Modelado de agente de servicios con IA para un operador económico autorizado en Colombia

Abstract

This graduate project from the Specialization in Information Technology Services Management at Universidad Santo Tomás aims to model an Artificial Intelligence-powered service agent (chat and voice) for an Authorized Economic Operator in Colombia, focused on improving customer service and advisory for internal and external clients in foreign trade services. To achieve this objective, the following development route was pursued: investigate available technologies to identify functional solutions applicable to intelligent agent modeling, establish the current advisory process flow to standardize conversation flow and AI service agent functionalities, and apply the AI service agent modeling as a use case to test functionalities and validate the proposed model's effectiveness in a real operational environment. Regarding highlighted results, a Minimum Viable Product (MVP) was successfully built and validated—an intelligent virtual assistant called VictorIA, with architecture based on the Llama 3 language model deployed on proprietary local infrastructure, demonstrating automated resolution rates, response accuracy, and response time reduction. The business model was constructed using Canvas methodology, identifying three priority segments (AEOs, Customs Brokerage Societies, and exporting SMEs) with a differentiated value proposition in security, agility, and compliance. Technical alternatives were analyzed across six critical functional blocks, justifying technological selection through comparative analysis. The digital transformation process was established and evaluated using MIT Sloan's Digital Maturity framework across five dimensions, and legal aspects were defined ensuring compliance with Law 1581/2012 and DIAN Resolution 089/2024.