Requires command-view access; approval and execution are separately gated by permission to approve operational actions and permission to execute operational actions, with action-specific permissions still enforced.

Steps

  1. Open AI Command Center and select the property. Use queue, status, risk and command-type filters to find the command; Awaiting approval and Ready to execute represent different states.

  2. Choose View details. Inspect expected impact, risk, confidence, explainability, proposed changes and evidence, verification and compensation state. Follow linked case/workflow records to resolve missing context.

  3. If authorised and appropriate, choose the available approval action. When prompted, enter Why is this action safe and necessary? and choose Confirm action. Approval authorises the scoped command; it is not the execution result.

  4. For an approved executable command, choose Execute and complete its confirmation. Verify the resulting execution and Verification outcome before treating the business action as complete.

  5. If a supported recovery is needed, choose Request rollback, review the proposed compensation and confirm only the intended action. Inspect compensation/verification afterward; unavailable rollback cannot be replaced with a generic undo.

  6. Open the linked case to record follow-up or the full audit to inspect evidence. A failed/unknown execution should be investigated before issuing another command.