Gmail Triage to Airtable + OpenRouter Auto Draft Replies (n8n)

Turn Gmail support emails into Airtable records and AI-written drafts—automatically (with human review when needed)

This n8n workflow monitors your Gmail inbox, logs new support messages into Airtable, and uses OpenRouter (with an OpenAI fallback) to categorize, summarize, and generate Gmail draft replies—without sending them. If the AI confidence is low or review is required, it alerts your team for approval.

What this workflow does

• Runs continuously: triggers every minute to fetch inbox messages, and also runs every 15 minutes to check for stuck Airtable records.
• Pre-processes messages: normalizes email bodies (converts HTML to text when needed) and skips newsletters, auto-replies, bulk mail, and messages from configured sender patterns.
• Avoids duplicates: processes emails one at a time and searches Airtable by Gmail Message ID to skip messages already logged.
• Creates an Airtable intake: inserts/updates a record with status Processing, then sends email content to an OpenRouter chat model.
• Generates structured output: expects JSON containing category, confidence, summary, and a draft reply. Falls back to OpenAI if needed.
• Drafts in Gmail: validates the AI output, updates Airtable with results, and creates a Gmail draft in the original thread (no auto-sending).
• Escalates for review: sends a Gmail review alert when confidence is below a threshold, the category requires review, or the AI requests human review, then marks the record as alerted.
• Handles failures: upserts/updates Airtable with Error and emails troubleshooting details on processing failures.

Use cases

• Support teams who want AI-assisted replies while keeping a human approval workflow.
• SaaS operators tracking incoming requests in Airtable and standardizing response categories.
• Automation engineers building reliable Gmail + Airtable + LLM triage pipelines in n8n.

Technical details

• n8n nodes: Gmail, Airtable, if, set, code, no op
• LLM integration: OpenRouter chat model with OpenAI fallback
• Core logic: HTML-to-text normalization, duplicate prevention via Gmail Message ID, Airtable status tracking, Gmail draft creation, alerting, and error upserts