Amazon’s planned pilot of Agility’s Digit puts a humanoid robot into a demanding real-world test: can a machine that walks on two legs contribute to warehouse work at the scale the company needs? The trial could offer clues about where bipedal robots fit, while leaving open whether Digit will become part of Amazon’s broader operations.
A trial signals interest, not a rollout
Amazon announced the test at its Delivering the Future event in Seattle. The company has already invested in Agility through its $1 billion Industrial Innovation Fund, which named the startup among its first five recipients in April of last year. That investment points to interest in the technology, but it does not guarantee adoption.
Amazon Robotics chief technologist Tye Brady described the fund as a way to explore what is possible and examine practical examples. He said the company is interested in walking robots and wants to learn what works and what does not. A pilot gives Amazon a chance to gather that evidence before deciding whether the robot belongs in its operations.
Early partnerships can also change direction. Agility previously piloted work with Ford while exploring last-mile delivery, then shifted its focus for Digit toward warehouses and factories. That history is a reminder that a test can help a company clarify where a product makes sense without leading to a permanent deployment.
Amazon measures automation by its effect on work
Amazon’s robotics program has largely centered on wheeled autonomous mobile robots, or AMRs. Its 2012 purchase of Kiva Systems helped establish the foundation for that work, and the company now has 750,000 AMRs across its warehouse network. Amazon has also developed other systems, including the Sparrow picking arm.
Those machines reflect a practical approach to automation: a system has to improve productivity to earn a place in the network. The goal is not simply to add a novel machine, but to find ways to move goods to customers more quickly and improve safety for employees.
That bar matters for a humanoid robot because a successful demonstration in one setting does not settle whether it can work across a large operation. Amazon would need to understand how Digit performs alongside its existing equipment and workflows, and whether the benefits justify integrating a different form of mobility.
Human-shaped spaces create a possible opening
Many warehouses and factories were built for people. Shelves, aisles, stairs and uneven terrain reflect human use, which can make them difficult environments for some automation systems. A robot with legs could, in principle, move through spaces designed around people without requiring every facility to be rebuilt for a particular machine.
This potential may be especially relevant in brownfield sites: existing workplaces that were not designed around a specific automation system. A humanoid robot could offer flexibility there, though the source of that appeal is also a challenge. Walking and handling objects in varied environments must prove useful in practice, not just seem well suited to the space.
Amazon can build facilities around its own needs, so the constraints of existing buildings may matter less to it than to other companies. Even so, equipment that fits into current workflows with little friction would be valuable. The pilot can help reveal whether Digit offers that kind of fit.
Digit is one path to mobile manipulation
Brady said Digit is not the only way Amazon might combine mobility and object handling. The company already works with mobile robots and robotic arms, so another route could be to bring those capabilities together. A wheeled base carrying an arm, for example, could also act as a mobile manipulator.
That broader possibility puts the Digit pilot in context. Amazon is examining how movement, sensing, computing and the ability to handle objects might work together. The question is not only whether a robot looks human, but whether its combination of capabilities helps with identification, manipulation or sortation.
Other companies, including 1X, Figure and Tesla, are also pursuing humanoid robots. Agility has a head start and funding, and its new factory in Salem, Oregon, is claimed to be capable of producing up to 100,000 Digits a year once fully online. Production capacity, however, does not by itself show that robots can perform useful work at scale.
If Digit proves effective in Amazon’s pilot, it could strengthen interest in humanoid workers more broadly. If it does not fit the company’s workflows or is not ready for Amazon’s scale, that would still leave open other approaches to mobile manipulation and bipedal robotics. The trial’s value lies in the practical evidence it can provide about what these machines can do.