Military AI and human control
Theatre-level Control vs. Lower-level Control
Whether a Theatre HQ is 'green flagged' or not, it makes its own assessment of the threats posed and adjusts its deployment accordingly. Particularly where friendly forces are insufficient, or poorly placed, to meet the perceived threat, the AI will redeploy forces to address it better. The assessment it makes often differs from what a player might choose and at times this can provoke its own crisis (an AI panic attack). In these circumstances using AI control at any level can be perilous.
However, AI control can still be very effective if Theatre AI is simply detached from the rest of the chain of command and subsidiary (now more independent and compliant) HQ are 'green flagged' - without the Theatre AI involved or linked, lower level HQ focus solely on the player set objectives which can be adjusted to deliver the desired results.
The player may then choose what formations are handled by the AI and at what level the AI acts - this then allows the player to set objectives for the AI at different levels of detail.
Rebels and partisans
AI control of garrison and occupation forces lifts some of the weight off the player's shoulders and is both easily arranged and effective.
Key provinces with ports, cities (IC centres), airports and resources should normally be protected by dedicated garrisons as should known choke points and pinch points along the supply train. Garrison or Militia divisions of two brigades are more than sufficient except where the threat comes from a major power.
In addition, a mobile, fast-response force of Cavalry and/or Militia divisions should always be available. Again two brigades is sufficient for these divisions unless the external threat is significant. Where the threat comes from a neighbouring or enemy major power, Infantry should be used instead of Militia.
Military Police brigades can be added to help suppress partisans. Nearby Light Armour, Tactical Bombers/Combat Air Support and Paratroopers can always be added later if major revolts occur and need to be suppressed quickly.
A dedicated HQ is needed at the appropriate level depending on the size of the garrison/anti-partisan force involved along with any necessary subsidiary HQ. The HQ should be 'green flagged' to activate AI control, set on a defensive stance with objectives set for (a) each of the key (Garrisoned) provinces and (b) the ungarrisoned but otherwise important provinces that are left to be protected by the mobile reserve. An objective should also be set inside any nearby enemy country.
If key provinces are unevenly spread or not of equal value (e.g. exposed ports), a cluster defence network can be set up with Garrison Corps covering critical groups of provinces (e.g. around the ports) and Cavalry/Militia Corps covering the rest. Each Corps HQ can then be given specific objectives related to the provinces they need to focus on and it will generally carry these out unless it believes a greater threat exists.
As explained above, if the Garrison HQ are linked in a chain of command to the Theatre HQ, the Theatre AI will use its awareness of the overall threat to manage the deployment. It is worth testing whether the AI's approach tally's with the players own by 'green flagging' the garrison at their highest level of command (while it is attached to a Theatre HQ) to see what the AI does - if it starts to redeploy the (fixed) garrison then switch it back off, cancel the errant orders at individual unit level, detach the garrison forces at their highest level from the chain of command and then 'green flag' them at the corps level - it will then simply focus on the threats/objectives the player has identified/set. Denizen 15:44, 13 October 2009 (UTC)
Be careful whenever you green flag HQ's. You have to check on them every now and then because sooner or later the AI is bound to do something stupid and blow your magnificent plan to smithereens. Something crazy I've seen was a panzer army detached from the theatre HQ and set to blitzing stance on Amsterdam... It did not blitz at all and sent two full corps to surround partisans in Sweden (it didn't even fight them and Sweden was in another theatre... go figure...). Other weird things will happen sometimes : they may not attack at all, or re-shuffle a perfectly acceptable force distribution to create a blatant weakness in your lines. If you attach fleets and wings to them, they will often be re-arranged into different units. Don't expect the AI to use those assets particularly wisely either : it's perfectly capable of fighting your air force and navy into the ground by never letting them rest if you don't check their status regularly. Of course this all works well if you have overwhelming superiority. But if you don't you might check at some point to realize that half of your navy has mysteriously disappeared.
Weaknesses and workarounds
To make things work well with AI automation you need to use the preparation stance a few weeks before declaring war in order to let the AI organize the forces as it sees fit. Don't try to organize things neatly with a corps in each province, it will never agree with you. It likes to shift units around a lot. Also try to set as few objectives as possible. More objectives mean more confusion. Never ever set a corps more than a single objective, and don't give more than two or three to an army. Setting objectives to corps isn't great though. It's nearly as bad as micromanaging everything yourself in terms of the attention it requires and you constantly have to deal with the fact that automated corps do not cooperate with each other (most of the time they ignore each other).
Remember to update objectives as your forces advance. If you set a series of objectives that go deep into enemy territory but are roughly aligned, an AI army will follow it relatively well but if objectives aren't removed as they are reached it will leave garrisons behind, thereby diminishing it's combat strength at the front.
The AI is useful to avoid micromanaging easy battles. The German invasion of Poland, for instance, can easily be automated because it's a given victory. When facing difficult odds though, you need to outsmart the AI in order to win and that can only be achieved by commanding the armies yourself. In the same line of thought, automation works fairly well at exploiting breakthroughs. The slight problem is that it won't make that breakthrough by itself when facing equal or superior forces. For example, if a war gets gridlocked with both armies wearing themselves out in a trench-warfare sort of battle or an endless series of attacks and counter-attacks it's probably a good idea to turn off the AI, concentrate your strength at a weak point, break through, and turn the AI back on. Otherwise it will happily bleed your manpower to death for no result whatsoever.