I think the biggest issue is that people dont like to look at the object through phone when it is much clearer here in reality. So looking for themselves and looking to show it to assistant naturally split into two separate processes. While looking “for yourself” the phone is randomly dangling in the hand, making the stream sea/sickening.

So I thought, VR headsets have a good see-through mode, which could also be streamed. It also could easily display a pointer from remote. Thus both processes would merge into one, and you could directly comment on what is on focus.