In my years as a simulator instructor, I found more simulator crashes were caused by pilots who had insufficient knowledge of advanced coupler modes than anything else. Autopilots are a powerful aid to pilots, but only if all aspects have been thoroughly taught during training and if the pilots maintain a thorough knowledge of the system through a mix of self study and continuation training from their training organisations. Real life is not just a computer game whose only consequences are having to reset a system.
It's my belief that insufficient time is given over to this on both the theory and practical side in both conversion and recurrent training. For the sake of a few extra hours of simulator training time offline and in a classroom, operators are trying to save money in the wrong place, especially when compared to the overall cost of obtaining and operating a very expensive, highly sophiscated airborne system. In the end this attitude is self-defeating because accidents lead to hugely increased insurance premiums and lost contracts.
I remember some years ago in a previous life when I was a training manager for quite a sizeable operation, the operations manager was of the opinion that we should reduce annual recurrent simulator training by 2 hours per pilot to save money. I told him that, au contraire, we should increase it by at least 2 -4 hours per pilot to save accidents and lives and spend at least an extra day at the simulator using the GFSS to improve knowledge. Sadly, the compromise was that we just retained the system we had. Unfortunately, when it comes to a fight between training and operations/commercial, the latter nearly always wins - until there's an accident.