TL;DR: Autonomous agents are multiplying fast, but weak workforce identity controls are still the foundation problem: roughly half of employees hold excessive or privileged access, and overloaded certification workflows push people toward rubber-stamp approvals, according to Clear Skye. AI governance will fail if the humans provisioning it remain under-governed.
At a glance
What this is: This is a Clear Skye analysis arguing that AI agent governance is being built on unresolved workforce identity problems, especially excessive access and approval fatigue.
Why it matters: It matters because IAM and IGA teams cannot govern autonomous access cleanly if the human processes that provision, approve, and certify it already produce risky access decisions.
Context
AI agent governance is not starting from a clean slate. The article argues that the underlying problem is workforce identity control, where people still provision access, approve entitlements, and certify permissions with too much friction and too much standing privilege.
In practical IAM terms, the issue is not just policy design but whether users can make accurate access decisions at all. If certification workflows are overloaded or hard to understand, the control may exist on paper while day-to-day behaviour drifts away from governance intent.
Clear Skye frames this as a sequencing problem rather than a technology problem. The article’s position is that workforce identity is the foundation layer for AI governance, and that is a typical failure pattern in organisations trying to scale governance faster than usability.
Key questions
Q: What breaks when workforce identity reviews become too complicated?
A: When reviews are too complicated, people stop validating access and start clearing work. That creates rubber-stamp approvals, hidden privilege creep, and a mismatch between policy and reality. The control still exists in the system, but the organisation loses confidence that certifications are actually reducing risk across human and delegated access.
Q: Why do AI agents complicate privilege governance?
A: AI agents complicate privilege governance because they can request access, use tools, and execute actions faster than human review cycles. That creates a larger attack surface for over-permissioning and misuse. Teams need continuous monitoring and strict task scoping so an agent cannot retain access beyond the approved workflow.
Q: How can teams tell whether access governance is actually working?
A: Look for short revocation times, low rates of stale entitlements, and repeatable access review outcomes across systems. If accounts remain active after role changes or offboarding, governance is not effective. Good measurement focuses on whether access is removed when it stops being justified.
Q: Should organisations sequence AI governance after workforce identity cleanup?
A: Yes. AI governance becomes much harder when the human provisioning layer is already overloaded with excessive access and unclear review workflows. Organisations should stabilise workforce identity first so agent access is inherited from a governance model that people can actually use and understand.
Technical breakdown
Why excessive human access becomes an AI governance problem
AI agents do not govern themselves into safe access. People provision the agent, approve its permissions, and inherit the same access culture that already exists for the human workforce. When employees hold more access than their jobs require, the agent often begins with an over-permissioned baseline. That turns workforce identity from a background IAM issue into the control point that determines whether agent governance is bounded or bloated from the start. The technical flaw is not the agent alone. It is the access model feeding the agent.
Practical implication: review the human provisioning path before treating AI agent access as a separate governance stream.
Why certification workflows fail when users cannot understand entitlements
Access review quality depends on the reviewer understanding what is being certified. When entitlement structures are opaque, reviewers cannot reliably tell whether access is still needed, so they approve to close the task rather than to validate the risk. That creates a governance illusion: the workflow completes, but the access decision is not materially tested. In identity terms, the control is nominally present yet operationally weak because comprehension is too expensive. This is especially damaging where AI agent permissions inherit from human roles or delegated approvals.
Practical implication: simplify entitlement presentation so reviewers can actually distinguish necessary access from inherited excess.
How workflow friction turns governance into help-desk behaviour
The article’s usability point is important because people adapt to friction. If access requests and certification steps are awkward, users route decisions through the help desk or approve quickly to remove the burden. That shifts governance from deliberate control to throughput management. The result is not just inefficiency. It is a predictable failure mode where controls on paper no longer match real decision-making. For IAM and IGA teams, the mechanism matters more than the complaint: overcomplicated processes create compensating behaviour that quietly weakens access oversight.
Practical implication: treat workflow friction as a control risk, not only a user-experience defect.
NHI Mgmt Group analysis
Workforce identity is the upstream control plane for AI agent governance: autonomous access cannot be governed well if the humans provisioning it are already operating with excessive privilege and low-quality approvals. The article is right to invert the usual sequence because agents inherit the access culture that created them. The implication is that AI governance programmes must start with the human identity layer, not bolt it on later.
Approval fatigue creates governance drift before any agent is deployed: when certification and request flows are too complex, reviewers stop performing real validation and start clearing work queues. That means the organisation is not actually governing access, only recording that governance happened. The practitioner takeaway is that process usability is now a security property, not an administrative convenience.
Identity adoption is a control, not a soft factor: if people do not use the workflow as designed, the formal policy is not the operational policy. That is especially true in ServiceNow-centric environments where governance depends on user participation. The broader lesson is that identity programmes fail as much through friction as through missing rules.
AI governance inherits every weakness in human IGA: if workforce access is already overextended, agent access will compound the problem rather than contain it. That is a structural governance issue, not a feature gap. Practitioners should read AI-agent oversight as an extension of workforce identity maturity, because the same provisioning habits shape both.
Right-sized access must precede autonomous delegation: the market is moving toward AI control frameworks, but the mature programme will still be the one that can prove human entitlements are understood, reviewed, and current. Clean agent governance depends on clean workforce governance first. The operating principle is simple: sequence the foundation before the automation.
From our research library:
- Nearly 60% of IT leaders cite restrictive cost and complexity as a weakness of legacy identity governance, according to the 2025 State of Identity Governance Report.
- Read next: IGA Buyer's Guide
What this signals
Identity adoption is now a security control: governance that users quietly route around is not a durable control model, even if the policy stack looks complete on paper. For AI agent programmes, that means the human provisioning path and the certification experience deserve as much design attention as the agent policy itself.
Workforce identity maturity sets the ceiling for agent governance: if a team cannot explain, review, and right-size human access, it will struggle to govern delegated or autonomous access with any consistency. The practical signal is to treat AI oversight as an extension of IGA maturity rather than a separate AI project.
For practitioners
- Tighten workforce entitlement modelling Map which human roles are actually provisioning or approving AI agent access, then remove inherited privilege that is not required for those tasks.
- Redesign certification views for comprehension Present reviewers with plain-language entitlement context, business ownership, and recent usage so approvals reflect understanding rather than task completion.
- Measure approval quality, not just completion rates Track how often reviews are approved without changes, how many items are escalated to help desk, and where reviewers cannot interpret the access being certified.
- Treat AI provisioning as part of workforce IGA Fold agent provisioning and delegation into the same governance operating model used for workforce identity so access decisions are not managed in separate silos.
- Reduce workflow friction before scaling automation Remove unnecessary approval steps, repeated data entry, and opaque role naming so the access path stays usable enough for deliberate decision-making.
Key takeaways
- The central risk is not AI agents alone, but the fact that they are being built on workforce identity processes that already tolerate excessive access and weak review quality.
- The article’s core evidence is behavioural: overloaded access workflows encourage help-desk deflection and rubber-stamp approvals, which makes formal certification unreliable.
- Practitioners should treat human identity governance as the foundation for AI agent control, because access quality at the source shapes everything that follows.
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 addresses the attack surface, NIST SP 800-53 Rev 5, NIST CSF 2.0 and NIST SP 800-63 set the technical controls, and ISO/IEC 27001:2022 defines the regulatory obligations.
| Framework | Control / Reference | Relevance |
|---|---|---|
| NIST SP 800-53 Rev 5 | IA-5 — Authenticator Management | Credential lifecycle and review quality shape the workforce access baseline described here. |
| Recommendation — Apply IA-5 discipline to keep workforce credentials and approvals aligned with current access need. | ||
| NIST CSF 2.0 | PR.AA-05 — Access Permissions, Entitlements and Authorizations | The article is fundamentally about entitlement quality and reviewability across workforce and delegated access. |
| Recommendation — Use PR.AA-05 to keep access approvals understandable, reviewable, and right-sized before AI delegation expands. | ||
| NIST SP 800-63 | SP 800-63C — Federation | The piece references human identity workflows and their operating environment, where federation and access flow matter. |
| Recommendation — Align identity flows under SP 800-63C so governance decisions remain traceable across access paths. | ||
| ISO/IEC 27001:2022 | A.8.2 — Privileged Access Rights | The article highlights excessive workforce privilege as the upstream issue undermining governance. |
| Recommendation — Review privileged access rights under A.8.2 before extending governance to AI agent provisioning. | ||
| OWASP Agentic AI Top 10 | ASI03 — Identity & Privilege Abuse | The article’s AI-agent concern is that weak human governance can translate into over-privileged agent behaviour. |
| Recommendation — Map delegated agent permissions to ASI03 and constrain privilege inheritance from human approvers. | ||
Key terms
- Workforce Identity Security: Workforce identity security is the control layer that protects employee access across hiring, onboarding, support, role change, and offboarding. It combines identity proofing, access governance, and recovery controls so an attacker cannot exploit business processes to obtain or restore trusted access.
- Certification workflow: Certification workflow is the sequence of tasks and approvals used to confirm access remains appropriate. For large organisations, the workflow must be reliable, understandable, and auditable, otherwise reviewers delay decisions or approve without enough context, which weakens governance.
- Approval Fatigue: The point at which repeated approval requests cause users to stop evaluating each one carefully. In agent governance, this is a control failure mode because the human reviewer becomes desensitised, making the oversight layer ineffective even though the workflow still appears compliant.
- Delegated Access: Delegated access is permission granted to one identity to act on behalf of another user, service, or system. In NHI environments, this usually appears in OAuth-connected apps and automation tooling. It is powerful, but it must be tightly scoped and reviewed because it can persist long after the original business need ends.
Deepen your knowledge
NHI governance, agentic AI identity, and machine identity lifecycle are core topics in our NHI Foundation Level course, the industry's only accredited NHI security programme. If you are building or maturing an IAM programme, it is worth exploring.
Published by the NHIMG editorial team on July 5, 2026.
Updated on October 11, 2026.
NHI Mgmt Group, the independent authority on Non-Human Identity, IAM, and Agentic AI security. nhimg.org