What it is
Platforms like n8n, Make and Zapier let you assemble automation visually (with a mouse, out of nodes): "an email arrives, an AI node extracts the data, creates a task and posts a notice to Slack". The AI node (an LLM) has become a standard building block in these builders.
Where it came from
Workflow automation has been around for a long time (Zapier since 2011), but adding LLM nodes and AI agents in 2023-25 turned it into a rocket. n8n (fair-code, self-hosted) became the favourite of technical teams, while Make is the visual favourite of marketers.
Why it took off
n8n raised $60M in March 2025, has around 200k active users and 3,000+ enterprise customers; 75% of customers use AI tools they built themselves. By the end of 2025 came a $180M round at a $2.5B valuation. ARR grew from roughly $7.2M to roughly $40M in a year. The reason: gluing AI to your own apps without a developer is mass-market demand.
How to use it today
- Take one manual chain (processing inbound requests, content posting, reports) and rebuild it out of nodes.
- Pick n8n if you need self-hosting and control over data; pick Make/Zapier if simplicity and ready-made integrations matter more.
- Add an LLM node for the "smart" steps: classification, extraction, drafting a reply.
- Log errors and set up retries — automation with no failure handling is dangerous.
What to watch out for
The more complex the flow, the more brittle it is: one service changes its API and the whole chain breaks. The cost of LLM calls inside a loop can creep up unnoticed. For personal data, choose self-hosting (n8n) and do not push sensitive material through someone else's cloud without vetting it.