n8n Workflow: Safe AI Draft Replies for Reviews + Slack Escalation

Turn every review into a brand-safe AI reply draft—then escalate risk to Slack

This n8n workflow automatically converts Google, Play Store, Zomato, and Amazon reviews into polished reply drafts using an OpenAI-compatible model—while enforcing guardrails so the AI never posts something risky. Low-confidence or sensitive drafts are sent to your owner via Slack for approval.

What this workflow does

• Receives a review via webhook (business, platform, author, rating, and review text).
• Validates the input and uses an LLM to classify sentiment and topics.
• Drafts a reply mentioning one concrete detail from the review (so replies feel specific, not generic).
• Applies deterministic “safe reply” code guardrails that check:
  • Reply length (max words)
  • Blocked phrases (e.g., refund/compensation/legal wording)
  • Low star ratings (configurable escalation threshold)
  • The model’s own “needs attention” flag
• Logs approved drafts as ready_to_post to a sheet or CRM via webhook.
• Escalates risky drafts to Slack by posting the draft to your owner through a Slack incoming-webhook.

Use cases

• Restaurants/cafes replying to Google reviews without risking inappropriate language.
• App brands drafting Play Store responses while preventing refund or legal statements.
• E-commerce teams turning Amazon feedback into consistent, on-brand replies with human review for low ratings.
• SaaS and automation operators needing guardrailed LLM outputs for customer communications.

Technical details

• LLM (OpenAI-compatible): Works with OpenAI, Groq, OpenRouter, or Ollama.
• n8n nodes: webhook, if, code, http request, respond to webhook, and a sticky note.
• Setup: Configure baseUrl, model, blockedPhrases, maxReplyWords, and escalateAtOrBelowRating. Add Slack via incoming-webhook in Alert Owner (Slack). Send approved replies to Log Approved Reply (Sheets/Airtable/CRM webhook).