Automation Workflows with n8n

n8n has become my go-to tool for connecting AI services to business workflows. Here's what I've learned.

When to Use n8n vs. Custom Code

n8n excels at orchestrating API calls across multiple services. Use it when:

You need to connect 3+ external services (Google Sheets, Gmail, Slack, etc.)

The workflow logic is primarily branching and conditional

You need non-technical stakeholders to see and understand the workflow

Use custom code when:

You need complex data transformations

The workflow has tight latency requirements

You need to handle edge cases programmatically

n8n's retry mechanism is good but you should supplement it with:

Error workflows that send notifications on failure

Dead-letter queues for failed items

Idempotency keys to prevent duplicate processing

The most reliable pattern for LLM calls in n8n:

1. Collect and format input data

2. Call LLM with structured prompt + few-shot examples

3. Parse response with a validation step

4. Retry with modified prompt on parse failure

5. Log everything to a spreadsheet

This pattern has been running in production for months with 99.5% uptime.