I think definitely your second example, possibly your first, could be accomplished with "Soft" AR. Remote personal interaction would still be reliant on the latency of the network/medium, and without touch, I'm not sure how much more use superimposing someone believably into my field of vision would be as opposed to having a display or a holographic avatar of some sort that doesn't require full Hard AR.