n8n + AI: Detect angry customers for New Emails Automatically

Every time a new customer email lands, this workflow hands it to Claude, which scores the sentiment and flags anything angry for a human. You get at-risk conversations caught before they churn without lifting a finger.

difficulty Advancedsetup 30 minresult At-risk conversations caught before they churn
  1. 1

    Trigger when a new customer email lands

    A Gmail Trigger starts the workflow the moment a new customer email lands.

  2. 2

    Hand it to the AI agent

    An AI Agent node running Claude scores the sentiment and flags anything angry for a human. The prompt is pre-written and scoped to this one job so results stay consistent.

  3. 3

    Deliver the sentiment call

    The result is sent to Slack — with the original for context.

  4. 4

    Add a safety net (optional)

    Insert an IF/human-approval step before anything is published or emailed, so you stay in control until you trust the output.

Frequently asked questions

How much does the AI part cost to run?

Pennies. Each run sends a small amount of text to the model, so most teams spend a few dollars a month even at hundreds of runs. You set the model and can cap volume in n8n, so costs never surprise you.

Can I trust the AI to get the sentiment call right?

The prompt is tightly scoped to one job and you can add a human-approval step in n8n before anything is sent or published. Start with review-mode for a week, and once you trust the output, switch it to fully automatic.

Does this need n8n Cloud or can I self-host?

Either works. The workflow is a standard n8n export, so you can run it on n8n Cloud or on your own self-hosted instance — you only connect your own app credentials.

About this recipe. Recipes on FlowRecipesHub are written for business owners, not developers, and are tested before publishing — how recipes get made. Some ingredient links are affiliate links that cost you nothing — full disclosure.