Unlike Google, at least they're not trying to get their customers to wear their nerd helmets in public. I've been trying to come up with a way it would be useful for me for the last week and I also keep coming up empty-handed. Worse for drafting as you've already pointed out. Almost all my design sketching today is done on paper and with physical models, to be able to easily manipulate real objects in space, sometimes in collaboration with other people who doesn't know AutoCAD/Rhino/Sketchup. Replacing it with virtual objects kind of defeats the purpose. Which leaves representation and presentation. I can see how "immersion" can help sell a project to a client, but I also see hundreds of roof goats that will have to be put down at later stages. Working with plan, section-elevation and perspectives (even a digital model fly-through) allows you to direct attention to what you think is the design's strengths while hand-waving away the things you haven't had time to think through or talk to an engineer about. Plus, who the hell has time or money for it.
I don't see it happening, honestly. There's no reason for it, adoption is low, and the one aspect of Google Glass that made me go I fucking want that was an AR motorcycle helmet. There is exactly no part of a motorcycle's gage cluster that fits within the DoD's sight window. Putting essential stuff up in your field of view where you can reference it when you need to is actually a great application of AR. And like I said, heads-up displays on cars are, in my opinion, a net good. I think they're largely dumb on fighter aircraft because fighter aircraft have been refining heads-up displays for fifty fucking years.