Pharmaceutical supply chains represent one of the most operationally demanding environments for artificial intelligence deployment, requiring continuous demand forecasting, expiration tracking, procurement automation, and multi-location coordination under conditions that permit minimal margin for error. This paper presents the design, implementation, and real-world deployment outcomes of Restocks AI, an autonomous multi-agent AI platform engineered to replace manual pharmaceutical supply chain operations at national scale. The system integrates predictive inventory engines, multi-agent decisioning frameworks, distributed task-processing infrastructure, and real-time data pipelines to autonomously manage end-to-end supply chain operations across geographically dispersed pharmacy networks. The platform was deployed across two pharmaceutical organizations in Uzbekistan spanning a combined 427 pharmacy locations across all 14 administrative regions, including one of the country's largest private pharmacy networks (150+ locations) and the largest government-operated pharmaceutical institution (277 locations). Deployment results demonstrated 70% reduction in manual operational workload, 30% acceleration of replenishment cycles, 65% reduction in pharmaceutical expiration waste, 70% reduction in stockout incidents, and a combined financial impact of 812 billion Uzbek sum. These results demonstrate that autonomous multi-agent AI systems can effectively manage pharmaceutical supply chains at national scale, offering a replicable architecture with significant implications for healthcare infrastructure, public health outcomes, and the broader application of multi-agent AI in critical industries.