How to Use n8n for AI Automation in 2026 — Beginner's Guide

n8n scored 82/100 in our testing. It's an open-source workflow automation tool that connects apps, APIs, and AI models without writing full backend code. Think Zapier but with actual code control — and a free self-hosted option.

Here's how to start building AI automations with it.

n8n vs Zapier vs Make: When to Choose n8n

Choose n8n if you need: AI model integration, self-hosting for data privacy, custom code in workflows, or complex branching logic.

Getting Started: Cloud vs Self-Hosted

Option 1: n8n Cloud (Fastest)

1. Sign up at n8n.io
2. Choose a workspace — the free trial includes full feature access
3. You're building in 2 minutes

Option 2: Self-Hosted (Free, More Control)

npx n8n

Or with Docker:

docker run -it --rm --name n8n -p 5678:5678 n8nio/n8n

Access at localhost:5678.

Self-hosted is free forever and keeps your data local — important for workflows handling sensitive information.

Understanding n8n's Interface

Every n8n workflow is a visual graph:

• Nodes: Individual steps (trigger, action, transformation)
• Connections: Arrows linking nodes — data flows along these
• Trigger node: Always the first node — starts the workflow
• Canvas: The visual editor where you build

Click + to add a node. Search by app name (Gmail, Slack, OpenAI) or function (HTTP Request, Code, Merge).

Your First AI Workflow: Email → AI Summary → Slack

This workflow reads new emails, summarizes them with AI, and posts to Slack.

Step 1: Gmail Trigger

1. Add node → search Gmail Trigger
2. Set trigger: New Email
3. Connect your Google account via OAuth
4. Set polling interval: every 5 minutes

Step 2: OpenAI Node

1. Add node → search OpenAI
2. Operation: Message a Model
3. Model: gpt-4o-mini (fast and cheap for summaries)
4. Prompt:

Summarize this email in 2-3 bullet points. Note the sender, key request or information, and any action required.

Email:
{{ $json.snippet }}

The {{ $json.snippet }} references the email text from the previous node.

Step 3: Slack Node

1. Add node → search Slack
2. Operation: Send a Message
3. Channel: your target channel
4. Message:

📧 New email from {{ $('Gmail Trigger').first().json.from }}

{{ $json.message.content }}

5. Connect Slack via OAuth

Test and Activate

1. Click Test Workflow to run with a real email
2. Check each node's output in the execution panel
3. Fix any errors (usually OAuth or missing field references)
4. Click Activate to run on the schedule

3 High-Value AI Automation Workflows

Workflow 1: AI Content Repurposing

Trigger: RSS feed (new blog post published) Step 1: HTTP Request → fetch the article content Step 2: OpenAI → generate LinkedIn post, X thread, and email newsletter excerpt Step 3: Google Sheets → log content to a tracker Step 4: Slack → notify team with the generated content for review

Workflow 2: Customer Support Triage

Trigger: Gmail/Help desk (new support email) Step 1: OpenAI → classify urgency (urgent/normal/low) and category (billing/technical/general) Step 2: Code node → route based on classification Step 3 (urgent): Slack → immediate alert to support team Step 3 (normal): Add to support queue in Notion/Airtable Step 4: Gmail → send acknowledgment email with estimated response time

Workflow 3: AI Research Assistant

Trigger: Google Sheets (new row with research topic) Step 1: HTTP Request → Perplexity API (search the web for the topic) Step 2: OpenAI → summarize findings and identify gaps Step 3: Google Docs → create a research brief document Step 4: Gmail → email the brief to the requester

The Code Node: Where n8n Gets Powerful

The Code node lets you write JavaScript to transform data between steps:

// Example: format an array of items into a readable list
const items = $input.all();

return items.map(item => ({
  json: {
    formatted: `• ${item.json.title}: ${item.json.description}`,
    timestamp: new Date().toISOString()
  }
}));

This is what separates n8n from Zapier — you can do anything with data, not just map fields.

Connecting AI Models

n8n supports these AI integrations natively:

• OpenAI: GPT-4o, GPT-4o-mini, Whisper, DALL-E
• Anthropic: Claude 3.5 Sonnet, Claude Haiku
• Google: Gemini 1.5 Pro/Flash
• Ollama: Local models (self-hosted only)
• HTTP Request: Any AI API with a REST endpoint

For complex AI chains (multi-step reasoning, tool use), use the AI Agent node — it handles tool calling and multi-step workflows automatically.

See also: n8n review → | Make review → | Best automation tools →