AI Gmail Auto-Reply Workflow with AWS Bedrock (n8n)

Automatically classify Gmail emails and generate AI replies with AWS Bedrock (n8n)

This n8n workflow watches your incoming Gmail inbox, uses AWS Bedrock (Amazon Nova Lite) to decide whether each message should be answered or escalated, and then sends a ready-to-use reply directly in the same email thread—saving you time on repetitive support and follow-ups.

What this workflow does

• Polls Gmail every minute for new incoming messages that match your configured Gmail search filter.
• Extracts key email details—including thread ID, subject, and a snippet of the body—to build prompt context.
• Sends the email content to AWS Bedrock (Amazon Nova Lite) to generate a structured decision: ACTION: REPLY or ACTION: ESCALATE, along with a reply when allowed.
• Parses the model output into aiAction (REPLY/ESCALATE) and aiReply (ready-to-send text).
• If aiAction = REPLY, it sends the generated response back via Gmail as a reply in the same thread.
• Logs outcomes: AUTO_REPLY_SENT after a reply, or HUMAN_REVIEW_REQUIRED when escalation is selected.

Use cases

• Auto-respond to inbound inquiries that match a specific Gmail query (e.g., support requests in a shared inbox).
• Handle first-line email communication while escalating complex cases for human review.
• SaaS operations teams that want consistent, policy-driven reply formatting with clear escalation rules.

Technical details

• Trigger/Integration: Gmail trigger (with configurable search query filter)
• AI layer: AWS Bedrock using amazon.nova-lite-v1:0 (ensure the model is enabled in your AWS region)
• Logic/nodes: if, set, code, gmail, and supporting nodes (including sticky note for guidance)
• Setup requirements: Gmail OAuth2 credential with permissions to read and send replies; AWS credential with permission to invoke AWS Bedrock.