AI Email Dispatcher: Classify, Process, and Route Emails with Multiple Agents

This n8n template automatically classifies incoming emails (Sales, Support, Internal, Finance, Promotions) and routes them to a dedicated OpenAI LLM Agent for processing. Depending on the category, the agent will draft a reply and send it for internal or customer support emails, or generate a summary, notification, or draft reply for a human to review for Sales, Finance, and Promotions emails.

This is the ultimate workflow for managing a high volume of emails, ensuring every message gets a fast, appropriate, and structured AI-powered response or initial analysis.

How it works: The Gmail Trigger node monitors your inbox for unread emails. The Text Classifier node uses AI to read the email body and subject to categorise it as Internal, Customer Support, Sales, Promotions, or Admin/Finance. The workflow then uses a Split In Batches node and multiple If nodes to send the email data to a specific, specialised OpenAI Agent based on the classification. Each AI Agent has a custom system prompt to ensure it processes the email with the right rules, tone, and output format. Finally, the workflow either automatically replies via Gmail (for Internal and Support) or creates a draft/summary and sends a notification (for Sales, Finance, and Promotions) for human review.

Set up takes about 10–15 minutes: You will need two Gmail credentials (one for the trigger, one for actions), an OpenAI API Key connected to the five LLM Agent nodes, and Telegram or Slack credentials for sales and finance notifications. You should also create the required Gmail labels (Internal, Customer Support, Promotions, Admin/Finance, Sales Opportunity) before activating the workflow.

Easily customisable: Add or remove email categories by modifying the Text Classifier node and adding matching LLM Agents. Swap OpenAI for Gemini or Anthropic Claude. Extend actions by creating tasks in ClickUp, leads in Salesforce, or entries in a Google Sheet when a Sales Opportunity email is detected.