An agent can send payment instructions that cannot be reversed. Which control suits this action best?
Why A is the answer
Human approval required before the action executes
Human-in-the-loop effort belongs on actions that are high risk and irreversible. Logs and audits only tell you after the money has gone.
In detail
- The action cannot be reversed, so a problem must be caught before it happens.
- Human approval belongs on actions that are high risk and irreversible.
- Show the reviewer enough context to decide quickly.
Use case
An illustrative scenario, not a real incident.
An agent prepares payments to suppliers. Nothing leaves until a finance officer approves it in a queue, showing the amount, the payee and the reason.
See it in different ways
flowchart TD
A{"Can the action be reversed?"} -->|Yes| B{"Low impact?"}
A -->|No| C["Human approval required"]
B -->|Yes| D["Automatic, with logging"]
B -->|No| E["Automatic, with review sampling"]stateDiagram-v2 [*] --> AssessAction AssessAction --> Automate: low value, reversible AssessAction --> HumanApproval: high risk, irreversible Automate --> Logged HumanApproval --> Logged Logged --> [*]
sequenceDiagram participant P0 as Agent participant P1 as Approval queue participant P2 as Finance officer participant P3 as Bank P0->>P1: Proposed payment P1->>P2: Review P2->>P3: Approved, so send
Key point: Put people in front of irreversible actions.
Why the other options are wrong
- B. A disclaimer shown in the interface before the agent runs
- A disclaimer does not stop a bad payment.
- C. Detailed logging of every payment for later review
- Logs only tell you after the money has gone.
- D. A monthly sample audit of what the agent has sent
- A monthly audit finds problems far too late.