At the moment we have one command per side containing all the units of that side. Both sides/commands use the same AI thread for processing their AI events - ie assessing the situation, developing plans, reacting etc. So at the moment all the events get banked up in a queue and processed serially. Multithreaded AI would see us say having one thread for each command and having multiple commands per side. So you may end up with say eight commands and eight AI threads. These threads would then be assigned to one of the CPU cores on your machine. Now some machines may only have four cores, others may have more. At any one time we could process say four or more AI events instead of the current one. It is a way of spreading the processor load so there is concurrent processing being done not serial processing. Or in other words utilising the existing processing capacity.