Thanks Kyle, and that's actually the kind of advice I was looking for, i.e. avoid MS UM for VR.
I'm focused more on dedicated VR vendors - Nuance, SpeechSwitch, and any others that may be out there - who provide Enterprise-class VR, not "and it does VR too" providers like MS. Maybe that's why our perceptions about usability are so different.
As I mentioned, I've had pretty good success recently with high-end, speaker-independent VR, especially when you can limit the "vocabulary" and the VR system is context aware, knowing what responses are reasonable or possible given the call flow.
As for pronunciation of names, again, the dedicated vendors seem to be keenly aware of these problems. SpeechSwitch, for example, provides a phonetic editor for a username, so you can help the system deal with names like Nguyen.
I suspect there aren't a lot of folks choosing IP Office that are willing to spend the money for high-end VR. We are, and in retrospect, I'm probably fishing in the wrong pond asking this question here. But thanks for your thoughts!
Geoff