GermanyTechnology and telecom2026 – ongoingn8n
AI Workflow Builder that assembles automation workflows from plain-language requests
n8n's AI Workflow Builder produced a first workflow in about three minutes instead of 10-30 minutes
Problem
Building a workflow in n8n meant finding the right nodes, connecting them in order and configuring each one. Users often copied suggestions from an outside chat model into n8n and fixed the errors that followed, because that model knew nothing about their actual workflow.
Approach
n8n built a chat interface inside its product, powered by Claude. A first design that asked the model for a complete workflow in one pass failed, because the model could not hold the full schema and node logic at once. n8n switched to a tool-based design in which Claude adds, connects and configures nodes one at a time through function calls. The first version reached production about two months after the project began, and the design later grew into a multi-agent system.
Results
Takeaway
Splitting the job into small tool calls worked where one large generation did not. Engineers still finish the last part of each workflow by hand.
Categories
- Industry
- Technology and telecom
- Business function
- IT and R&D
- Innovation type
- Product and service innovation
- AI technology
- Generative AIAI agents
- Adoption
- Off-the-shelf tool or platform
- Evidence
- Vendor case study
Sources
Source summaries
Anthropic's customer story describes how n8n built its AI Workflow Builder on Claude using a tool-based design. It reports that a first workflow version takes about three minutes instead of 10-30 minutes, and that the tool delivers roughly 80% of a target workflow. The page carries no date; the Internet Archive first saved it on March 9, 2026.
Full details are in the original sources.
