Most KPIs can be measured similarly to regular testing. However, the throughput of that testing will be much different.
What this means is that you are going to be exposing more low-level errors in the application than if you were running a standard test cycle.
Where you might have had one tester work a month on testing the application against 2 different setups (Say PC and MAC), you are now able to have one machine set up to run through the same test cycle in 1 day, and if you use something like VMWare, you could have tested this against a multitude of different system configurations, virtually.
So what this means is that you are now able to get at KPIs which would have been very difficult to measure before, such as MTBF, or error frequency, and you can also draw relations to reliability since you can, essentially, have the testing running 24/7 and all you need is someone to track the errors and measure these KPIs.
Performance is another huge one. You should be able to track most performance issues during a manual test cycle, but there may also be exceptions which cause drastic changes in system performance only after X number of hours of continuous use.
I would advise that you start by looking at what you want to know, then to ask yourself (or the staff) what questions that needs to be answered to give you that information. From those questions you should be able to derive the KPIs to meassure.
Just to have the KPIs and trying to figure out what they mean is for me, just a statement that you are running around blindly not knowing why you are measuring in the first place. Remember that the management do NOT want KPIs from you they whant information that is worth the amount of money they put into testing.