Decomposing SaaS into Hierarchical Agents: Center–Subcenter–Edge Architecture for the Agentic Era

Working Paper v0.2.5. This paper is part of the Agentic Governance Series. This paper proposes a structural decomposition of SaaS architectures in the agentic era, distinguishing three functional layers: Center, Subcenter, and Edge. We argue that AI-enabled agents fundamentally transform enterprise software by reallocating execution, verification, and liability across these layers. The core innovation is the introduction of the Operated Evidence Protocol (OEP): a portable, verifiable operational evidence layer that links contracting, insurance, and governance. Rather than treating SaaS as pure functionality provision, we model it as a governance stack in which responsibility, verification, and risk absorption become allocatable design variables. The paper formalizes conditions under which centralized control is efficient versus when subcenter specialization or edge autonomy dominates. We show how insurance, auditability, and evidence portability allow higher agentic delegation while preserving accountability constraints. This architecture provides the institutional backbone for AI-native enterprises and connects directly to the companion market-design paper on Insured Operated-System Markets. Companion paper (Markets): DOI 10.5281/zenodo.18666718.