I’ve actually been ignoring the state machine altogether, and just scripting things in by hand currently. Are you defining the variables for transitions between said animations in the state machine (so between run/idle & attack/hit)? I’m attaching a screenshot example of how I have my eventgraph setup for my AI currently. From what you see in the attached image, it goes into a branch with a check if the AI is alive or not, if it is, then it fires off the Play Anim Montage node with the HItAnimation referenced.
You can always just script in the AI’s logic, situational animations in the event graph for the moment, and use Play Anim Montage to trigger animations. Thats what I’m doing for the time being. It’s so far been working flawlessly, and its been less of a headache overall then working with the state machine. Not sure if that is ideal long term or not, but its one way of handling it right now.