Building an AI agent inside a 7-year old Rails application
I work on a large Rails monolith used by social workers, where access control and correctness matter more than experimentation.
For a long time, I assumed AI agents didn’t fit this kind of system. After RubyConf, I realized the problem wasn’t AI — it was how I was thinking about where it belongs.
I wrote about building the first agent inside our existing architecture.
https://catalinionescu.dev/ai-agent/building-ai-agent-part-1/
Post a comment