A distribution center in the UK has already mastered this:
> The Ocado Andover CFC is a three-storey grid of storage crates containing groceries. On top of the grid a fleet of 1,100 robots grab items and deliver them to “pick stations” where “personal shoppers” assemble orders for delivery. According to Ocado, an order of 50 items takes five minutes to pick and pack.
http://www.canadiangrocer.com/top-stories/a-sneak-peak-at-th...
https://www.youtube.com/watch?v=4DKrcpa8Z_E
Google "automated grocery distribution", there's plenty of examples.
This is essentially unskilled labour moving boxes from one point to another. The exception is vegetables/fruit where a human discards the bad stuff before it goes into another box for delivery on a truck. 95% of that process could be automated. Including eventually the delivery drivers in between the distribution center <-> retail endpoint pickup centers (where consumers and last-mile Instacart/delivery drivers can go in person for pickup - I'm assuming humans will be needed for home delivery for a long time).
Instacarts biggest drawback is the cost and how they nickel-and-dime you, but GroceryGateway is run by the grocery store itself, and it's much better and actually helps you save money by shows you all the deals in the interface (imagine sorting a grocery store by discounts and price, as you walk through it gets more expensive). There is very little markup vs the stores, I've found it to be cheaper in practice, especially factoring in my time.
A fully integrated/automated version of this with far less urban real estate + labour costs will bring the prices down significantly.
I have a feeling only smaller luxury/farmers market/vegetable+fruit+meat only (+ semi-automated pantry section) shops will exist in the future. You don't really need to pick up most stuff by hand, but an in-person browsing UX can still be useful. Ikea stores are halfway to figuring this out.