AI agents create a new identity problem before they create an AI problem.
Autonomous software can request data, invoke tools and act across enterprise systems. The security model must answer who or what is acting, on whose authority, for what purpose, with which permissions and for how long.
Extend the identity control plane — do not bypass it.
Fiero's thesis is straightforward: the same disciplines that make workforce IAM reliable — authoritative identity, lifecycle, least privilege, policy, ownership and auditability — must evolve to cover agents and other non-human actors.
This makes existing Okta and Microsoft Entra architecture strategically important. Agent adoption should strengthen the case for modern IAM, not create a parallel credential layer that security teams cannot govern.
Questions the architecture must answer
- Who owns each agent?
- How is human authority delegated?
- What can the agent access, and why?
- How short-lived can access be?
- How is human → agent → system activity attributed?
- How are access, lifecycle and exceptions reviewed?
AuthorizationOkta / Entra
Identity is the control plane for the agentic enterprise.
Start by making the current IAM environment trustworthy, understandable and extensible.
Explore services ↗