Auto-label Gmail with Ollama AI and draft smart replies

Who is this for

This workflow is built for professionals, founders, freelancers, and anyone drowning in email who wants to automatically triage their inbox using AI — sorting emails into categories, applying Gmail labels, and generating reply drafts for important messages, all running locally with Ollama (zero paid API costs).

What this workflow does

It watches your Gmail inbox for new unread emails, uses a local AI model (Ollama) to classify each email into one of six categories (Urgent, Action Required, Follow-up, Newsletter, Automated, or Spam/Promotional), automatically applies the matching Gmail label, generates reply drafts for high-priority emails, and logs every processed email to Google Sheets for analytics.

How it works

• Gmail Trigger watches your inbox for new unread emails at a configurable interval (default: every 30 minutes).
• Extract Email Data parses the sender, subject, body, attachments, and metadata into clean fields — handles all Gmail node output formats.
• AI Classifier (Ollama, running locally) analyzes the email content and returns a structured classification with category, priority, confidence score, summary, suggested action, sentiment, and a reply draft for important messages.
• Extract Classification robustly parses the AI response (handles markdown fences, nested objects, and edge cases) and merges it with email metadata.
• Smart Router sends each email down one of six dedicated paths based on its classification.
• Gmail Label nodes automatically apply the correct label to each email in your inbox.
• Reply Draft nodes create Gmail drafts for Urgent and Action Required emails — ready for you to review, edit, and send from your Drafts folder.
• Log Entry tracks every processed email in Google Sheets with category, priority, confidence, summary, and timestamp for analytics and review.

Setup steps

• Gmail OAuth — Connect your Google account. Grant read, modify, and compose permissions.
• Create Gmail Labels — In Gmail, manually create these labels (the workflow applies them automatically): 🔴 Urgent 📋 Action Required 💬 Follow-up 📰 Newsletter 🤖 Automated 🚫 Spam-Promo
• Get Label IDs — Use the Gmail API or n8n's Gmail node (operation: "Get Many Labels") to find each label's ID. Update the labelIds in each Gmail Label node with your actual IDs.
• Ollama — Ensure Ollama is running locally (http://localhost:11434) with your preferred model pulled (e.g., ollama pull mistral). Change the model name in the Ollama Chat Model node if needed.
• Google Sheets (optional) — Connect your Google Sheets OAuth credential and set a spreadsheet ID in the logging node. Create a sheet with columns: Date, From, Subject, Category, Priority, Confidence, Reason, Summary, Key Action, Has Reply Draft, Sentiment, Auto Archive, Gmail ID, Processed At.
• Test — Send yourself a few test emails (urgent request, newsletter, promotional) and run the workflow manually to verify classification accuracy.

Requirements

• Gmail account with OAuth credentials (read, modify, compose permissions)
• Ollama running locally with a pulled model (mistral, llama3, gemma2, etc.)
• Google Sheets account with OAuth credentials (optional, for logging)
• n8n instance (self-hosted recommended for Ollama connectivity)

How to customize

• Swap mistral for llama3, gemma2, deepseek, or any Ollama-compatible model.
• Add more categories by editing the AI system prompt and adding outputs to the Switch node.
• Change the reply draft tone (formal, casual, friendly) in the AI system prompt.
• Add Slack or Telegram notifications for urgent emails by branching from the Label nodes.
• Add auto-archive for newsletters and spam by adding a Gmail "Remove Labels" node (remove INBOX label).
• Decrease the polling interval for near-real-time processing.
• Add sender whitelist/blacklist logic in a Code node before the AI classifier.