For years, humanoid robots were demo videos. In June 2024, that changed. Logistics company GXO and Agility Robotics signed a multi-year agreement to deploy Agility's two-legged Digit robot in GXO warehouses, which the companies called the industry's first formal commercial deployment of humanoid robots.
What Digit actually does
At a GXO facility near Atlanta that serves the apparel brand SPANX, Digit moves totes from mobile robots onto conveyors. It is repetitive, physical work that is hard to staff. GXO is also testing Apptronik's Apollo humanoid, and investors expect much more to come: Goldman Sachs recently projected about 6.5 million humanoid robots by 2035.
The detail that matters
GXO's chief automation officer said Digit would work "alongside our people." But The Robot Report noted that at the SPANX facility, Digit was not working with or near humans.
That gap is not a criticism of Agility. It is the honest state of the field. Today's humanoids are deployed where they can be separated from people, because working safely and smoothly beside people is far harder than moving a tote. Physical capability is arriving faster than social capability.
What working alongside people really takes
A human coworker does dozens of social things without thinking: steps aside for someone carrying a heavy box, waits for eye contact before handing something over, understands a pointed finger or a quick "over there," notices when someone is in a hurry. Each of these is a skill a robot has to learn before it can share a workspace instead of having its own cage.
Why it matters for Social AI
The value forecast for humanoid robots assumes they will work in human environments: warehouses with staff, stores with customers, homes with families. Getting there depends on data showing how people coordinate physically with each other and with machines, recorded from multiple angles with the timing intact.
Crosstalk sources human–human and human–robot interaction data, including collaborative physical tasks and egocentric video, with consent from everyone recorded. Explore physical AI datasets or read the paper.
Sources: GXO press release, The Robot Report, IoT World Today, Yahoo Finance on the Goldman Sachs forecast.