Triaging comments and DMs and drafting on-brand replies
A busy account collects hundreds of comments and DMs a day — questions, praise, complaints, and trolls mixed together — and keeping up in the brand's voice is relentless. Only 17% of social pros currently use AI to respond to comments and messages (Metricool), the lowest-adoption use on the list, precisely because it is sensitive. Used carefully, AI can triage the queue and draft replies for a human to approve.
You are a community manager for {{brand_name}}, whose voice is: {{brand_voice}}. Below are incoming comments and DMs. For each one: - Classify it: question / praise / complaint / troll / sales lead / crisis-risk. - Draft a short reply in our voice, EXCEPT for anything legal, medical, safety, or crisis-related — for those, write [ESCALATE] and a one-line reason instead of a reply. Known policies and FAQs you may use: {{policies}} Comments and DMs: {{comments}} Rules: - Never promise refunds, discounts, delivery timelines, or policies I have not given you. If a good reply needs a fact or policy I did not provide, write [NEED POLICY: what you need] instead of guessing. - Do not argue with or feed trolls. For hostile comments, suggest either a brief neutral reply or "ignore," and never match their tone. - Keep replies short, warm, and human. No corporate filler. - Never invent an apology for something that didn't happen or admit fault you can't verify.
Fill in your details and the prompt updates live — then copy.
1) Question. Reply: "Great question! Our rain jacket runs true to size — if you're between sizes, most folks size up for layering. Happy to help you pick." 2) Complaint. Reply: "So sorry for the wait, [name]. That's not the experience we want. Can you DM your order number and we'll track it down today? [NEED POLICY: current shipping SLA so we don't overpromise a date]" 3) Praise. Reply: "This made our day — thank you! Get out there this weekend." 4) Troll / crisis-risk. [ESCALATE] Public accusation of fraud; suggest a calm one-line reply plus a manager review, don't argue in the thread.
The full workflow
- Paste your current policies and FAQs so the AI has the real facts to reply with
- Run the prompt to classify and draft, but never let it post replies automatically
- Resolve every [NEED POLICY] flag and route every [ESCALATE] item to the right human
- Personalize approved drafts — add the person's name and any account-specific detail yourself
- Log recurring questions so you can update the FAQ and the policy list the AI works from
Watch out for
Never paste a customer's personal information — full names, order numbers, addresses, DM contents — into a consumer AI tool. Strip identifiers before drafting, and keep customer records in your own support system.
A wrong or off-brand reply is public and permanent. Complaints, safety issues, legal threats, and potential PR crises must go to a human, not an auto-reply — the FTC also holds brands responsible for misleading claims made on their accounts, including in replies.
Where this comes from
Every use case on this site is grounded in real reports from working social media managers — not invented by us.