This is a bit off-topic, but it seems you hit the problem pretty thoroughly in your main post.

I am curious, have the AI heroes ever been a problem to you? I know my sovereign, depending on build, may do the majority of damage in my empire. If I am going to use a hero, bother leveling it up with its own army, it will generally be able to do something significant. I have never seen an AI champ/sov do anything signficant in a battle.

I keep capturing sovereigns, high-level sovereigns 16-17 on Hard/Expert, that have random or breadth-first trait selection, so despite being endgame level they don't have any traits over two-levels deep in any tree. They will completely fill out the general tree every time. This means no attention has been paid to leveling strategy. I know strategic map army usage is critical, if the army never gets there it is a non-starter, but it seems like in most successful strategies, some heroes/sovereigns need focused leveling because powers/spells are such powerful tools. They are suped-up one figure units without focused leveling (though with crazy hp in your example).

Wow, I can see why you have to rely exclusively on units on insane. I was not expecting them to be so useless, but since you are always behind the tech curve on that difficulty, it makes sense. I don't normally play an unmodded sov outside of challenging/hard, but if the AI was already significantly up the tech tree in the early phase, I don't think there would ever be a sweet spot where heroes would be effective.

On the lower difficulty levels, the lack of HP boosts mean that AOE spells from an evoker type sov can decimate entire armies in the relatively early game. The sheer number of hit points on those late level troops you are posting would make that far less devastating, but games are usually effectively over before troops like that can be produced. I don't think I have ever seen an AI troop with 150+ hitpoints.

I have never found a melee hero nearly as effective. Again, they have to hit a sweet spot ahead of the tech curve, but it must be much before, and it is not easy to get them to dominate a battlefield. But in a game that doesn't hit the highest levels of tech progression they can be useful.

I wasn't expecting a criticism of hero progression, but I didn't take in to account the effect of the game difficulty you were playing on. I still think one of the big weaknesses of the AI is not taking advantage of heroes, but I don't think that has much effect on Insane difficulty because they are apparently outclassed too quickly.

I fully agree with you Nick, this is quite a big problem that really hampers the AI. I think early days in the FE beta Brad spent quite a bit of time in making the ai armies come together at logical places creating some nice armies. For some reason this seems to have deteriorated over time. It's one of the weakest AI points currently: too many easy kill lonely units or heroes.

Those are his own troops, not AI faction troops. As such, I believe that those don't receive any health boosts aside from structures like the Guild Grocer, faction abilities like Tough and Ironeer blood, Aura of Vitality, and a few special pieces of equipment. Base health value for trained units is six per figure, plus two per level per figure, so it isn't that hard to get a single unit with a maximum health of 20 or more per figure. If you go in for the health-boosting traits and equipment, you can increase that by something like 10 health per figure, and if you have a few towns with Guild Grocers or find a few Wild Game resources, it isn't that difficult to get a faction-wide health bonus of 20% or more. In my current game, I have four-figure units that don't have any health boosting traits or equipment and which were not trained at a location with Aura of Vitality, yet at level 10 they have 125 health (six base health, plus 20 health from levels, plus 20% health from Guild Grocers and Pastures, multiplied by four figures) - twice what my level 14 sovereign has.

The biggest problem for champions against troops (especially for melee champions) is that troop health increases by 2*figure_count per level before bonuses and penalties, and most champions won't have access to weapons and armor which are significantly better than (or in many cases even equal to) the equipment used by troops, which means that melee champions often deal damage equivalent to one- or two-figure trained units while being at least as vulnerable to troop weapons as the troops are to the champion weapons, and except for spells which scale per figure in the target unit (like Blizzard) or which are already enormously powerful (like Mana Blast, assuming you have a large mana reserve), most mage spells simply won't keep up with late-game unit health. This is before getting into the various city spells that improve units trained there - it isn't uncommon to have a three essence fortress, and in my current game I have a four-essence fortress, and a site that I can turn into a five-essence fortress, and I have the Auras of Grace, Vitality, and Might, as well as Heart of Fire available for city enchantments. It would take a truly exceptional champion to compete with even a level 1 unit that comes out with all of that. Granted, I've never seen the AI take full advantage of these enchantments or of fortresses.

Some updates/clarifications, since my previous post in this thread:

-When I mentioned my lvl 20 sov just cast heal/slow/haste, and while he's maxxed out 3 spell lines, they're water, earth, and air.  Not fire, else he'd be casting fire spells as they're fairly powerful.  I depend on air/earth/water in Insane games, for their strategic spells.  It's more important to damage/delay/scatter enemy armies than use fire spells in tac combat, at least for my playstyle.

-My regular units shown above have no 'refined' stats.  They're 'men' race.  The spear maiden in now lvl 15:

I have researched all techs except warg riding, and of the 'refined' techs I've only researched the gold-enhancing one IIRC.

Here's some of the units I'm facing, all Kraxis:

The strength of the AI units at Insane isn't bad, in fact its pretty good.  I can sorta match them now that:

-I've researched all techs

-I've carefully protected my early units so that although, for example, the spear maidens only have leather armor (and not the boots/gloves, instead the soldier armor), they have decent stats

-I dedicated my first city as a fortress (3/3/3, with scrying pool for 4 essence) with all the possible unit-enhancing buildings/spells/etc. -- quality over quantity.  The challenge is surviving until this unit-producing city is online (need place-holder units until it is) then the challenge is keeping the 'keeper' units alive so they can level up.  And then the challenge is getting sufficient gold to upgrade their equipment (where possible -- you can't upgrade accessories for example, or add horses).

The 'my units vs AI units' comparison is good.  It's the 'my regular units vs my champs' comparison that might need tweaking.  Even with decent bows/swords/etc. I'd rather cast slow/haste/heal than have champs/sov do damage, as it's essential to keep my regular units alive -- they need leveling to compete and that takes time, they are relatively slow to build, and they cost a lot of crystal/horses/metal/etc., so keeping them alive is important.  Slow helps delay enemy units approach (and on maps with choke points, allows stopping an entire flank), allowing time to kill off enemy units a few at a time.  Haste gives regular units more attacks -- a force multiplier, and if you're going to multiply a force it should be your strongest force.

Fire spells are different, they can do good damage, but given the choice of finishing off 1 enemy unit (best case for using fire spell) or healing a quality regular unit that might otherwise die before the next chance to heal it, I'll pick the latter every time.

AI sovs can do pretty good damage.  My previous post gave them too little regard.  Since then Magnar wiped out one of my stacks, all by himself.  On the other hand, it was one of my 'placeholder' stacks -- 1 fairly useless champ (Flurin lvl9 assassin) and 8 strong melee units (it was a pretty impressive win for Magnar nonetheless).  Had the stack been one of my 'main' armies (4 melee, 4 ranged, champ/sov healer) then Magnar would have died easily, I think, even if he was in a full 9 unit army (as Kraxis and Yth has previously).

On the other hand, that city was Yth's last city and Krax was attacking Yth.  Yth is gone, so Krax declared on me, and just happened to have his armies right there from his previous war.

Magnar declared on me about 40 turns ago, and I never saw one of his armies.  Krax is between Magnar and I, but still, Magnar should/could have pre-positioned armies to hit me, and if Magnar couldn't for some reason then he shouldn't have declared.

So, there's been improvement (no pioneers/scouts in armies, initially the AI armies are mostly full and come in bunches), but there's still a lot left to improve (subsequent armies are piece-meal and incomplete, declaring without preparing and even without attacking, etc.).