Implementation of Artificial Intelligence for Multimodal Search in a Sales and Inventory System: A Case Study in a Peruvian Hardware Store

Meza N.A.N.; Salcedo M.A.J.; Rivera R.N.M.; Contreras R.D.O. · CSAI 2025 - Proceedings of 2025 9th International Conference on Computer Science and Artificial Intelligence (2026) · DOI: 10.1145/3788149.3788161

Resumen

This study implemented a sales and inventory system with multimodal artificial intelligence (text, image, and voice search) in a Peruvian hardware store, integrating the Gemini AI model with the Model Context Protocol (MCP) to optimize inventory management and improve customer experience in the specialized hardware sector. Through a pre-post quasi-experimental study over 2 months and 1 week, using three-layer architecture (PHP, JavaScript, MySQL) and scientific metrics including the specifically developed Economic Growth Indicator (ICE), the study achieved a 23% increase in sales (S/475,265 to S/584,578), 31.3% improvement in inventory turnover, 27% reduction in non-moving stock, 42.4% decrease in customer service time, and an ICE of 20.8% indicating significant integrated economic growth, where voice search was preferred (43%) although text search showed higher precision (94.3%). The results exceed previous retail studies with a projected ROI of 544.6% and investment recovery in 1.86 months, with technical categories (electric tools, plumbing) showing the greatest benefits, demonstrating the transformative potential of multimodal AI in specialized Peruvian retail and providing a replicable methodological foundation for other sectors.

Instituciones: Universidad Continental