Evaluating applications of AI for siting and permitting energy projects

What makes a permitting or development problem 'suitable' for  AI? And better yet, how can we characterize the problems that AI will be better at solving than the status quo today?