Governing Agentic AI in Enterprise Operations: Architectural “Rails” for Safe, Deterministic, and Compliant Autonomous Systems
This paper argues that the introduction of agentic AI requires a substantial expansion of traditional enterprise architecture principles to address new behavioral, security, and governance risks emerging from non-deterministic AI systems interacting with heterogeneous operational platforms-ERP, HCM, CLM, asset management, workflow engines, and domain-specific applications.