A company is deploying a generative AI chatbot for customer support. They want to define the assistant's persona, restrict it to only answer questions about their products, and instruct it to refuse off-topic requests — all before any user types a message. Which configuration should they use to establish these instructions?