Browsing by Author "Calle Cabrera Alex Xavier"

Now showing 1 - 1 of 1
  • No Thumbnail Available
    Item
    Implementación de un asistente conversacional inteligente conectado a un sistema de comercio electrónico para compras automatizadas en COAVPRO
    (Instituto Superior Tecnológico Particular Sudamericano, 2026) Calle Cabrera Alex Xavier; Carriel Cepeda Elkin Sebastián
    This degree project examines the impact of integrating conversational artificial intelligence technologies within the e-commerce sector, specifically in COAVPRO, as a response to the deficiencies associated with traditional online shopping processes, such as manual product browsing, lack of customer support outside business hours, and the absence of a personalized shopping experience that led to customer frustration and purchase abandonment.The main objective of this research is to design and implement an intelligent conversational assistant integrated into COAVPRO's e-commerce system, capable of automating the sales process through natural language interaction, real-time stock and price verification, and personalized multichannel attention. To achieve this, an applied and mixed research methodology was implemented, including a rigorous analysis of the commercial environment, iterative development using the SCRUM methodology across four sprints, and the conceptual and technical design of a solution that integrates a local large language model through Ollama, semantic vector search using PostgreSQL with pgvector, and the Model Context Protocol (MCP) to guarantee data integrity throughout every interaction.The most relevant results indicate that the use of MCP tools completely eliminated the generation of incorrect information regarding prices and stock availability, achieving zero hallucination cases in critical transactional data. Furthermore, the semantic search system reached 84.7% precision in natural language product queries, and the automatic resolution rate of customer consultations exceeded 81% without human intervention, while providing uninterrupted 24/7 attention through the Telegram messaging channel.This study demonstrates that the integration of large language models with transactional e-commerce systems through MCP architectures not only improves operational efficiency and customer satisfaction, but also establishes a reliable, scalable, and cost-effective technological alternative to drive digital transformation in local e-commerce, positioning conversational commerce as a viable strategy for businesses seeking to modernize their sales channels.