AI and Predictive Analytics for Supply Chain Risk Management: Opportunities for U.S. Manufacturing Resilience

Recent shocks in the global supply chain such as pandemic related shutdowns and semiconductor shortages have shown the susceptibility of the U.S. manufacturing supply chains. Conventional risk management practices that are mostly based on historical averages and reactive practices have not been effective in unpredictable environments. This paper examines the connection between artificial intelligence (AI) and predictive analytics and reinforcing risk identification, forecasting, and resilience in the U.S. manufacturing industry. We assess the performance of predictive models including ARIMA, Prophet, and Random Forest using the secondary data from U.S. Census Bureau, Bureau of Labor Statistics, and publicly available logistics datasets. The most important measures of resilience are forecast error (MAPE, RMSE), inventory turnover, order fulfillment, and recovery time. The findings indicate that predictive analytics can greatly reduce errors in predictions as well as enhance the outcomes of resilience with the Random Forest being better in terms of reducing forecast error by up to 30% as well as increasing order fulfillment rates by 15% compared to traditional models. A model-derived association analysis using simulation-based performance metrics further indicates statistically significant directional relationships between predictive analytics capability and resilience indicators. These results highlight the value of predictive analytics as an input to operational decision-making. To managers, the study illustrates the use of digital tools to help them get actionable information on the risks of disruption and to policymakers, it is part of wider efforts to strengthen national supply chain resilience. This study can enhance the empirical understanding of AI-facilitated models and introduce feasible resilience metrics.