So in the absence of answers to that kind of question I think the best answer might be small numbers of control and tech types sent down to the surface to run localized units of war robots. (It's all very good to have a giant 'command center' in the sky) but at some point after you land you need eyes & brains on the ground. People who can make split second decisions at close quarters for your platoons and companies etc of robots in real time.
This might especially be the case because they won't/don't understand us at first contact i.e. we're 'alien' to them. Fighting their own kind? - no problem their robots can probably be at least partially trusted with a degree autonomy about tactical issues . Fighting us though? that means re-programming their actions on the run. And you cant do that if you don't see what works and doesn't in real time.
They do the calcs and decide - again that its time for an armistice.
The plus to this idea as I see it is room for lots of discussion around small unit tactics, spoofing and deception, booby traps and ambushes etc.