If this is designed to replace shop floor staff, it will fail. The tests will highlight that app adoption will be low and behaviour to ask an employee (if one is visible) will be a shopper's first reaction. Why do I say this? It's easier and faster to talk than reach for phone, open an app, type, read the response etc. It's the staff that need a handheld oracle so that they can accurately and promptly answer these questions.
Knowing about stock availability has nothing to do with AI. A good inventory system (RFID) that's properly integrated does that.
Sharing my bias (and full disclosure), I have a business in RetailTech, personalisation software that matches clothes to shoppers. Importantly, although it's an application, it's designed to work on a tablet device, IN THE HANDS OF THE SALES ASSISTANT.
I'm certain the future will prove that those retailers who invest in their store teams — by giving them great tech to give fantastic service and efficiency — will be the winners. AI can certainly play a big role in this, though I see greater benefits in use throughout the supply chain. Sadly, this AI test on the shop floor suggests it's driven by money saving (less staff) as opposed to genuine shopper benefit. The value is actually helping the customer so they buy (and more). It also triggers loyalty and the shopper returning.
I don't think a DIY app for shoppers is the answer. Particularly if they're hunting for Pokemon at the time... ;0o