I built an n8n workflow that acts as a first-line AI support agent. It receives customer tickets via webhook, tries to resolve them using your knowledge base and order history, and either sends an auto-reply or escalates to the right human team — all automatically.
How it works
-
Customer submits a ticket via webhook (works with any website form, chat widget, or API call)
-
n8n parses the ticket and runs two lookups in parallel: knowledge base search + order history
-
GPT-4o-mini analyzes everything — classifies the category (billing, technical, shipping, etc.), assigns priority, detects sentiment, and writes a response
-
If AI can resolve it → sends a branded HTML email reply to the customer, logs to Google Sheets, posts to Slack
-
If it can’t → escalates to the right Slack channel (#billing-support, #tech-support, #security-urgent) with the AI’s draft response and internal notes so the human agent can respond faster
Smart escalation
The AI doesn’t just dump everything into one queue. It routes based on the issue:
-
Billing/refund issues → always escalated to #billing-support
-
Account security → immediately escalated to #security-urgent
-
Technical issues beyond the KB → #tech-support
-
Everything else → #support-escalations
Each escalation includes the AI’s draft response, confidence score, and internal notes.
What you need
-
n8n (self-hosted or cloud)
-
OpenAI API key
-
SMTP credentials
-
Google Sheets + Slack bot token
-
Optionally: a knowledge base API and order/CRM API
Cost
With GPT-4o-mini, it’s about $0.90/month for 50 tickets/day. Very cheap to run.
Get it
Full workflow JSON, setup instructions, and detailed docs: https://github.com/ivansiyanko/n8n-support-ai
Just import the JSON into n8n, set up your credentials, and activate. Happy to answer any questions or hear suggestions for improvements!