I've sold some AR projects to customers. Just product visualizations for trade fairs etc. nothing too fancy. It's cool and fun to work with but my takeaway was

I've sold some AR projects to customers. Just product visualizations for trade fairs etc. nothing too fancy. It's cool and fun to work with but my takeaway was that for many practical applications it's pretty cost/time intensive to work on the non-technical parts of the projects (tuning the models and scenes so they look photorealistic). The devil is in the details. For example our typical projects involved getting CAD data from customers and turning it into a fairly simple AR-App. Most of the time was spent on fixing the models in various ways (and that's fairly hard to automate imo). UX is also really tricky. Reminds me a bit of the cost of working with data vs. the actual models in ML. The models and tuning them and playing with them is the cool stuff but most of the competitive advantages (imo) come from mastering the data part. I learned a lot from my ventures into AR and it was fun and I've kept a mindset of pretending there was an AR device in contact-lense form to brainstorm what one could do with this breakthrough every now and then. But there's really not much beyond "novelty effect" and gaming I see as interesting products today. Even in areas where AR solutions are deployed in industry (warehousing and logistics), I feel that less glamorous solutions would provide similar benefits. All that being said, long term I'm quite optimistic that the technology will bring very interesting changes and things I cannot imagine right now.