The automation stack consists of the Claude large language model, the n8n workflow tool, and an MCP server acting as middleware to handle user queries and API responses.
Use methods like team interviews, Pain Point Bingo, or AI-driven questioning to find repetitive, error-prone tasks that are worth the effort to automate.
Before building, verify tool support via native n8n nodes or REST APIs, determine the authentication method like OAuth or API keys, and decide between personal or technical user accounts.
Provide a detailed prompt with context, clear rules, and do's and don'ts, then ask the AI to produce a concept for approval before it generates the actual n8n workflow.
Run the initial AI-generated workflow, identify and fix errors like credential issues or logic bugs by providing feedback to Claude, and repeat until the automation works as expected.
After deploying the working automation, consider adding advanced features like robust error handling with notifications and state rollback capabilities.
Jobs that call for the skills explored in this talk.