Retention Human Review Queue

Scenario

Use as the final handoff for customer success, account management, or product teams that need to review a risk before contacting a customer.

Input and output

Input is a redacted account reference, audited risk result, evidence references, allowed intervention types, and owner. Output is a review packet with priority, proposed next step, due date, human decision, and an explicit no-automatic-action state.

Execution and recovery

Run python scripts/queue.py examples/input.json. If evidence is held, add the missing redacted source or change the owner and rerun. No email, discount, refund, cancellation, or plan change is executed.

Privacy boundary and acceptance

Use placeholders and coarse values. Acceptance requires one queue item per account, preserved reasons, PENDING human decision, and automatic_action=false.