I dig the AI, its pretty fun in the current state. Even if it doesn't really play to nuances and specialized rules, as a blunt intrument it usually cobles something reasonably effective together for a landgrab, if you can just find that threshold were a boost in income overcomes its other deficits. Sometimes that can go by side, its land/air game is usually a lot stronger than its naval game, so for maps where the balance really hinges on the water its easier to exploit the AI there. Sometimes one side is more dependant on that kind of thing than the other. A lot of times for second world war games that's the Allies, though I think in Iron War its a little inverted from the norm, since the Europe Axis side and mainly W. Germany can fall apart pretty quickly when Allies get the upperhand around the channel, whereas Japan is a little better at managing a land war even after the IJN is pushed around. I think AI Allies have the edge there probably, and its more entertaining to play as Axis because of the way the opener is set up. Not sure how the world map compares to the Europe one, but might be a fun change if you wanted to try one where AI Allies seem a bit stronger than AI Axis under vanilla conditions. But in general I think a lot of maps could probably support an AI challenge, its just a matter finding the right number for a handicap. So for WaW if the AI is doing well as Axis maybe the vanilla challenge is good for the one side, but Allies might need something extra. Usually I'd go for a modest increase (probably percentage increase for PUs at the lower end 10-20% and see if it makes a difference for their ability to trade more effectively, then just see how high you can go before its ridiculous hehe.) The spread you end up with there probably tells you a rough ballpark of how dependant the map is on specialized rules, higher level purchasing strategies and such, compared to the basic WW2 maps where the AI does pretty well for itself. So for the difference in complexity between say v2 and v3, or v5 and Global, there is probably a number for the later where the HardAI will become competative again, even if its not prioritizing added things like objectives or VCs or tech.
I have noticed more recently that the Hard AI will now bomb semi regularly, if it has enough bombers, but whether they buy them or not seems to depend more on the production frontier, like low production but high cash scenarios. Not sure a map that's overly dependant on consistent SBR for balance would work well at this point. The AI also struggles with things like intercept or scrambling where those dynamics are utilized heavily or with bases that use stuff like dynamic movement boosts. So G40 type solos really fall apart there. Objectives you can probably overcome with an income bonus but the AIs handicap in not managing an Air/Naval bounce from bases is something you can't really compensate for as well. Probably every map would have exploits when the combat mechanics rely on map position (any kind of terrain effect reducing or improving movement.) The AI will do what it can do from the current position (if it ends up on those kind of tiles) but won't build like a competant strategy around the map. Its pretty good at projecting power with air and cheap fodder though, so if bombers are cheap they will spam them pretty well for dark skies, when the income outpaces production capacity. Happens in Iron war when the AI gets backed off the front, or is knocked down to just a lone factory. Otherwise I see them mainly bought for the AI air umbrellas around island territories. The AI is actually surprisingly good at projecting air power vs enemy naval under those circumstances. They will make some clutch strikes to sink fleets, but then often don't follow up with re-establishing fleets or transport capacity of their own. See it happen to Japan at the home island or like with USA out of England which is kind of cool, but once the AI is boxed in it becomes a lot more conservative with its TUV, so its more likely to withdraw than trade TUV at disadvantage, even if a human definitely would to try and turn a game around. When the chips are down HardAI goes for like the slow bleed instead of going out with a bang hehe. But sometimes that can make for entertaining attrition endgames that last way longer than you might see otherwise.
Anyhow, yeah I like Frostion's maps for the AI, since many were built out with the idea that the AI control some stuff already. I think it works on probably any map, but definitely better where the income/production or basic victory conditions are set up more along classic economic domination lines rather than the TKO victory conditions of v2 and later. Right now the AI doesnt understand VCs, so what matters to it is really the income/production and distance to nearest enemy capital. It will knock off TUV when it has the advantage or to take over territory along the way, but won't airblitz to take a capital or secure a specific VC for the win or things like that. I think for the naval game one place where it struggles on most maps is with blocking, and also stacking (like around carrier groups for example) to cover key zones like canals or straights. It spams transports pretty wild over time and launches more than you'd think to pretty good effect, but that only gets you so far, if a particular zone is do or die for naval balance due to map dynamic movement features that the AI doesn't get. But even still, I had a good time messing around with Iron War.