Driving animation in Behavior Tree vs State Machine

Would it be feasible to completely drive animations from a behavior tree, rather than using a separate animation state machine?

I can see the downside of mixing the animation system and the behavior system. But there is also the benefit of not having to setup the connections between the systems.

What problems do you see with not using a state machine and controlling animations straight from the behavior tree (limitations, performance, complexity, debugging, flexibility) ?

Hi I’m currently tackling a similar issue and was wondering which approach you used and why?