Journal of Pioneering Artificial Intelligence Research

Abstract

Global supply chains are increasingly exposed to compounding disruptions spanning geopolitical events, commodity price volatility, natural disasters, trade policy shifts, and regulatory compliance risk. Existing enterprise procurement tools address these domains in isolation, requiring analysts to manually synthesis fragmented signals across disconnected systems. This paper presents NAUTILUS Terminal, an agentic supply chain intelligence architecture that unifies real-time multi-source data ingestion, large language model (LLM)-assisted classification and recommendation, and structured risk fusion into a single decision-support interface. The system decomposes procurement intelligence into seven autonomous API agents covering commodity pricing, geopolitical incidents, seismic events, wildfire propagation, maritime port status, foreign exchange dynamics, and trade news, whose outputs are synthesized by a central reasoning agent into actionable sourcing directives, risk scores, and compliance assessments. We describe the system architecture, data integration methodology, LLM orchestration strategy, and evaluate the system across 40 commodity categories. NAUTILUS demonstrates that agentic composition of heterogeneous live data streams, mediated by structured prompting and deterministic reasoning, produces procurement intelligence that is broader in coverage and lower in latency than conventional enterprise solutions.