Capstone Part 3: AI Support Brain

In Part 3, we teach Nexus-One how to "Read." We will monitor an inbox for customer questions, use AI to summarize them, and send a clean alert to the support team.

1. The Gmail Trigger (Review Module 7)

Add a Gmail node.

• Operation: Watch for new emails.
• Search Query: subject:(Support OR Question OR Help).

---

2. The AI Chain (Review Module 10)

Add an OpenAI (or Anthropic) Chain node.

• Prompt: *"You are a support assistant. Below is an email body. Please:
  1. Summarize the user's problem in 5 words.
  2. Determine the sentiment (Happy, Neutral, Angry).
  3. Suggest the best department to handle this (Tech, Billing, Sales)."*

---

3. The Extraction Step (Review Module 10, Lesson 4)

Ensure your AI node is returning JSON.

• Add an Output Parser or simply use a Set node to map the AI's response to fields like summary, sentiment, and dept.

---

4. The Slack Notification

Add a Slack node.

• Message: 🔴 NEW SUPPORT TICKET Problem: {{ $json.summary }} Sentiment: {{ $json.sentiment }} Assigned: {{ $json.dept }}

---

Exercise: The Support Test

1. Send yourself a "Test" email with the subject "Help! My login is broken."
2. Run your n8n workflow. Does the AI correctly identify "Login issues" as the summary?
3. What department did it assign for an "Email Password Reset" request?
4. How would you "Pin Data" (Module 3) to test your Slack message without waiting for a new email?
5. Research: How would you add a link to the original Gmail message in your Slack alert?

---

Summary

You have given Nexus-One a "Brain." By delegating the reading and categorization of support tickets to AI, you ensure that your team only ever looks at structured, meaningful summaries instead of a messy, overflowing inbox.

Next Lesson: Part 4: Error Handling and Health Monitoring.