Runtime Governance Architecture: Consistent Governance Execution for Enterprise Systems and Autonomous Agents

Enterprise environments increasingly consist of heterogeneous systems, distributed automation pipelines, and autonomous agents that continuously propose and execute system mutations. To enable consistent governance across such environments, this paper introduces the Runtime Governance Architecture (RGA), a framework that compiles governance intent into canonical artifacts and evaluates them through a deterministic governance execution model. Maintaining consistent governance across these environments is difficult because identity semantics, action representations, policy languages, authorization models, and audit mechanisms differ across platforms. We first identify the governance functions that must be performed consistently across enterprise stacks and the challenges of implementing them in distributed environments. We then introduce the Artificial Intelligence Governance Control Plane (AGCP) as a reference implementation of the governance execution model defined by RGA. AGCP centralizes governance evaluation into a deterministic pipeline capable of producing consistent authorization decisions and ordered lifecycle evidence. Within this model, enterprise systems and autonomous agents operate as interacting state machines whose transitions are governed by RGA. This architecture reframes enterprise governance as a runtime transition-control system in which proposed system mutations are evaluated through a deterministic governance pipeline prior to execution.