AI Email Support Agent with RAG & Human Approval

Unlock Hyper-Efficient Email Support with this AI Agent

This AI-driven workflow automates your email response process while keeping you in complete control. It monitors an email inbox, intelligently summarizes new messages, and then uses a Retrieval-Augmented Generation (RAG) approach to draft accurate replies. By connecting to a Qdrant vector store populated with your company's knowledge, the agent ensures responses are factual and relevant. Before any email is sent to the original sender, a draft is sent to a designated approver via Gmail. With a simple 'Yes' or 'No' click, you decide if the response goes out, perfectly blending AI speed with human oversight.

Key Features & Benefits

• AI-Powered Drafting: Leverages powerful LLMs (like GPT-4o-mini and DeepSeek) to generate professional, context-aware email replies in seconds.
• RAG Integration: Connects to a Qdrant vector store to ground responses in your actual company documents, FAQs, and knowledge base, ensuring high accuracy.
• Human-in-the-Loop Control: Includes a crucial one-click approval step via Gmail, giving you full control over outgoing communication and maintaining brand voice.
• Automated Summarization: Quickly digests long email threads to create a concise summary for the AI agent and for your quick review.
• End-to-End Automation: Handles the entire process from email receipt to final, approved response, freeing up your team for high-value tasks.

• An n8n instance (Cloud or self-hosted).
• API keys for an LLM provider (e.g., OpenAI, OpenRouter).
• A Qdrant vector store account and API key, populated with your knowledge base documents.
• An IMAP-enabled email account to monitor for incoming messages.
• A Gmail account for the approver (required for the 'Send and wait for response' feature).

1. Import the workflow into your n8n instance.
2. Configure the 'Email Trigger (IMAP)' node with your email account credentials.
3. In the 'Qdrant Vector Store' and 'Embeddings OpenAI' nodes, enter your Qdrant and OpenAI credentials. Update the collection name to match your setup.
4. Configure the LLM nodes ('OpenAI' and 'DeepSeek R1') with your respective API keys.
5. In the 'Send Draft' Gmail node, set the approver's Gmail address in the 'To' field and add your Gmail credentials.
6. Configure the final 'Send Email' SMTP node with credentials for your sending address.
7. Activate the workflow.