n8n AI Nodes: 10 Automations Using LLMs in Your Workflows

AI in n8n isn't just a novelty — it's a force multiplier for your automations. Whether you're using OpenAI's API or running models locally with Ollama (zero cost), n8n's AI nodes let you chain LLM calls with 400+ other integrations to build workflows that think, write, analyze, and decide.

Here are 10 real automations you can build today, ranging from zero-cost local setups to production-grade API pipelines.

The AI Node Stack in n8n

n8n has a growing family of AI nodes:

| Node | What It Does | |---|---| | OpenAI | Chat completion, embeddings, image generation | | Anthropic Claude | Claude 3.5 Sonnet for complex reasoning | | Ollama | Local models — Llama 3, Hermes 3, Mistral, Gemma | | Google Gemini | Google's multimodal models | | AI Agent | Multi-step reasoning with tool use | | Vector Store | Embeddings + retrieval for RAG pipelines | | Text Classifier | Sentiment, categorization, language detection |

The killer feature: You chain these with HTTP Request nodes, Webhooks, Databases, and 400+ other integrations. Your AI isn't a chatbot — it's an automated worker.

1. Auto-Reply to Customer Emails with Context

Nodes needed: IMAP Trigger → OpenAI → Gmail Reply

Every incoming customer email gets parsed, classified by intent (refund, question, complaint, praise), and auto-replied with a context-aware response. The AI reads the email thread, pulls relevant order data from Stripe, and crafts a response that sounds human.

IMAP Trigger → Set (extract body) → OpenAI (classify + draft reply)
→ Switch (route by intent → urgent vs standard template)
→ Gmail Reply

Cost: ~$0.002 per email with GPT-4o-mini. Or $0 with local Ollama (Hermes 3 8B).

2. RSS Feed → AI Newsletter Draft

Nodes needed: RSS Read → Ollama/OpenAI → Google Docs → Gmail

Every morning, pull articles from your favorite RSS feeds, send each through AI summarization (headline + 3 bullet points + key quote), compile into a Google Docs newsletter draft, and email yourself for review.

Why this wins: 2 hours of manual curation becomes a 5-minute review session.

Schedule Trigger → RSS Read (10 feeds) → Item Lists (flatten)
→ Ollama (summarize each) → Google Docs (compile draft) → Gmail (notify editor)

Pre-built? Yes — our RSS → AI Newsletter Draft template is ready to import.

3. Sentiment Analysis Router for Feedback

Nodes needed: Webhook → Ollama/OpenAI → Switch → Slack

Customer feedback enters via webhook (Typeform, Google Forms, Intercom). AI analyzes sentiment and routes accordingly:

• Positive → Slack #testimonials (marketing amplifies it)
• Neutral → Google Sheets (trend tracking)
• Negative → Slack #urgent-support @here (respond within 2 hours)

Webhook → Set (extract feedback text) → AI Sentiment Analysis
→ Switch:
  Positive → Slack #testimonials
  Neutral → Google Sheets
  Negative → Slack #urgent-support @here

Cost: ~$0.001 per analysis with GPT-4o-mini. $0 with Ollama.

4. Generate SEO Meta Descriptions from Blog Content

Nodes needed: Google Docs/Webhook → OpenAI → Google Sheets

Paste a blog post URL or Google Doc. AI reads the content, extracts the core topic, and generates: SEO title (under 60 chars), meta description (under 155 chars), 5 target keywords, and a suggested H2 structure.

Webhook (receive blog URL) → HTTP Request (fetch content)
→ OpenAI (generate SEO metadata) → Google Sheets (log all metadata)

5. Code Review Assistant

Nodes needed: GitHub Trigger → OpenAI/Claude → GitHub Comment

When a PR is opened, AI reviews the diff and posts an inline comment with: potential bugs, style suggestions, performance concerns, and missing tests.

GitHub Trigger (new PR) → HTTP Request (fetch diff)
→ Anthropic Claude (code review) → GitHub (post review comment)

Why Claude over GPT-4o: Claude 3.5 Sonnet consistently produces better, more actionable code reviews. It understands context better and hallucinates less about code behavior.

6. Product Description Generator for E-Commerce

Nodes needed: Airtable Trigger → OpenAI → Shopify/Airtable Update

When a new product is added to Airtable with basic specs, AI generates: a compelling product description, 5 bullet-point features, SEO title, meta description, and Instagram caption. Updates your store automatically.

Airtable Trigger (new product) → Set (format specs)
→ OpenAI (generate copy variants) → Shopify (update product)
→ Airtable (update with generated copy)

7. Meeting Transcription → Summary + Action Items

Nodes needed: Google Drive Trigger (new audio/video) → OpenAI Whisper → OpenAI Chat → Notion

Drop a meeting recording into Google Drive. Whisper transcribes it. AI extracts: meeting summary, action items (with assignees), decisions made, and follow-up questions. Creates a Notion page with everything formatted.

Google Drive Trigger (new file) → OpenAI Whisper (transcribe)
→ OpenAI Chat (extract summary + action items)
→ Notion (create meeting notes page)
→ Gmail (send summary to attendees)

8. Data Enrichment Pipeline

Nodes needed: Google Sheets Trigger → HTTP Request (Clearbit/Hunter) → OpenAI → CRM Update

New leads from a signup form are enriched with AI: company info from Clearbit, personalized email opener, lead score based on job title + company size, and recommended next action.

Google Sheets Trigger (new lead) → HTTP Request (Clearbit enrichment)
→ OpenAI (personalize messaging + score lead)
→ HubSpot (update contact + create task)

9. Local AI ChatOps for Your Team

Nodes needed: Webhook → Ollama → Discord/Slack

Deploy a Discord or Slack bot that answers team questions using a local LLM. Zero API costs, zero data leaving your network. Perfect for internal docs queries, onboarding questions, or a "rubber duck" for developers.

Webhook (receive chat message) → Ollama (generate response)
→ HTTP Response (return to chat)

Hardware needed: Any machine with 16 GB+ RAM for 8B models, 32 GB for 70B models. A used RTX 3060 12GB runs Llama 3 8B at 50+ tokens/second.

10. Automated Competitor Intelligence

Nodes needed: Schedule Trigger → HTTP Request → Ollama/OpenAI → Google Sheets → Slack

Every Monday, scrape competitor websites, pricing pages, and blog RSS feeds. AI summarizes: any price changes, new features launched, shift in messaging, new blog topics. Compiles into a Google Sheets report and sends a Slack digest.

Schedule Trigger (Monday 8 AM) → HTTP Request (scrape competitor pages)
→ Ollama (summarize changes) → Google Sheets (append weekly report)
→ Slack (send digest)

Local AI vs Cloud AI: Which Should You Use?

| Factor | Local (Ollama) | Cloud (OpenAI/Claude) | |---|---|---| | Cost | $0 (use existing hardware) | $0.002–0.015 per 1K tokens | | Privacy | Data never leaves your server | Sent to third-party servers | | Speed | 20-50 tokens/sec (GPU) | 50-100+ tokens/sec | | Quality | Good for classification, summarization | Better for creative writing, complex reasoning | | Setup | Docker run ollama/ollama | API key only | | Models | Llama 3, Mistral, Hermes 3, Gemma | GPT-4o, Claude 3.5, Gemini |

My rule of thumb: Use local AI for classification, extraction, summarization, and internal tools. Use cloud AI for customer-facing content, complex reasoning, and when you need the absolute best quality.

Setting Up Ollama for n8n (5 Minutes)

On your n8n server (or any machine on the network):

# Install Ollama
curl -fsSL https://ollama.com/install.sh | sh

# Pull a model
ollama pull llama3:8b        # Meta's Llama 3 — great all-rounder
ollama pull hermes3:8b       # Nous Research Hermes 3 — strong at instruction following
ollama pull mistral:7b       # Mistral — fast and efficient

# Verify
ollama run llama3:8b "Hello, are you working?"

In n8n, add an Ollama node, set the base URL to http://your-server-ip:11434, select your model, and start building.

The Bottom Line

AI nodes turn n8n from a workflow tool into an autonomous operations platform. You can build automations that think, write, analyze, and decide — while keeping costs near zero when using local models.

The 10 automations above are just the starting point. Once you chain AI with databases, APIs, email, and Slack, the only limit is what you can describe in a prompt.

Ready to build? Browse FlowForge's AI-powered templates →

---

Running local models? Check out our RSS → AI Newsletter and Sentiment Analysis Router templates — both work with Ollama for zero-cost AI.