A Hybrid Agentic AI Framework for Intelligent Supply Chain Analytics

Summary

Supply chain planners often need database querying, KPI analysis, demand forecasting, and performance diagnosis, tasks that span data engineering, operations research, and domain expertise. This paper proposes a hybrid agentic system in which a coordinator agent interprets user intent and delegates work to specialized agents. The architecture supports both open-ended exploratory questions and deterministic, structured workflows, while keeping domain logic in specialist agents and prompts. The authors describe the design as modular and auditable, with prompt-centric development intended to reduce the cost of extending functionality. They evaluate it in a test environment replicating multi-echelon inventory management operations. The multi-agent design reaches 90% accuracy, comparable to a single-agent baseline, while reducing input-token usage by roughly fourfold. Case studies demonstrate interpretable detection of suboptimal decisions and automated forecast optimization. The results suggest that agentic architectures can make supply chain decision-support tools more scalable and accessible, although the reported evaluation is tied to the described test environment.