Handle a busy inbox with faster, safer replies. Incoming emails are read, summarized, and a short response is drafted. A human review step approves the message before it goes out.

The flow watches your mailbox with IMAP. Email HTML is converted to Markdown so language models read it clearly. A DeepSeek model creates a brief summary. An OpenAI model writes a professional reply under 100 words, using company facts pulled by an agent from your vector database. The draft is emailed to a reviewer with the original message and the AI reply shown together. If approved, the system sends the reply back to the original sender using the same subject and addresses.

You only need a Gmail or Outlook inbox and API keys for OpenAI and DeepSeek. Set the approver email and adjust the prompts to match your tone. Teams use this for support tickets, sales leads, and partner questions. Expect response time to drop from 20 minutes to 2 minutes per email while keeping full human control over what gets sent.

• IMAP email watcher pulls new messages from your inbox
• Markdown node converts HTML emails to clean text for better model input
• DeepSeek chat model summarizes each email in about 100 words
• OpenAI model drafts a short professional reply under 100 words
• Agent step can look up business details from your vector database for accurate answers
• Approve Email step emails the draft to a reviewer and waits for a decision
• If node routes only approved drafts to send and stops the rest
• SMTP sending uses the original subject and addresses for correct threading
• Set node maps AI output into the final email body

• Reduce manual reply work from 20 minutes to 2 minutes per email
• Automate about 70 percent of routine responses with a human approval step
• Lower errors by showing the original message and the AI draft side by side
• Handle up to 5 times more inbound emails without adding staff
• Connect Gmail or Outlook with OpenAI and DeepSeek in one flow

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, Microsoft Outlook, OpenAI and DeepSeek. See the Tools Required section above for links to create accounts with these services.
3. In the n8n credentials manager, create an IMAP credential for your mailbox. If using Gmail, create an app password and enable IMAP. If using Outlook, use your Microsoft 365 IMAP settings. Double click the Email Trigger node, choose your credential, and test the connection.
4. Create an SMTP credential for sending. Double click the Send Email and Approve Email nodes, select the SMTP credential, and send a test message to confirm delivery.
5. Open the OpenAI node, click Create new credential, paste your OpenAI API key from the official API page, and select the gpt 4o mini model.
6. Open the OpenAI Chat Model node, create a credential for DeepSeek, add your API key from the DeepSeek account page, and select the deepseek chat model.
7. In the Email Trigger node, set the inbox folder and make sure it reads only unread emails to avoid duplicates.
8. In the Approve Email node, set the reviewer address for approvals. Confirm the message shows both the original email and the AI draft.
9. Open the Write email node and adjust the system message and word limit to match your brand tone and compliance needs.
10. Run a test: send a sample email to your inbox, confirm the workflow creates a summary and draft, check the approval email, approve it, and verify the reply is sent to the original sender.
11. Troubleshoot: if nothing triggers, check IMAP settings and app passwords. If approvals do not resume the flow, make sure the workflow is active and that the approval email is not in spam. Review the Executions list for errors.