If you've been curious about AI automation but felt overwhelmed by the tooling, n8n is your gentlest on-ramp. In this walkthrough, I'll show you how I built a simple but production-ready lead qualification agent that saves my client roughly 12 hours per week.

What we're building

A workflow that watches an incoming email inbox, enriches each lead using publicly-available data, scores them against a fit rubric using GPT-4, and pushes qualified leads directly into a HubSpot pipeline with a Slack notification to the sales team.

Why n8n over Zapier or Make

Three reasons I reach for n8n first:

• Self-hostable. You own your data, your workflows, and your uptime. No per-task pricing anxiety.
• Code when you need it. Zapier gives you blocks. n8n gives you blocks plus a JavaScript node when the logic gets tricky.
• AI-native. n8n has first-class LangChain integration, OpenAI/Anthropic nodes, and vector store support built in.

The workflow, step by step

Step 1: Trigger on new email

I use the IMAP node to poll a dedicated inbox (leads@yourcompany.com). Every 5 minutes, n8n checks for unread emails and passes them into the workflow.

Step 2: Extract structured data with GPT-4

The raw email body goes into an OpenAI node with a structured extraction prompt. GPT-4 pulls out: name, company, email, phone, use case, budget signal.

Pro tip: use the "response_format: json_schema" mode. It guarantees the model returns valid JSON matching your schema - no more regex parsing.

Step 3: Enrich with public data

Using the company domain, an HTTP request node hits Clearbit's free API to grab company size, industry, and location. This is what turns a name into a qualified persona.

Step 4: Score against your rubric

Another OpenAI node with a scoring prompt. You describe your ideal customer profile once, and GPT-4 returns a 0-100 fit score with reasoning. Anything above 70 goes to sales.

Step 5: Route the outcome

An IF node splits the flow: qualified leads (>70) create a HubSpot deal and post to #sales-hot in Slack. Unqualified ones get a polite auto-reply and a soft archive.

What this costs to run

My client's workflow processes about 400 leads a month. Costs:

• n8n self-hosted on a $5 DigitalOcean droplet
• OpenAI API: roughly $8 per month at current GPT-4o pricing
• Clearbit free tier: 25 lookups/month free, then pay-as-you-go

Total: under $15 a month to save a salesperson's afternoons.

What could go wrong (and how I handle it)

AI workflows fail in weird ways. My non-negotiables:

• Error routing: every node has an error output that pipes to a monitoring channel. If something breaks, I know in seconds.
• Human checkpoint: for the first 2 weeks, every scored lead gets a manual thumbs-up/down in Slack. This trains me on where the scoring drifts.
• Cost caps: a monthly usage cap on the OpenAI key prevents a runaway loop from bankrupting anyone.

The bigger picture

This one workflow saves 12 hours a week. But the real win is compound: once your team sees what's possible, the ideas start flowing - support ticket triage, invoice extraction, content research, meeting notes into CRM entries. That's where AI automation pays off.

If you'd like help identifying the highest-ROI automation in your business, book a free 30-minute call. I'll walk your workflows with you and tell you exactly what's worth automating first - even if we never work together.