I think that would be very optimistic! The AI's placement and movement of troops is often poor in both attack and defence anyway; we've all noticed that. As a software engineer of old, I think I would find the challenge to improve it quite enormous. You'd need to choose overall between many set troop patterns from quickly changing heuristics, and you'd still need to be flexible to adapt to changing terrain and opposition.
That's right: in the city! 
True.
You don't know that, one way or the other. Programming any AI to deal with human interaction is very hard indeed. Try it yourself. As a parallel, after more than a decade of trying to get human interaction bots to be convincing (http://alice.pandorabots.com/ was one of the first), they're still VERY easy to break. You'd think that language and conversation would be an easy win, but it's not. I'd be kinder on the Civ AI programmers. It's a hard job and we're probably many years away from anything convincing.
Oh, it's important enough all right. It's just not cost effective for a small games software company to fix by themselves. Their payback would never be able to fund them enough to hire a large enough team to do this.Isn't that the story of this game though? Programming a competent AI is really the only important task the development team had. ... I don't see this getting fixed, as it clearly wasn't important to begin with.
