U.S. pharmaceutical supply chains are increasingly exposed to disruption risks including manufacturing downtime, logistics failures, demand surges, and regulatory bottlenecks. These disruptions can propagate rapidly and result in drug shortages, delayed patient access, and elevated operational costs. This study proposes and evaluates an autonomous AI decision system that integrates real-time signals from manufacturing, inventory, logistics, and quality operations to reduce disruption detection time and accelerate recovery actions. The proposed architecture combines predictive analytics for early disruption sensing with reinforcement learning–based decision policies for adaptive response, including production resequencing, inventory rebalancing, and dynamic shipment rerouting. A bounded autonomy governance layer is incorporated to ensure that high-impact decisions (e.g., batch release exceptions, cold-chain risk decisions, compliance deviations) are escalated to human experts for validation under FDA-aligned constraints. The approach is assessed using scenario-based simulation experiments representing three common disruption categories in pharmaceutical supply chains: (i) production delays, (ii) transportation failures, and (iii) regulatory/quality constraints. Results demonstrate that autonomous decision execution improves resilience outcomes by shortening the sense–decide–act cycle, reducing stockout exposure windows, improving on-time delivery performance, and maintaining compliance through structured human-in-the-loop approval gates. The findings support a practical roadmap for deploying safe, explainable, and regulation-aware autonomy in U.S. pharmaceutical supply networks. All evaluations are conducted under controlled scenario-based simulation settings to ensure repeatability and comparability between baseline and autonomous policies.
Sharma* et al. (Thu,) studied this question.