Agility Robotics Digit in Amazon warehouses:

Agility Robotics deployed Digit robots in Amazon warehouses and the deployment isn’t a pilot. It’s operational.

The robots pick up yellow totes from conveyor belts and carry them to shelving units. Over and over. The task is simple. The task is also the exact task that millions of warehouse workers do every day.

I watched the deployment video three times. What struck me wasn’t the robot’s capability (it’s doing basic material handling, not surgery). What struck me was the ordinariness of it. Digit walks up to a conveyor. Picks up a tote. Walks to a shelf. Places the tote. Walks back. The rhythm is industrial. The movements are functional. There’s no flair. No demo polish. Just work.

That ordinariness is the point.

The economics

A warehouse worker in the United States earns roughly $15-20 per hour. With benefits, overtime, and turnover costs (warehouse work has among the highest turnover rates of any industry), the fully loaded cost is higher. Warehouses operate in shifts: day, night, weekend. Staffing all shifts is a perpetual challenge.

Digit works 20 hours per day (with 4 hours for charging and maintenance). No benefits. No overtime. No turnover. No shift scheduling headaches. The robot doesn’t call in sick. It doesn’t quit after three months because the work is physically exhausting.

The cost comparison isn’t straightforward because the robots are expensive upfront and require maintenance and software updates. But the trajectory is clear. As the robots get cheaper and more capable, the economic case for human warehouse labor weakens.

This isn’t a prediction about the distant future. It’s an observation about a deployment that’s happening right now, in a warehouse that ships your Amazon packages.

What Digit can and can’t do

Can: walk on flat surfaces, pick up rigid objects of standard size, navigate predefined routes, avoid obstacles, operate for 20 hours with one charge cycle.

Can’t: pick up irregularly shaped objects reliably, handle fragile items, work in cluttered environments, improvise when the layout changes, communicate with human coworkers naturally.

The “can” list covers about 30-40% of warehouse tasks. The “can’t” list covers the rest. That ratio will shift over time as the robots improve, but right now, Digit is useful for a specific subset of tasks, not all warehouse work.

That’s fine. That’s how deployment works. You start with the tasks the robot can do reliably. You expand the task list as capability improves. You don’t wait for the robot to do everything before deploying it for something.

What this means for workers

I keep going back and forth on this. The optimistic view: the warehouse jobs being automated are the hardest, most physically demanding, highest-turnover positions. Nobody aspires to carry totes for 10 hours. Robots taking these jobs frees humans for work that’s more creative, less physically punishing, and better compensated.

The realistic view: the people doing these jobs need these jobs. They’re not being “freed.” They’re being displaced. The transition from “carry totes” to “better job” isn’t automatic. It requires retraining, education, support systems that may not exist.

Both views are true simultaneously. The automation improves efficiency. The displacement creates hardship. These aren’t contradictory. They’re two aspects of the same change.

I don’t think the solution is to stop deploying robots. The economics won’t allow that. I think the solution is to plan for the displacement with the same rigor that we plan for the deployment.

We’re very good at planning deployments. We’re historically bad at planning for the humans affected by them.

Digit is in the warehouse. The totes are moving. The economics are working. The harder question, what happens to the people who used to carry the totes, is one that the technology can’t answer. That one’s on us.

Related thinking: