n8n WhatsApp Support Agent with AI Memory (OpenAI, Sheets)

Turn WhatsApp messages into AI replies—while keeping the full chat history in Google Sheets

This n8n workflow receives Meta WhatsApp webhook messages, stores each conversation per phone number in Google Sheets, generates short context-aware responses with OpenAI, and supports a human handoff via email when needed.

What this workflow does

• Receives WhatsApp webhooks from Meta: accepts a POST webhook event and parses the payload into phone number, message text, and message type.
• Filters out non-message events: ignores delivery/read receipts and only continues for real inbound messages.
• Loads and rebuilds conversation memory: pulls prior rows from Google Sheets for the sender and reconstructs the last N turns of history, then logs the incoming message with bot/human status.
• Routes the request:
  • If the user requests a human, it triggers a human handoff.
  • If no text is available (unsupported media), it returns an unsupported-media reply.
  • Otherwise, it generates a reply using OpenAI based on your knowledge base and the conversation history.
• Sends the reply back via Meta Graph API and appends an outgoing log entry to Google Sheets.
• Sticky human handoff: emails a teammate once with the message and recent history, and stops bot responses for that contact.

Use cases

• SaaS support teams handling inbound WhatsApp requests with AI answers and quick escalation to a human.
• Operators who need auditable conversation logs in Google Sheets per phone number.
• Automation engineers building WhatsApp-to-AI support flows with context retention.

Technical details

• Integrations: Meta WhatsApp Webhooks (Meta Graph API), OpenAI, Google Sheets, Gmail.
• n8n node logic: if, set, code, no op, switch, gmail.
• Required auth: Meta Graph API header auth using Authorization: Bearer <YOUR_META_ACCESS_TOKEN> and the correct Meta phone number ID.