ERP System Integration Architecture with Artificial Intelligence Models for Supply Chain Management and Inventory Control Optimization

Abstract

The increasing complexity of supply chain management (SCM) and the volatility of customer demand pose major challenges to traditional inventory control mechanisms embedded in Enterprise Resource Planning (ERP) systems, which are largely deterministic and limited in predictive capability. To address this limitation, this study adopts a Design Science Research (DSR) approach to design, develop, and evaluate an ERP–Artificial Intelligence (AI) integration architecture aimed at optimizing supply chain operations and inventory control. The proposed architecture employs a modular and loosely coupled framework that utilizes real-time transactional data from core ERP modules, including purchasing and inventory management, as inputs for a machine learning–based predictive engine. A Long Short-Term Memory (LSTM) model is implemented to capture non-linear temporal demand patterns and improve forecasting accuracy. In addition, a bidirectional data flow mechanism enables automated feedback of predictive outputs into ERP decision parameters, supporting closed-loop and data-driven inventory decision-making. The artifact is evaluated through case-based inventory simulations and backtesting using historical ERP data. The results demonstrate an average improvement of approximately 15% in demand forecasting accuracy compared to conventional ERP statistical methods, along with a significant reduction in stock-out frequency and improved inventory responsiveness. This study contributes prescriptive architectural knowledge for integrating AI into ERP systems and demonstrates how AI-enabled ERP architectures can transform SCM and inventory control into adaptive and predictive decision systems.