AI-Native ERP Systems: A Design Science Framework for Intelligent Enterprise Decision Automation

Abstract

Enterprise Resource Planning (ERP) systems underpin the operational fabric of modern enterprises, yet their predominantly rule-based, static architectures constrain adaptability in dynamic market conditions. This paper proposes and formalizes the concept of AI-native ERP—an architectural paradigm in which artificial intelligence is not an optional overlay but a foundational, deeply integrated layer spanning data ingestion, intelligent decision-making, autonomous workflow execution, external integration, and human interaction. We present a five-layer reference architecture, a formal decision-engine model combining Bayesian-calibrated machine learning prediction, constraint-satisfaction rule validation, and multi-objective utility optimization, and a risk-aware escalation mechanism parameterized by a composite risk function. Six enterprise use cases—spanning logistics, accounts payable, tax compliance, procurement, financial operations, and cross-functional intelligence—are analyzed to demonstrate practical applicability. Comparative analysis suggests the proposed architecture has the potential to substantially reduce manual process dependency and improve decision quality relative to AI-augmented ERP systems. Challenges including model governance, adversarial robustness, explainability, and organizational change management are critically examined. Future directions encompassing agentic ERP, federated enterprise learning, and self-healing architectures are delineated.