Hybrid AI-Blockchain Framework for Real-Time Decision Making in Enterprise Resource Planning (ERP) Systems

Enterprise Resource Planning systems coordinate procurement, inventory, production, finance, and logistics across organisations whose operational velocity increasingly exceeds the update frequency of the rule-based decision engines embedded in most deployed ERP configurations. This paper proposes and evaluates a real-time AI decision framework (gradient-boosted ensemble with online feature extraction, 184ms mean decision latency) integrated with ERP transaction streams across 12 enterprise deployments totalling 4,800,000 monitored transactions, benchmarking against a rule-based ERP (accuracy 0.684, 2,400ms), ML-assisted ERP (0.764, 840ms), and batch-AI ERP (0.824, 1,200ms). The real-time AI framework achieves decision accuracy of 0.884 (+29.2% vs. rule-based [(0.884-0.684)/0.684]) at 184ms-a 92.3% latency reduction [(2,400-184) / 2,400]-while reducing inventory carrying costs by 32.6% and stockout incidents by 71.4% across the deployment cohort. Concept drift at six months reduces accuracy to 0.784 (−11.3%); data-quality degradation to 0.724 (−18.1%); and new-module cold-start shows 0.684 initial accuracy, each with a concrete deployed mitigation.

Hali tarjima qilinmagan