AI-Enabled Decision Intelligence for Managing Material Shortages in Engineer-to-Order Supply Chains: A Reproducible Simulation Study

Abstract

Background: Engineer-to-order (ETO) manufacturing faces material shortages arising from changing requirements and uncertain supply. This study develops Artificial Intelligence–Supply Chain Decision Intelligence (AI-SCDI), linking shortage prediction to feasible interventions under cost and capacity constraints. Methods: Six specialists assessed framework content and parameter plausibility. A locked experiment used 12 synthetic cohorts and 6000 paired Monte Carlo replications per policy, supplemented by model-recovery diagnostics, matched policy ablations, calibration assessment, and joint sensitivity analysis. Results: Histogram gradient boosting achieved a mean precision–recall area under the curve of 0.7587, compared with 0.5399 for logistic regression and 0.7415 for random forest. Under baseline assumptions, AI-SCDI reduced disruption time by 30.0% and synthetic operating cost by 19.1% relative to no proactive intervention. At baseline capacity, estimated-net-benefit selection and abstention contributed more to cost reduction than impact weighting alone. Learner rankings varied across generating mechanisms, horizon-specific calibration showed risk overestimation, and mean net benefit was negative in 19 of 64 informed-parameter scenarios. Conclusions: AI-SCDI provides a reproducible, capacity-constrained decision heuristic. Its benefits remain conditional on synthetic assumptions; external validation and empirical cost calibration are required before industrial effectiveness can be claimed.