They're replacing some "meatbags", in the next decade. And they only add Kiva to new warehouses, they haven't been retrofitting. So they'll have "meatbags" for a while. If they can build a compelling warehouse assistant HUD they'll then sell it to others and then expand into larger markets like everything else they do?

If they can get something compelling for 5 years that saves 10% on labor, that's billions of dollars and established expertise in a major market.

Um.

As a programmer I am in the job-destruction business. I think a society with fewer jobs and more automation can be a better society for everyone provided we simultaneously shift in the direction of more of a social safety net for the unemployed and hopefully a UBI. But right now a lot of people are very frightened by that job destruction and they're not totally wrong to be. It is causing a lot of short-term suffering, and a lot of people blame us.

Let's not start calling the people we unemploy "meatbags", ok?

Ignoring the tone-deafness of "meatbag", human workers will continue to be part of Amazon's systems for a long time to come, so it makes sense to develop things for humans.

Robots are - and will continue to be in the foreseeable future - massively expensive, require long lead times, and slow to scale. In the current state of the US economy I can hire several thousand humans to do a job practically overnight, while buying/renting several thousand industrial robots requires both considerably more money and time.

Amazon's business is highly seasonal, which naturally favors options with low initial cost but high marginal cost (humans) rather than high initial cost and low marginal cost (robot). Near-total replacement of a high-variance business process with robots is likely impractical and will continue to be for a long time.