A working multi-agent architecture in large enterprises

AI Hype aside, how many of you have truly seen a working multi-agent deep embedding in large enterprises or large complex environments?

If you have, what's your stack/architecture?