Large Action Models actually

I’ve been trying to understand where Large Action Models actually fit in real software systems, not just in theory. Most of what I’ve seen so far is either demos or very controlled environments, but I recently read this breakdown that explains how they’re supposed to coordinate actions across tools instead of just generating text: https://www.trinetix.com/insights/what-are-large-action-models-and-how-do-they-work . It reminded me of a small internal tool I worked on where we had an “AI layer” deciding which API endpoint to call next, but it was still very rigid and rule-driven. I’m wondering if anyone here has actually seen something like this running in a real production environment where the model has some real autonomy, or is it still mostly just structured automation with a smarter interface?