only helps reduce expenses related to external advisors and operational workload, but also enables the development of a more organized, efficient, and, above all, preventive tax management system. Keywords: Tax Assistant, Taxpayer, Artificial Intelligence, Obligations