They're at the 1980s toy computer level of usefulness - fun to play with, a few useful applications, but hopelessly clumsy and underpowered in terms of real user needs.
No one was sure what real user needs would be in the 80s. It turned out the all-time killer USP was a global data network, which could be accessed through keyboard+screen terminals and pocket devices, and which replaced a lot of slow paper and phone call transactions of all kinds with near-instant access.
VR/AR is currently a wart on that. It won't come into its own until there's some equivalent new killer USP, which isn't just a different way of accessing what's there already.
I'd guess there's going to be some kind of live AI rendering of both real and simulated interactions. But it's going to have to be far beyond what we're used to today to be interesting.
No it very much matters what they can do, and it's "not much". If they could do a lot more any of these products would be very popular.