I've been researching the current landscape of AI agent infrastructure for enterprises and noticed an interesting divide emerging. On one side you have open-so

I've been researching the current landscape of AI agent infrastructure for enterprises and noticed an interesting divide emerging. On one side you have open-source frameworks like LangChain, LangGraph, CrewAI, and Microsoft's AutoGen. They offer a lot of flexibility and developer control, but deploying them in production often means building a lot of infrastructure around them — security, observability, RBAC, deployment pipelines, etc. On the other side there are enterprise platforms that try to package the full stack together: visual agent builders, orchestration, governance, monitoring, and compliance. Platforms like SimplAI are positioning themselves more like an "agentic operating system" rather than just a framework. The trade-off seems to be: Open source → flexibility, control, no licensing cost Enterprise platforms → speed to production, governance, and operational tooling What I'm curious about from people here: • Has anyone successfully deployed LangChain/CrewAI agents in production at enterprise scale? • How much infrastructure did you end up building around it? • Are managed platforms actually reducing engineering time or just shifting the complexity elsewhere? Would love to hear real production experiences rather than demos or prototypes.