The Agentic Trust Cycle: An Operating Model for Securing and Scaling the Agentic Enterprise

Enterprises deployed AI agents at scale in 2024–2025; the governance primitives required to run them safely — cryptographic agent identity, semantic screening with organizational floors, ephemeral sandboxes, enterprise-owned agent registries — arrived in 2025–2026, opt-in, into estates that predate them. The result is governance debt by construction. This paper proposes an operating model for paying that debt down and never re-accruing it: the Agentic Trust Cycle, four connected disciplines — a governance framework, control-plane operationalization, an agent production factory, and organizational change management — bent into a continuous cycle around business transformation. It argues that governance debt and the agent-scaling problem are one problem seen from two desks, that the scarce capability is no longer designing a control plane but operating one, and that best practice is unified governance with federated enforcement. It defines seven runtime controls for a secure agent runtime, maps each to the published standards that require it — NIST AI RMF 1.0, NIST AI 600-1, NIST AI 100-5, ISO/IEC 42001, OWASP, MITRE ATLAS — and describes a companion reference implementation. Written for enterprise security, platform, and transformation leaders, and for the consulting and vendor ecosystems that serve them. Views are the author's own; vendor capabilities reflect public documentation as of September 2026.