AI Resilience Report forPostal Service Mail Sorters, Processors, and Processing Machine Operators

At the same time, the Postal Service is adding new kinds of robots to help people. For example, USPS now uses “flex rover” robots: a worker scans a package and places it on the rover, and the rover drives the mail to the right bin ^[2]. Each rover can carry hundreds of packages an hour, easing the workers’ load.

Tasks that are still very hands-on – like dumping heavy sacks of mail onto conveyors – remain mostly done by people. (Though researchers are already testing AI-guided robots for this, such as robotic arms that pick parcels from a pile and put them on a belt ^[3].) In short, most of the scanning and sorting work is already done by machines, with people supervising, handling special cases or doing the very physical jobs ^{[1] [2]}.

However, heavy automation also faces costs and challenges. Buying and installing machines is expensive, and they don’t always plug in smoothly. An audit found that almost half of the new driverless carts were hardly used, because plant layouts and high mail volumes made them hard to use ^[4].

This shows that training staff and setting up systems well is as important as the hardware. Labor concerns also play a role: workers’ unions have protested cuts to sorting jobs, and in 2020 the postal workers’ union even filed a grievance when hundreds of sorters were planned to be removed ^[1]. Socially and legally, people expect reliable mail service too, so any AI must maintain quality.

Overall, it seems likely that mail-processing will keep adding AI and robots where they save time and money. But because mail delivery must stay quick and accurate for everyone, and many tasks still need a human touch (like solving tricky errors or doing maintenance), adoption will probably be gradual. The good news is that even as machines handle more of the routine work, skilled postal workers will still be needed to oversee the systems, fix problems, and serve customers – roles that are hard to automate ^{[1] [2]}.