I think that at some point AI will enter the domain of repetitive physical jobs, like for example an experienced butcher will train an AI arm that eventually will learn to slice meat correctly since that's an activity that normally works with a reasonably standardized product(the animal carcass).

Keeping that in mind, honestly for me it's hard not to be very skeptical that it would make any economic sense in the next many decades to have an expensive humanoid robot performing tasks with a very low aggregated value like moving boxes one by one. Just because it's a robot performing an economic activity it doesn't necessarily mean it will be productive or competitive.