Which metric should a Generative AI Engineer monitor for a customer service LLM application that answers customer inquiries after it has been deployed in production?