Back to workflows
Security & ITExecute within guardrailsTeam

Legacy access provisioning

Translate an approved access request into system-specific steps with segregation-of-duties checks.

What this workflow does

Translate an approved access request into system-specific steps with segregation-of-duties checks. It validates identity and permissions, simulates the bounded action, and executes only the approved change with verification and rollback evidence.

What you gain

Provisioned access with verification and expiry

What the AI agent changes

How this worked before

API and RPA automations could repeat known clicks, but they were brittle when the interface or situation changed.

What the AI agent changes

The agent combines identity, approval, segregation-of-duties, and least-privilege rules before preparing a bounded access change.

Agent trigger

A manager-approved access request passes identity verification.

Inputs

  • Process context: identity, role catalog, approvals, SoD rules, and target state
  • Approved policies, ownership, and exception rules

Agent flow

  1. 1

    Verify identity, entitlement, and the current policy version.

  2. 2

    Simulate the requested action and calculate its bounded effects.

  3. 3

    Prepare a permitted least-privilege provisioning plan with side effects and rollback conditions.

  4. 4

    Execute the approved action, verify the result, and log the change.

Human decisions

After step 3

The accountable process owner approves a permitted least-privilege provisioning plan.

Outcome

  • Provisioned access with verification and expiry
  • Evidence, exceptions, and audit trail

Guardrails

  • Block execution when identity, evidence, or rollback conditions are incomplete.
  • Require fresh approval for irreversible or high-value actions.

Risks and mitigations

An incorrect or over-scoped action can change a material record.

Use least-privilege tools, dry runs, hard limits, and post-action verification.

Sources and evidence