So I'm changing some stuff in EB, but maybe I should think on it some more...

[Ludens]

A few thoughts:
-Why increase the price of Scythian horse archers? Are they noticeably cheaper than alternative low-level HA units? Unless you mean steppe HA in general, in which case it may impede the steppe factions that are already struggling with a low income.
-The team removed the higher forge-levels because they did not fit in the EB statting system. They actually considered getting rid of the forge entirely, something I support. EB unit stats work on the WYSIWYG-principle, and the differences between units are subtle. If a unit wears a normal chainmail suit, he gets the same base armour as all other normal-chainmail units (there are other modifiers for armour value, but they are not important in the example). It's not unrealistic that a good armourer improves the quality, but +3 is a bit much: a +3 chainmail suit would approach the protection of a muscled plate in the EB system. For that matter: muscled plate and similar high-end armour represents the peak of the armourer's art, so I doubt their quality could be improved substantially. Back to the example: high-quality chainmail would be considerably more expensive than the standard stuff, and armour already represents the main cost in equipping a unit, so recruiting a unit with upgraded armour should also cost more.
-Type III governments are rarely used by players: maybe you could look at balancing those?
-There is no way to make the A.I. competitive with the player when it comes to expansion. The human mind is simply far better at planning. I don't think there is any solution beyond unbalancing the game further in favour of the A.I.

Your ideas on improving A.I. navy use sound sensible. Let me know how they turn out. Reducing the costs of siege equipment and elephants also sounds good, although keep in mind that in reality only the wealthiest states could afford such units.