Correct wrong AI answers from the inbox

Mark an AI answer as wrong and add or approve the correct response directly from the conversation, so future answers improve without manually editing sources. When the AI gives a wrong or incomplete answer, the teammate reviewing the conversation should be able to flag it right there in the inbox. They would select the incorrect passage, mark it as wrong, and either type the correct answer or approve a suggested correction. Behind the scenes, that correction feeds back into the AI's knowledge. The next time a customer asks the same question, the AI gives the updated answer instead of repeating the mistake. This turns every wrong answer into a learning moment, without asking anyone to leave the inbox, open a separate content editor, or remember to update a help article later. This matters because wrong answers erode customer trust quickly. If a visitor gets bad information once, they are less likely to trust the AI the next time. A fast correction loop — right from the conversation where the mistake happened — keeps quality high and makes the AI more reliable over time.