Why do traditional identity governance tools struggle when organisations add cloud platforms and AI agents?

Traditional identity governance tools struggle because they depend on fragmented connectors, manual mapping, and periodic review cycles that cannot keep up with changing entitlements and schemas. As identity sprawl increases, access decisions become harder to validate in real time. That creates delays, coverage gaps, and a weaker ability to prove who has access and why.

Why This Matters for Security Teams

Traditional identity governance was built for bounded systems with relatively stable roles, approvals, and review cadences. Cloud platforms and AI agents break those assumptions. Cloud entitlements shift as infrastructure is recreated, while agents can request tools, chain actions, and change their own operating context in ways that static role models cannot express. The result is not just more identities, but identities whose access needs are runtime-dependent and hard to predict.

This is why manual certification and connector-led visibility degrade so quickly. Current guidance suggests that identity controls for autonomous workloads need to move from periodic validation to continuous evaluation, especially where machine identities can trigger downstream actions. NHIMG research on the Ultimate Guide to NHIs shows that 97% of NHIs carry excessive privileges and only 5.7% of organisations have full visibility into their service accounts, which is consistent with the operational blind spots many teams face when cloud and agentic workloads scale together. In practice, many security teams encounter privilege sprawl only after an agent or service account has already expanded its reach beyond the original approval scope.

How It Works in Practice

The practical fix is not to force cloud apps and AI agents into a human-centric governance cycle. Instead, teams should treat workload identity, task context, and policy evaluation as the decision inputs. For cloud, that means capturing ephemeral entitlements, infrastructure-as-code drift, and service-to-service access in near real time. For agents, it means authorising the action the system is trying to take, not just assigning it a broad role once and hoping the role remains accurate.

That approach aligns with emerging guidance from the NIST AI Risk Management Framework and the OWASP Agentic AI Top 10, both of which emphasise runtime risk evaluation and stronger controls around autonomous behaviour. In agentic environments, best practice is evolving toward intent-based authorisation, just-in-time credential issuance, and short-lived tokens that expire when the task ends. Workload identity primitives such as SPIFFE, SPIRE, or OIDC-backed service identities matter because they prove what the workload is, not just what secret it possesses.

• Use short TTLs for secrets and tokens so access ends with the task, not the quarter.
• Map cloud and agent permissions to runtime context, such as request source, tool usage, and environment state.
• Prefer policy-as-code evaluated at request time over static approval matrices.
• Separate human access reviews from machine access attestations, because the evidence model is different.

NHIMG’s OWASP NHI Top 10 reinforces the same operational point: static credentials and broad standing access are poor fits for systems that can adapt, pivot, and call tools autonomously. These controls tend to break down when agents operate across multiple cloud accounts and SaaS tools because entitlements change faster than review cycles can validate them.

Common Variations and Edge Cases

Tighter cloud and agent identity controls often increase operational overhead, requiring organisations to balance faster automation against stronger containment. Not every environment can adopt the same pattern immediately, especially where legacy apps, long-lived service accounts, or vendor-managed integrations still depend on static secrets. In those cases, current guidance suggests reducing blast radius first, then phasing in short-lived credentials and runtime policy checks where integration points allow it.

There is also no universal standard for agent governance yet. Some teams use coarse role definitions for initial rollout, while others move directly to task-scoped authorisation and ephemeral delegation. The right choice depends on how much autonomy the agent has, how sensitive the target environment is, and whether downstream actions can be reversed. NHIMG’s Top 10 NHI Issues is useful here because it highlights how over-privilege, poor rotation, and weak offboarding compound one another across cloud and non-human identities.

For higher-risk deployments, teams should also align agent governance with the CSA MAESTRO agentic AI threat modeling framework and continue to monitor incident patterns through the 52 NHI Breaches Analysis. The common failure mode is assuming a connector inventory equals governance maturity, when the real control question is whether access can be explained, constrained, and revoked at the moment it is used.

Related resources from NHI Mgmt Group

• Why do legacy identity tools struggle as organisations add more non-human identities and AI-driven access?
• Why do organisations struggle to maintain effective identity governance across fragmented application environments?
• How should organisations decide whether to integrate AI agents with other security platforms through protocol-based connections?
• Should organisations prioritise cloud identity governance before expanding privileged access controls across applications?