It depends on how accurately the avatar is able to represent important information: emotion, attention, state of mind, etc. There's a lot of bandwidth in looking at a face (and bodylanguage as well), that's where the value in face-to-face meetings is.