There's a category of business work that's too bespoke for off-the-shelf SaaS but not complex enough to justify custom software: lead enrichment pipelines, invoice processing, internal reporting, data syncing between systems. n8n, the open-source workflow automation platform, combined with LLM APIs, has become our go-to solution for eliminating this work entirely.
What Makes n8n Different
Unlike Zapier or Make (formerly Integromat), n8n is self-hostable and code-friendly. You can write JavaScript nodes for custom logic, iterate on complex branching workflows visually, and connect to any API with an HTTP node. For technical teams, this is the right balance of visual speed and code power.
The AI Node Pattern We Use Most
The most valuable pattern we've found is what we call "classify and route." Incoming data (emails, form submissions, support tickets) is passed to an LLM with a structured classification prompt. The LLM returns a category and extracted fields as JSON. The workflow then branches based on that classification — routing to different email templates, Slack channels, CRM pipelines, or database entries.
A client's invoice processing workflow went from: receive PDF by email → manual data entry into accounting system (15 minutes per invoice, 40 invoices/week) to: receive PDF → extract data with Claude → validate against PO database → auto-approve or flag for review. Net time saved: approximately 9 hours per week.
Deployment Considerations
Self-hosted n8n on a small VPS ($20–40/month on Digital Ocean or Hetzner) handles most workloads comfortably. For high-throughput or mission-critical workflows, consider n8n Cloud or Kubernetes deployment with horizontal scaling. Always set up execution logging and error alerting — a silent workflow failure is worse than a loud one.