Organisations should move when privileges are created and consumed inside short runtime flows that leave little or nothing for periodic review. At that point, certification alone cannot govern the decision, because the useful control point is the moment access is issued.
Why issuance-time controls become the right control point for agents
Access reviews work best when there is a stable entitlement set to certify. Agents often do the opposite: they request, use, and discard access inside a narrow runtime window. When the decision that matters is “should this action be allowed now?”, the control has to move closer to issuance, where scope, duration, and context can be evaluated before the agent acts.
That shift matters because certification is retrospective. It can confirm whether a standing permission should exist, but it cannot prevent a short-lived overreach that already happened during execution. Issuance-time controls turn access from a periodic governance question into an active authorization decision.
For agent workflows, the practical signal is simple: if the access is created for a task, consumed by a task, and removed after the task, the governance model should treat it as a request-and-decision problem rather than a review-only problem. Just-in-Time Access and Zero Standing Privilege Guide is useful here because it frames the same timing problem for temporary privilege.
What changes in the control design at issuance time
Issuance-time control means the organisation decides access at the moment it is granted, not later in a certification cycle. That usually requires tighter scope, shorter duration, and stronger context checks than a broad access review process ever enforces. For agents, the important question is not only who or what owns the identity, but whether the requested action is narrowly bounded enough to be safe for the runtime event.
This is where reviews and issuance controls complement each other rather than compete. Reviews remain valuable for discovering drift, orphaned access, and recurring design flaws. But once an agent’s useful authority is ephemeral, the durable control objective shifts to preventing excess access from being issued in the first place. Access Reviews and Certification Guide helps anchor the retrospective side of that model, while AI Agent Authorisation Guide covers per-action decisions and delegated authority for agents.
In practice, issuance-time controls are strongest when they are tied to the actual action being requested, the minimum time needed, and the least privilege required to complete that action. That is much closer to operational authorization than to annual or quarterly hygiene.
When organisations should make the switch
The switch becomes warranted when periodic review no longer matches the lifecycle of the privilege. Three signs usually appear together: the agent’s permissions are short-lived, the permissions are task-specific, and the useful evidence exists at request time rather than at review time. At that point, keeping access reviews as the primary control creates false confidence because the control is looking in the wrong place.
Organisations should also move when access is being assembled dynamically from tokens, delegated permissions, or temporary elevations that are hard to reconstruct after the fact. In those cases, an access review may tell you that a role exists, but not whether the correct decision was made for the exact action the agent performed. Privileged Access Management Guide and AI Agent Observability, Audit and Incident Response Guide are relevant because they connect authorization, session control, and post-event traceability.
That does not mean reviews disappear. It means the operating model changes: reviews become a backstop for entitlement hygiene, while issuance-time controls become the primary gate for agent execution. The more autonomous and high-frequency the agent flow, the more this shift is justified.
Risk and Threat Considerations
When organisations keep relying on periodic reviews for short-lived agent access, they create a window where excessive privilege can be used before anyone certifies it. That weakens containment, makes privilege creep harder to see, and increases the chance that an agent can perform actions that exceed the intended task scope.
Failure mechanism: The access decision is made too late, so the agent can receive and use authority before the next review cycle notices that the entitlement was too broad, too long-lived, or too easy to reuse.
Impact: Unnecessary exposure accumulates at runtime, and the organisation loses the ability to stop over-privileged actions at the point they matter most, during execution.
Standards & Framework Alignment
This section maps relevant standards and security frameworks to the operational risks and controls described in this guidance.
OWASP Agentic AI Top 10 and OWASP Non-Human Identity Top 10 address the attack and risk surface, while NIST SP 800-53 Rev 5 and NIST CSF 2.0 set the governance and control requirements practitioners need to meet.
| Framework | Control / Reference | Relevance |
|---|---|---|
| NIST SP 800-53 Rev 5 | IA-5 — Authenticator Management | Controls the issuance and lifecycle of credentials used in short-lived access flows. |
| AC-6 — Least Privilege | Agents need narrowly scoped authority at the moment access is granted. | |
| AU-6 — Audit Record Review, Analysis, and Reporting | Reviews remain useful as backstop evidence for runtime authorization decisions. | |
| Recommendation — Issue and expire agent credentials on the shortest workable lifecycle. Constrain each agent request to the minimum required privilege. Review runtime audit evidence to validate issuance-time access decisions. | ||
| OWASP Agentic AI Top 10 | ASI03 — Identity & Privilege Abuse | Agent access decisions can be overbroad or reused beyond intended scope. |
| ASI02 — Tool Misuse | Issuance-time controls reduce the chance that agents misuse granted tools. | |
| Recommendation — Apply per-action authorization to prevent agent privilege abuse. Gate tool access at request time with task-scoped policy decisions. | ||
| OWASP Non-Human Identity Top 10 | NHI-05 — Overprivileged NHI | Short-lived agent access still becomes risky when it is granted too broadly. |
| Recommendation — Replace broad standing access with just-in-time, least-privilege issuance. | ||
| NIST CSF 2.0 | PR.AA-05 — Least Privilege | Directly supports shifting from periodic review to access issued only as needed. |
| Recommendation — Use least privilege at issuance instead of relying on later review. | ||
Practitioner Guidance
What to prioritise: Move first on agent flows where the permission is both high impact and short duration. Those are the cases where access review adds the least value and issuance-time policy adds the most.
What to verify: Make sure the system can express task scope, time bound access, and approval or policy context at the moment access is issued. If you cannot prove those attributes at request time, you are still depending on retrospective governance.
Common mistake: Treating issuance controls as an extra layer on top of reviews instead of a replacement for the primary decision point in ephemeral flows. For agentic access, the right question is often whether the entitlement should exist at all, not whether it can be reviewed later.
Practitioner takeaway: When the access exists only long enough to be used, governance has to move to the issuance event, because that is the last point at which the organisation can still shape the blast radius.
Related resources from NHI Mgmt Group
- When do NHI access reviews create more value than a one-time cleanup?
- When should organisations move from vault-centric PAM to real-time privileged access controls?
- How should security teams run access reviews for non-human identities?
- How do organisations reduce the dwell time of exposed credentials at scale?