I think I have a solution to a perceived problem with the AI. But first, thanks to Brad and Derek and the team that made this game. I love it, and have put in 150+ hours since launch. If this game never gets another update, I will still love it. But I had some brief ideas on how the AI could be a little more fun to play, if you'll hear them.

First of all, I find the Hard and Expert difficulties usually make the best games. I create a custom faction as my main opponent set on Expert, then have all other factions set to Hard. This is hard enough for me (for now), as I avoid any major exploits (I play a custom Mancer faction).

The AI's reliance on the faction power ranking makes it predictable. It will be a bully once its power surpasses yours by a threshold, and buddy up to you if yours surpasses its ranking significantly. No, this isn't ALWAYS the case, but the exceptions are rare.

Here's the main flaw: the comp simply cannot organize its troops as well as the player can, so I generally know I have won when 1 of 2 things happen: 

1.) I get a higher faction power ranking than the other AIs

2.) Someone declares war on me

Why do I know I am going to win when the AI declares war on me? Because I know I will steamroll into his poorly defended cities, and soon have extra territories that over time improve my faction power ranking significantly. With this extra power, I eventually conquer all the Ai's territory. The AI NEVER makes reasonable peace treaty offers,  instead demanding 5000+ gold for a peace treaty when it is in a tenuous position. Only once it has been mostly decimated will it suddenly be willing to pay for Peace - and at that point, why not finish the job and take his last city or two?

I appreciate the time and capital that goes into extra coding and updating this game. So with algorithmic simplicity in mind, here's my suggestion:

FIRST: Up the faction power ratio needed for the AI to decide to become the aggressor, so that it is more likely to go to war when it is truly ready to pose a significant invasion threat. The faction power calculation is generally accurate, but not accurate enough to guarantee a true advantage. Just because they have more troops and a better economy doesn't mean I can't take the "superior" player's closest city the turn after they declare war on me. 

SECOND: Make the AI a lot more willing to stop wars when it is winning - in exchange for tribute treaties. Having a second or third tier of tribute might make sense, too, if it is easy to code and add to the diplomacy interface. 

Regular (as in game): 10% of income

Harsh Sanctions: 30% of income

Puppet State: 50% of income

Alternatively, the AI could accept more reasonable amounts of gold for peace (not demand more than the player has), but then immediately badger the player for tribute treaties, or the war starts back up. 

THIRD: Allow ceding of territory. Not sure how algorithmically complex this would be, but I think it would make wars even more interesting. A dominant AI could demand you send them a city in lieu of war - and likewise you could give conquered territories to your allies, making them happy while saving you the headache of diverting more troops to defense of a distant newly conquered territory.

Ceding outposts would allow the AIs to get what they really want - resources. Likewise the player could bargain for crystal crags/shards instead of  taking cities.

BONUS: Can the AI be made more likely to fill EVERY slot available in a stack? Especially when it is in a war?

By making war less "all or nothing," the conflict and tension can be heightened, I anticipate. Quests and resources out in the wild could swing the balance, allowing a defeated tributary player to suddenly rise back up against their puppeteer state. In this type of environment, the Master Quest becomes a very interesting trump card.  

AND FOR THE RECORD: It's how the AI behaves toward the player that's the real issue. The AI interacts with the other AIs very well, in my experience.