I spent years in Amazon Fulfillment Tech, a lot of in the Picking software space. It's all about tradeoffs, and shipping costs.
Amazon will spend far more money getting the item from the warehouse to your front door than they will on labour costs getting the item into the box, ready to ship to you. Shipping is expensive, labour is relatively cheap when each item is touched by maybe 2-4 people for a few seconds each.
Robotics are getting better, but the typical storage bin in an Amazon warehouse is optimized for density, not for easy picking. Humans do a great job of picking items despite the storage density. We fish it out, feel around, use our incredible vision powers. We're great at it! Robots suck at that sort of thing (today).
Amazon could build a warehouse that is optimized for robotic picking. But if you built it with today's robots, you would have a fraction of the item storage density. That means you'd either need many times more warehouses or you'd need to reduce your selection. When you reduce selection, that means a different warehouse is handling those other orders, and it's probably further away from the customer.
Further away from the customer => higher shipping costs.