Subscribe to the Non-Human & AI Identity Journal

Automation tiers

Automation tiers are policy-defined levels that determine how far a system can act on its own before human approval is required. They are useful because they separate governance from implementation, allowing organisations to control risk without scripting every possible workflow.

Expanded Definition

Automation tiers describe how much autonomy a workflow, tool, or agent is allowed to exercise before a person must review, approve, or intervene. In practice, the term is less about the technology itself and more about governance: the same system may sit at a low tier for sensitive actions and a higher tier for routine actions. That is why automation tiers are often used alongside policy, risk classification, and change control rather than as a purely technical setting.

In cybersecurity and identity operations, the concept becomes especially important when software can make decisions that affect access, secrets, tickets, or infrastructure. A tiered model gives security teams a way to separate “can propose” from “can execute,” which is critical when an AI agent or orchestration tool has tool access. The most useful reference point is control design in NIST SP 800-53 Rev 5 Security and Privacy Controls, where human oversight, authorization, and accountability can be mapped to system behaviour.

Definitions vary across vendors when they describe autonomy, approval gating, or “copilot” behaviour, so organisations should treat automation tiers as a governance model first and a product feature second. The most common misapplication is assuming a tool is “low risk” because it is helpful, when the actual condition is that its action scope was never formally tiered against the sensitivity of the task.

Examples and Use Cases

Implementing automation tiers rigorously often introduces slower execution for sensitive workflows, requiring organisations to weigh operational speed against reduced approval risk.

  • A service desk assistant drafts password reset actions but cannot trigger the reset until a human operator approves the request.
  • An NHI workflow automatically rotates low-risk API keys, while high-impact secrets require dual approval before the change is applied.
  • An AI agent can gather logs and suggest containment steps, but it cannot isolate a host without a higher tier of authorization.
  • A cloud remediation workflow is allowed to quarantine misconfigured storage only when the policy engine confirms the asset is non-production.
  • An access review system can recommend role changes, but only identity administrators can finalize entitlement removal after checking business impact.

For teams building policy around autonomous actions, the distinction between suggestion and execution should be explicit in documentation, workflow design, and audit evidence. That is especially relevant when aligning to NIST control expectations for authorization, accountability, and system integrity, because a tier boundary is only meaningful if it is enforced consistently. In mature operations, the tiers are also tied to incident severity, so the same workflow may be fully automated for routine events and strictly supervised during heightened risk.

Why It Matters for Security Teams

Security teams need automation tiers because autonomy without boundaries creates hidden authority. When a system can act across identity, infrastructure, or incident response domains, an unclear tiering model can lead to overreach, accidental privilege escalation, or actions that are difficult to explain after the fact. This is particularly relevant for agentic AI and NHI-adjacent workflows, where tool access can turn a recommendation engine into an execution path if approvals are not carefully enforced.

The governance value of tiers is that they make escalation rules auditable. Instead of debating whether a workflow should be “automated” in the abstract, teams can decide which actions require confirmation, which need policy checks, and which are allowed to run continuously. That aligns well with control-focused security programs and with the intent of NIST SP 800-53 Rev 5 Security and Privacy Controls for accountability and least privilege.

Organisations typically encounter the real cost of weak automation tiering only after an autonomous action removes access, modifies production systems, or triggers an incident response mistake, at which point the tier model becomes operationally unavoidable to fix.

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 CSF 2.0, NIST SP 800-53 Rev 5 and NIST AI RMF set the governance and control requirements practitioners need to meet.

Framework Control / Reference Relevance
NIST CSF 2.0 PR.AC-4 Least-privilege access aligns to tiered authorization before autonomous action.
NIST SP 800-53 Rev 5 AC-2 Account management supports governance of who can approve or trigger tiered actions.
OWASP Agentic AI Top 10 Agentic AI guidance addresses tool-use boundaries and human oversight for autonomous actions.
OWASP Non-Human Identity Top 10 NHI governance covers service identities that execute automated workflows at different trust levels.
NIST AI RMF The AI RMF emphasizes governance, mapping well to policy-defined autonomy levels.

Map automation permissions to PR.AC-4 and require approval before higher-impact actions execute.