Robots made picking faster. Faster made the warehouse floor more dangerous.
Amazon Β· 2019βongoing
What they did
Amazon's fulfillment centers pair AI-driven inventory robots (the Kiva/Amazon Robotics system) with human pickers and packers working against algorithmically-set productivity rates, aiming for speed and accuracy no manual-only warehouse could match.
What happened
Multiple independent investigations β Reveal, the Strategic Organizing Center, and the Washington Post β found Amazon's robotic facilities had a serious-injury rate 54% higher than its non-robotic ones in 2019 (7.9 vs. roughly 5.1 per 100 workers), and more than double the warehousing-industry average. A December 2024 U.S. Senate investigation concluded Amazon knowingly avoided safety fixes that would have slowed throughput; the U.S. Attorney's Office (SDNY) is separately investigating whether Amazon concealed its true injury rates.
The so-what
The robots weren't the injury cause on their own β the algorithmically-optimized pace they enabled was. "AI made the process faster" is not the same claim as "AI made the process better for the humans still in the loop," and this is the case study that shows the gap between the two most starkly.