Your First AI Workflow with n8n

Overview

n8n is a powerful workflow automation tool that lets you connect AI models like OpenAI’s GPT to hundreds of apps without code. Build smart automations visually with a drag-and-drop interface.

What you’ll build: An AI-powered content moderation workflow that automatically analyzes user submissions and routes them based on sentiment.

Use cases: Content moderation, email auto-responses, data enrichment, customer support triage.

Time: 30 minutes

Prerequisites

n8n account (free at n8n.cloud) or self-hosted installation
OpenAI API key
Basic understanding of automation concepts

Setting Up n8n

Option 1: Cloud (Easiest)

Go to n8n.cloud
Create free account
Access your workflow editor

Option 2: Self-Hosted (Docker)

docker run -it --rm \
  --name n8n \
  -p 5678:5678 \
  -v ~/.n8n:/home/node/.n8n \
  n8nio/n8n

Access at http://localhost:5678

Connecting OpenAI to n8n

Step 1: Add OpenAI Credentials

Click Settings > Credentials
Click Add Credential
Search “OpenAI”
Enter your API key
Test connection
Save

Step 2: Create Your First Workflow

Workflow: AI Content Analyzer

Components:

Webhook - Receives content submissions
OpenAI - Analyzes sentiment
IF - Routes based on result
HTTP Request - Sends to appropriate endpoint

Building the Workflow

1. Add Webhook Trigger

Drag “Webhook” node to canvas
Set path: /analyze-content
Method: POST
Click “Listen for test event”

2. Add OpenAI Node

Search “OpenAI” and add node
Connect to webhook
Select credential
Operation: “Message a Model”
Model: gpt-4o-mini
Prompt:

Analyze this content for sentiment and toxicity:

Content: {{$json.body.content}}

Return JSON with format:
{
  "sentiment": "positive|negative|neutral",
  "toxic": boolean,
  "confidence": 0.0-1.0,
  "reason": "brief explanation"
}

3. Add IF Node

Connect to OpenAI node
Condition: {{$json.toxic}} === true
True branch: Flag for review
False branch: Approve

4. Add Action Nodes

For True (toxic content):

HTTP Request node
URL: Your moderation endpoint
Method: POST
Body: {{$json}}

For False (clean content):

HTTP Request node
URL: Your approval endpoint
Method: POST
Body: {{$json}}

Testing the Workflow

Click Execute Workflow
Send test data via webhook:

curl -X POST http://localhost:5678/webhook/analyze-content \
  -H "Content-Type: application/json" \
  -d '{"content": "This is a great product!"}'

Check execution log
Verify correct routing

Common n8n + AI Patterns

Pattern 1: Email Auto-Responder

Gmail Trigger → OpenAI (Draft Response) → IF (Confident?) → Gmail (Send)

Pattern 2: Data Enrichment

Database Trigger → OpenAI (Analyze) → Update Record → Slack Notify

Pattern 3: Content Generation

Scheduler → OpenAI (Generate) → Format → Post to CMS

Best Practices

1. Error Handling

Add “Error Trigger” node:

Catches workflow failures
Logs to monitoring system
Sends alerts

2. Rate Limiting

Use “Wait” node between API calls to avoid rate limits.

3. Cost Control

Monitor OpenAI usage:

Set max tokens
Use cheaper models for simple tasks
Cache frequent requests

4. Testing

Test with sample data first
Use n8n’s test execution mode
Monitor costs during testing

Example Workflows

Email Sentiment Analysis

Gmail Trigger
  ↓
OpenAI (Analyze Sentiment)
  ↓
IF (Urgent + Negative?)
  ↓ Yes          ↓ No
Slack Alert   Label Email

Twitter Trigger (New Mentions)
  ↓
OpenAI (Classify Intent)
  ↓
Switch (Intent Type)
  ↓ Question    ↓ Complaint    ↓ Praise
  OpenAI Reply  Alert Team    Thank User

Troubleshooting

Issue: OpenAI timeout

Increase timeout in HTTP Request settings
Use shorter prompts
Check API status

Issue: Webhook not receiving data

Verify webhook URL is correct
Check firewall settings
Test with curl first

Issue: High costs

Use gpt-4o-mini instead of gpt-4o
Add caching layer
Limit workflow triggers

Next Steps

Learn more:

Advanced workflows:

Multi-step AI reasoning chains
Custom function nodes
Sub-workflows for reusability

Found an issue? Open an issue!