ARIMA Forecasting with LLM-Powered Multi-Agent Coordination for Omnichannel Retail KPIs

This paper presents a novel artificial intelligence (AI) agent-based framework for forecasting key performance indicators (KPIs) in omnichannel retail environments. Traditional retail forecasting often relies on historical or budgeted KPIs, which limits the ability to respond proactively to market dynamics. We address this gap by integrating time series machine learning (ML) models-specifically, ARIMA-with large language model (LLM)-powered agents within a cloud-based architecture. Our approach enables automated, accurate forecasting and the seamless delivery of predictive insights directly into retailers' operational workflows. The methodology is validated using real-world omnichannel retail data, with performance evaluated at the store and department levels. The system further leverages AI agents for data analysis and reporting, offering actionable recommendations to enhance model accuracy and business outcomes. This work demonstrates the potential of combining advanced ML techniques with agentic reasoning to support data-driven decision-making, improve inventory management, and optimize customer experience across multiple retail channels.