This is a one-turn conversation: you ask for the prompt applied to the document, then write the AI's response yourself, with its predictions. Continue it as a conversation →
You write both sides. Predictions show what the model would say in each role.