I would argue that the bad UX of current phones and apps is a big obstacle to this happening now. Having to take your phone out of your pocket, unlock it, open

I would argue that the bad UX of current phones and apps is a big obstacle to this happening now. Having to take your phone out of your pocket, unlock it, open an app, point it at the object, wait for recognition, wait for data/processing, and only then see the result is a huge negative that demotivates most people from wanting this sort of experience on their phone. With an always on AR system with specialised hardware and software for object recognition and good depth mapping and hopefully a design that allows for multiple AR services to run and provide overlays concurrently - you should approach the level of usability that will allow this to happen.