Identity governance’s unfinished business still blocks agentic AI

At a glance

What this is: NHI Mgmt Group's analysis, drawing on READI's blog post: Identiverse 2026 showed that identity governance fundamentals still lag while agentic AI dominates vendor messaging.

Why it matters: Identity and IAM teams need to separate near-term governance debt from long-term innovation narratives, because unresolved coverage and provisioning gaps weaken both human and machine identity programmes.

Context

Identity governance programmes still fail most often at the basics: knowing which applications are covered, keeping provisioning current, and keeping authoritative systems in sync. The conference theme may have shifted toward agentic AI, but the operational backlog behind human access governance has not disappeared.

That gap matters because governance debt compounds across human, machine, and autonomous programmes. IAM and IGA basics still apply when teams are trying to scale access reviews, automate joiner-mover-leaver workflows, or decide whether a new control is actually reducing manual effort.

Key questions

Q: How do teams know if identity governance is actually closing access gaps?

A: Look for whether offboarding, recertification, and entitlement review remove access across every linked record, not just the primary directory entry. If duplicate identities, stale accounts, or unmanaged service credentials still exist after lifecycle actions, the programme is not closing the gap. True closure means the identity graph is reconciled end to end.

Q: Why do manual provisioning workflows create identity governance risk?

A: Manual workflows create inconsistent approval paths, delayed access removal, and poor audit evidence. They also make it harder to apply the same entitlement policy across joiners, movers, and leavers. In SaaS-heavy environments, the risk is not just inefficiency. It is that access decisions become too variable to govern reliably.

Q: How does identity governance change for AI systems with runtime access?

A: Identity governance for AI systems with runtime access must extend from static approval to continuous control of the agent's delegated scope, tool permissions and accountable human owner. runtime lifecycle governance covers issuance, session boundaries, logging, human review, and rapid revocation when the agent's behaviour, data access or purpose changes. Treat prompts, tools and tokens as governed entitlements.

Q: What is the best way to govern disconnected identity systems?

A: The best way to govern disconnected identity systems is to federate their records into a single operational view for identity governance, access reviews and privileged access oversight. That view should reconcile users, service accounts, groups, entitlements and certificates across each source system, while preserving the authoritative source for each attribute. Without reconciliation, orphaned access and duplicate identities persist.

Technical breakdown

Application coverage gaps are a governance control problem

Coverage gaps appear when an identity programme cannot reliably see, govern, or certify every application that issues or consumes access. The result is not just incomplete inventory. It is a broken governance loop where entitlements exist outside review cycles, approvals happen in side channels, and auditors keep finding shadow access paths. In practice, this usually reflects weak connectors, inconsistent data ownership, or an access model that was never designed for scale. See the NHI Lifecycle Management Guide.

Practical implication: Measure application coverage as a control objective, not an implementation metric, and track every uncovered app as unmanaged governance exposure.

Manual provisioning backlogs create identity drift

Manual provisioning is slow because each request depends on a person, a queue, and a series of repeated decisions. Over time, the delay creates identity drift, where accounts, roles, and entitlements no longer match current business need. In IGA terms, that drift becomes privilege creep, stale access, and delayed offboarding. The problem is not only speed. It is that manual handling makes the access state less trustworthy with every handoff. See the Joiner-Mover-Leaver (JML) Guide.

Practical implication: Treat provisioning backlog as risk accumulation and prioritise the highest-impact joins, moves, and leavers first.

Agentic AI raises the bar for lifecycle governance

Agentic AI changes the governance conversation because the identity subject can create, use, and retire access in ways that are harder to predict than a human workflow. That does not make lifecycle governance obsolete. It makes identity registration, delegation, and retirement more important, because the system can act at runtime rather than wait for a ticket. The same governance discipline now has to cover dynamic access consumption as well as static entitlement assignment. See the Agentic AI Identity Guide.

Practical implication: Extend lifecycle governance to any AI system that can initiate actions or request access at runtime, and require ownership before deployment.

NHI Mgmt Group analysis

Identity governance is still being judged on fundamentals, not vision. READI's Identiverse 2026 commentary shows that practitioners are still asking for coverage, provisioning, and system integration before they ask for more advanced automation. That is a sign the market has not finished converting governance from a set of manual tasks into an operational control plane. The implication is that roadmaps built around novelty will continue to miss the real buying trigger.

Application coverage remains the most consequential blind spot in IGA. A programme that governs only a fraction of the application estate is structurally incomplete, even if the workflows around that fraction are polished. The issue is not configuration quality inside the toolset; it is the gap between what the business uses and what the governance programme can actually see. Practitioners should treat coverage as the first test of whether IGA has operational authority.

Manual access handling is a scaling ceiling, not an implementation detail. Once provisioning, reconciliation, and exception handling depend on repeated human intervention, the programme becomes slower exactly when the environment becomes more dynamic. That creates a queue of stale access, delayed reviews, and unresolved exceptions that auditors notice every year. The implication is simple: governance that cannot keep pace with access change will always trail the business it is supposed to control.

Identity governance for autonomous systems will inherit the same old weaknesses unless the operating model changes. Access review processes assume access persists long enough to be reviewed; autonomous actors can create, use, and discard privileges before a traditional review cycle has even started. That assumption was designed for stable entitlement states. It fails when the actor can act at runtime and change its own access needs mid-session. The implication is that governance teams need to rethink what counts as an auditable state before they can safely extend IGA to autonomous systems.

Continuous evidence, not one-time configuration, is becoming the governance threshold. The recurring pattern across the conference floor was not a lack of ideas, but a lack of closed loops between identity data, provisioning, and certification. That makes evidence freshness as important as policy design. Practitioners who can continuously reconcile identity state will be better positioned to absorb both current IGA debt and future agentic pressure.

From our research library:

• 67% of organisations still rely heavily on static credentials despite the risks they pose to agentic AI deployments, according to the 2026 Infrastructure Identity Survey.
• Only 13% of organisations feel extremely prepared for the reality of agentic AI despite the majority racing toward autonomous adoption, according to the 2026 Infrastructure Identity Survey.
• Read next: NHI Lifecycle Management Guide

What this signals

Identity governance is moving from task automation to coverage management. Teams that treat IGA as a workflow project will miss the larger shift: the control problem is now whether every meaningful application, account, and delegated identity is inside the governance perimeter. READI's Identiverse recap is a reminder that coverage debt, not feature debt, is what still blocks mature programmes.

Runtime access will expose the limits of review-based governance. Access reviews assume there is a stable entitlement state to certify. As more systems behave like agents or act through delegated identities, that state becomes shorter-lived and less observable, which pushes governance toward issuance-time and ownership-time controls instead of post-hoc certification.

Coverage debt: the gap between the identity estate a programme believes it governs and the estate it can actually see, certify, and offboard. That gap is now the practical measure of whether identity governance can absorb both human backlog and autonomous access patterns.

For practitioners

• Audit application coverage first Build a definitive list of applications that issue or consume access, then compare it to the set currently governed by your IGA process.
• Remove manual steps from high-volume provisioning Prioritise the joins, moves, and leavers that generate the most queue time, exception handling, and stale access.
• Tie every access pathway to an owner Require named business and technical owners for each application, integration, and non-human access path before it enters the governance model.
• Extend lifecycle control to AI systems Treat any runtime system that can request or use credentials as an identity subject and define how it is registered, reviewed, and retired.

Key takeaways

• Identity governance programmes still fail most often at application coverage, provisioning throughput, and system integration rather than at high-level strategy.
• The article reflects a recurring enterprise pattern: advanced messaging often lands before the basics are operationally complete.
• For practitioners, the immediate priority is to make identity coverage, ownership, and lifecycle control measurable across both human and machine access.

Related NHI Mgmt Group resources: NHI Lifecycle Management Guide · Joiner-Mover-Leaver (JML) Guide · Agentic AI Identity Guide · Top 10 Agentic AI Identity Issues

Key terms

• Application Coverage: Application coverage is the share of an application portfolio that has been tested within a defined period. It helps security leaders see whether testing reaches only a small set of crown jewels or extends across the broader estate, including lower-tier apps and APIs that can still create meaningful risk.
• Identity Governance and Administration (IGA): A framework of policies, processes, and technology to manage and govern digital identities and their access rights. Increasingly extended to cover non-human identities alongside human users.
• Provisioning Backlog: A provisioning backlog is the queue of identity and access changes waiting to be applied to target systems. It usually contains account creation, role assignment, group membership and entitlement updates that have been approved but not yet executed, which can delay access delivery and leave joiner-mover-leaver changes incomplete.
• Delegated Identity: Delegated identity is when one actor acts on behalf of another with explicit permission and bounded authority. In AI-assisted commerce, it requires clear consent, limited scope, and traceable records so the retailer can distinguish authorised delegation from unauthorised automation.

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 responsible for identity security strategy or NHI governance in your organisation, it is worth exploring.