Example: Clean an AI-Written Reply Before Sending

Common tasks

• Clean an AI-written customer reply before sending.
• Rewrite a professor, recruiter, client, or teammate response so it feels accountable.
• Remove English and Chinese assistant phrases from a public comment or email.
• Check whether the reply promises too much or avoids the real question.
• Turn a vague AI response into a concise answer with a next action.

Recommended workflow

1. Find and remove wrapper lines that talk about the answer instead of answering.
2. Replace vague praise or apology with the concrete issue, decision, or next step.
3. Keep facts, dates, limits, and uncertainty that matter to the recipient.
4. Check whether the message answers who will do what by when.
5. Remove overpromises such as guaranteed, always, perfect, 一定, 完美解决, or 绝对不会 unless they are truly supported.
6. Read the final message as the recipient and cut anything that only exists because an AI assistant wrote it.

When not to use it

• Do not use cleanup to disguise authorship or evade disclosure rules.
• Do not delete necessary caveats, risk notes, or source limitations.
• Do not send a polished reply that still avoids the recipient's main question.
• Do not leave templated closings when the conversation needs a concrete owner or deadline.
• Do not publish cleaned text before verifying factual claims.