This is a suggestion to make the AI use/build things well... smarter... or at least, not stupidly. It'll probably contain some jargon that you'd only be familiar with if you dug into the xml files, so I'll apologize before hand if it doesn't make a lot of sense for those who don't.
It all started with this thread: http://forums.elementalgame.com/425251
For those who don't want to read the thread: Basically, the AI is doing things that it shouldn't. In the thread's case, it was casting curse on those with no def (which was pointless). That was when I explained that the AI doesn't actually know what the skill "curse" actually does, and that it would help if there were conditions to alter the skill's priority so that it doesn't use it when it isn't necessary.
It didn't take too long before I started seeing this behavior pop up all over the place in my own games (actually I've noticed for a while, but it's glaringly obvious now). For example, I watched Tarth builds an Inn in a nearby city... that would normally be a great thing (since it insists on pumping out useless units 98% of the time - but that's another matter)... except the city it builds the Inn in... is full of people. What it SHOULD have built, was a granary. But nope, it wants an Inn (and that's the ONLY building it built in that city for 250+ turns - at least build the barracks if you're pumping troops... jeez).
So what I'm suggesting is: instead of assigning these straight forward "10"s and "15"s value as priority for these buildings (and skills), that the AI doesn't really understand, what we should do is turn priority into a variable that can be defined by humans that knows what those things are suppose to do. In the case of the Inn, the variable would be a calculation based on the number of available food remaining in the city... say: available food/5 + 1 (to say that it's a bit better than pubs - due to construction time). So if you have a lot of food.. the priority is very high (makes sense, that's when the Inn would be most useful), and if the food supply is already low, then priority is also very low (so it won't build it when it really doesn't need to). The opposite for the Granary, whose priority could be based on a value minus the available food supply (to get higher priority when food is low) multiplied by the potential growth of the city (before the zeroing out). So if there is a lot of growth, and no available food, priority for granary is very high.
The same thing can be applied to skills. For "curse", you'd get the target's def and use that as its priority. So if they have high def, curse gets high priority. You can also do the same thing for other skills, like fire spells that has priority based on its target fire resists. High resist, low priority, a fire weakness? Super high priority. Makes sense, right?
I'm pretty sure that if you do this, you can get the AI to build and use skills in a much better manner.