AI is the last thing you do. It doesn't do much good to code AI while the game is being significantly altered. I wouldn't want Brad to put real time into the AI before getting some real playtest time so we can get feedback on whats fun, what we need more of, what we need less of.
The decision is, do we lock our design, do AI work, then release playtest for minor tweaks (mana cost of a spell, training cost of some weapons). Or do we will release for playtest feedback, be open to feedback for significant changes (like halving the amount of monster lairs on the map, doubling the amount of xp you can get from quests, or adding a new spell), then do the AI. This is Stardock, so you know what answer that will be. More open, more feedback.
We will get people who will speak poorly of us when the early beta version doesn't have good AI. But there will be far more people who understand and offer great feedback. And in the end the game will be better for it.
