Handle incoming emails faster with AI and a built in approval step. The system reads each message, finds answers from your company documents, drafts a short reply, and waits for a human to approve or suggest edits. It is ideal for shared support inboxes and busy teams.

First, documents in Google Drive are loaded, split into small parts, and embedded with OpenAI, then saved in a Qdrant collection. When a new email arrives through IMAP, the text is converted to clean markdown and summarized. An AI agent uses the Qdrant knowledge base to write a clear answer. A teammate gets the draft by Gmail and can reply with approval or feedback. A classifier checks the response. If approved, the message is sent by SMTP to the sender. If not, an Email Reviewer rewrites the text using the feedback and routes it again.

Replace QDRANTURL and COLLECTION, and use a Gmail address for the send and wait step. Expect faster replies, more consistent answers, and less manual typing. Great for customer questions, sales inquiries, and vendor emails where accuracy and tone matter.

• IMAP email intake to capture new messages in real time
• HTML to markdown conversion for cleaner AI input
• Email summary with a short 100 word overview
• Google Drive file load, split, and OpenAI embeddings
• Qdrant vector store used as a knowledge tool for answers
• Draft reply creation with a concise business tone
• Gmail send and wait for human approval via a simple reply
• Text classifier decides Approved or Changes needed
• Email Reviewer rewrites using human feedback
• Final delivery via SMTP to the original sender

• Reduce first response time from hours to minutes
• Automate up to 70 percent of routine email replies
• Improve answer accuracy by referencing approved docs
• Handle up to 5 times more emails with the same team
• Connect Gmail, Google Drive and Qdrant in one flow
• Keep human review in the loop for quality control

1. Import the template into n8n: Create a new workflow in n8n > Click the three dots menu > Select 'Import from File' > Choose the downloaded JSON file.
2. You'll need accounts with Gmail, Google Drive, OpenAI, Qdrant and DeepSeek. See the Tools Required section above for links to create accounts with these services.
3. In the n8n credentials manager, create a Gmail OAuth2 credential. Double click the Gmail node, choose the new credential, and follow the on screen steps to allow send and wait.
4. Create a Google Drive OAuth2 credential. Open the Google Drive nodes, pick the credential, and test by listing a folder.
5. Create an OpenAI API Key in your OpenAI account, then in n8n create an OpenAI credential and paste the key. Link it to the embeddings and chat model nodes.
6. Create a DeepSeek API Key in your DeepSeek account, then in n8n create a DeepSeek credential and attach it to the summarization model node if used.
7. Set up Qdrant access. In n8n create an HTTP Header Auth or Qdrant API credential with your Qdrant URL and API key, then attach it to the HTTP Request and Qdrant nodes.
8. Configure the IMAP Email Trigger node with your inbox host, port, username, and password. Test that new emails are detected.
9. Configure the SMTP Send Email node with your outgoing server, port, username, and password. Send a test email to confirm.
10. Replace QDRANTURL and COLLECTION in the HTTP Request and Qdrant nodes. Choose the Google Drive folder to index, and run the manual steps to create and refresh the collection.
11. Open the Text Classifier node and confirm the Approved and Changes needed categories. Adjust wording if your team uses different terms.
12. Run a full test. Send an email to the IMAP inbox, check the Gmail approval message, reply Approved, and verify the sender gets the final email.
13. Troubleshoot common issues. Use a Gmail address for the send and wait step, confirm Qdrant collection exists, and check OpenAI or DeepSeek API limits if embeddings or summaries fail.