As a final step before sending translations to clients, we can run them through an NLP product specifically focused on scoring translation quality called Modelfront (https://modelfront.com). A high-risk score would trigger automatic QA and notification to client about delays.