AI-Driven Demand Forecasting and Inventory Optimization for Resilient U.S. Manufacturing Supply Chains: A Data-Driven Framework

Abstract

This quantitative study examined how AI-driven demand forecasting and inventory optimization strengthened resilience across U.S. manufacturing supply chains. A longitudinal quasi-experimental design combined survey responses from 602 professionals with operational data obtained from 50 manufacturers, 4,860 product–location series, 758,160 weekly demand observations, more than six million inventory transactions, and 417 verified disruption episodes over 36 months. The measurement model demonstrated satisfactory reliability and validity, with Cronbach’s alpha values ranging from .842 to .931 and composite reliability values ranging from .887 to .944. Weighted absolute percentage error declined from 18.7% to 13.6% among AI-integrated manufacturers, compared with a reduction from 18.9% to 18.2% in the matched comparison group. Stockouts decreased by 34.5%, material shortages declined by 31.8%, and order fill rate increased from 91.8% to 96.1%. Inventory turnover rose from 6.2 to 7.4 annual cycles, while average inventory value, shortage costs, obsolescence expenses, and emergency transportation costs declined. Disruption-detection time decreased from 18.6 to 9.8 hours, response time fell from 31.4 to 18.2 hours, and recovery time declined from 12.8 to 8.1 days. Structural analysis explained 53.1% of forecast quality, 64.2% of inventory optimization, and 68.7% of supply-chain resilience. Forecast quality and inventory optimization significantly mediated the relationship between AI capability and resilience. Data quality, digital maturity, analytical competence, supplier integration, and environmental uncertainty produced significant moderating effects. The findings established that integrated AI forecasting improved inventory performance, operational continuity, and disruption recovery across heterogeneous manufacturing environments.