Intelligent Support Triage & Auto-Response Engine with Jotform, Airtable, Gemini

Description

AI-Driven Support Triage and Auto-Response Engine

This n8n workflow establishes a sophisticated, multi-stage support system. It automatically validates user identity, analyzes ticket severity and sentiment, and attempts to resolve lower-priority issues using an AI agent connected to a knowledge base (Pinecone). It ensures high-priority or complex issues are immediately escalated to the team via Slack.

Phase 1: Intake and User Validation

The workflow triggers whenever a new support ticket is submitted, ensuring that only registered users receive automated assistance.

Jotform Trigger: The process begins when a customer submits a support request via your designated Jotform.
Airtable User Check: The workflow immediately searches your Airtable "Users" table to verify if the provided email matches an existing account.
Validation Logic (If Node): * If User Found: The workflow proceeds to Phase 2 for analysis.
If User Not Found: It sends an automated Gmail message asking the user for the correct account email and updates the Airtable ticket status to "Need User Input"

Phase 2: AI Triage and Severity Analysis

Once validated, the issue is analyzed by a Senior Support AI to determine how it should be handled.

AI Agent (Triage): Powered by Google Gemini, this agent analyzes the "Issue Type" and "Description." It uses a Structured Output Parser to categorize the ticket by:
Severity: High, Medium, or Low.
Sentiment: Frustrated or Normal.
Airtable Logging: The analysis (Severity/Sentiment) is immediately appended to the support record in Airtable.
Escalation Path (If High Severity): If the issue is flagged as "High," the workflow bypasses automated response attempts and triggers a Slack alert to the team for immediate manual intervention.

Phase 3: Automated Resolution and Escalation

For non-critical issues, the system attempts to solve the problem using your internal documentation.

AI Agent (Resolution): This agent is equipped with a Pinecone Vector Store (containing your Support FAQs) and Gmail Tools.
Vector Search: The agent searches the FAQ database for a solution.
Success (>90% confidence): The agent automatically drafts and sends a polite solution via Gmail and updates the Airtable status to "Responded."
Failure (No solution found): The agent returns "MANUAL_REVIEW_REQUIRED," which triggers a Slack notification to the support channel, including deep links to the Airtable record and specific user details (Plan type, Start date).

Phase 4: Verification Loop (Email Reply Detection)

The workflow also includes a secondary logic branch to handle incoming replies from users who were previously asked for identity verification.

Gmail Trigger (Polling): Monitors for new unread replies from users with correct emails.
Verification Check: Searches Airtable for tickets currently in "Require User Input" status.
Status Restoration: Once the user replies, it notifies the team via Slack that a ticket has been "Verified" and moves the status back to "Pending" for regular processing.

Requirements

1. Automation & AI Platforms
n8n Instance:** Cloud or self-hosted.
Google Gemini API:** Required for the Triage and Resolution AI agents.
Pinecone:** A vector database containing your FAQ embeddings for AI retrieval.

2. Service Credentials
Airtable Personal Access Token:** To search users and update support entries.
Gmail OAuth2:** To send automated resolutions and monitor for user replies.
Slack OAuth2:** To send real-time alerts and escalation blocks to the team.
Jotform API Key:** To trigger the workflow upon form submission.

3. External Configurations
Airtable Schema:** Two tables are required:
Users: Containing Email, First Name, Last Name, and Current Plan.
Support Entries: Containing Submission ID, Status, Severity, Sentiment, and Message id.
Pinecone Index:** An index named support-faqs populated with relevant documentation embeddings.