I can only +1 on @Deltakosh answer
1 Like
I think the main issue is that implmenting AI is too specific to the individual project. I went through all major approaches starting with Behaviour Trees, followed by GOAP and UtilityAi. There is open source repos for each approach on GitHub. This stuff is in good supply
So it is more like getting the behaviour out of AI agents the way you want with whatever AI approach.
I do not think there is a way around trial and error. In my case, it took me 5 or so rewrites of my AI code and I eventually settled with a GOAP/UtilityAi mix.