Humanoid trials in logistics and retail: what GXO, Amazon and others have documented

Only a few logistics and retail companies have publicly documented humanoid robot work. The best-documented is GXO, which has run Agility Robotics' Digit since 2024 and in October 2026 said it would add two more US sites. Amazon has tested Digit, and Mercado Libre signed an agreement in December 2025. Every published task is simple: moving plastic totes.

This article collects what these companies and their robot suppliers have published, separates figures from plans, and explains what the record says about how mature humanoids are outside staged demonstrations. Nearly all the numbers come from the companies themselves. None of them has been independently audited.

GXO: from proof of concept to a multi-year contract

GXO, a contract logistics operator, is the most visible early user. In its June 2024 announcement of a multi-year agreement with Agility Robotics, the companies said Digit would work at a facility GXO runs for the apparel brand SPANX. The task was narrow. Digit takes totes from collaborative mobile robots and puts them onto conveyors.

The release said this followed a proof-of-concept pilot in late 2023. It also described the deal as a Robots-as-a-Service arrangement, meaning GXO pays for the robots as a service rather than buying them. Agility called it the first humanoid deployment of this kind, which is the company's own claim. No robot count or price was published.

On 26 October 2026, GXO and Agility Robotics announced an expanded partnership. Digit is due to go into GXO sites in Statesville and Lebanon, with deployments expected to begin in early 2027. The release said Digit has moved more than 100,000 totes at GXO's Flowery Branch site in Georgia since the 2024 agreement began. It also said Digit robots have logged more than 65,000 hours of real-world operation across Agility's commercial work. The new sites will start on the current Digit 4 and move to Digit 5 once that version is available.

What the GXO numbers do and do not tell you

One hundred thousand totes over about two years is a real operating record, but it is not a throughput figure. The releases do not say how many robots did the work, how many shifts they ran, or how many totes a human worker handles in the same time. The 65,000-hour figure covers all Digit deployments, not GXO alone.

There is also an inconsistency worth noting. In December 2025, Agility already described Digit as having moved more than 100,000 totes in live commerce operations. The October 2026 release uses the same headline figure for one GXO site. The two statements may cover different scopes, or the public figure may simply not have been updated. Either way, readers should treat these as rounded marketing milestones, not operating data.

Amazon: testing, not a rollout

Amazon's interest in humanoids is often overstated. Its newsroom article introducing the Sequoia system and Digit said Amazon would test Digit at a research and development facility south of Seattle. The first use case was tote recycling: picking up empty totes once all the items in them have been picked, and moving them on. Amazon said it sees an opportunity in mobile robots that can grasp and handle items. It described the work as exploratory and did not announce a fleet or a site-wide deployment.

Amazon also invests in Agility through its Industrial Innovation Fund, according to the same article. Agility continues to list Amazon among its Digit customers and investors. Neither company has published tote counts, hours or results from the Amazon tests in the material we reviewed. So the most accurate summary is that Amazon has tested Digit on one task. There is no public evidence of wider adoption.

Mercado Libre: an e-commerce agreement starting in Texas

Latin American e-commerce group Mercado Libre announced a commercial agreement with Agility Robotics on 10 December 2025. The first deployment is in San Antonio, Texas, with expansion into Mercado Libre's Latin American warehouses described as a future step. The stated tasks are again repetitive, physically demanding material handling, mainly moving totes.

The release frames the project around jobs that are hard to staff and around ergonomics. It gives no robot numbers, schedule beyond the initial site, or performance targets. It is best read as the start of a deployment, not a result.

Figure: logistics throughput, but no named customer

Figure AI published some of the most specific speed figures in this field. In a June 2025 post on its Helix model in logistics, Figure said its robot handled a package in about 4.05 seconds, down from about 5.0 seconds. It said the robot scanned barcodes successfully about 95% of the time, up from around 70%. Figure also reported that increasing demonstration data from 10 to 60 hours improved both speed and barcode success.

Two cautions apply. The post is more than 12 months old, and it describes work "in a logistics environment" without naming a customer or site. Figure itself said the model was still in a low-data regime. These are useful benchmarks for one package-handling task, not evidence of a commercial logistics contract. For Figure's better-documented factory work at BMW, see our overview of humanoid pilots in car plants.

The limits the industry admits

Some of the most useful information comes from the cautions, not the milestones. In February 2026, TechCrunch reported that Toyota had contracted seven Digit robots for a Canadian factory after a year-long pilot. The task was unloading totes of parts from automated tuggers. The same article made three points that apply equally to warehouses:

• Humanoids strong enough for heavy lifting are, in the article's words, still too unreliable to work autonomously around people.
• Agility's chief technology officer said the cost of deploying a robot can be far higher than the price of the robot itself. Agility says it uses AI tools to cut deployment time and configuration work.
• Real humanoid deployments in industrial settings remain rare. Seven robots at one plant counts as notable news.

That matches the pattern above. Every documented logistics task involves standard containers, fixed pick-up and drop-off points, and very little interaction with the public.

Reading the commercial claims

Agility, the supplier behind most of these pilots, has strong reasons to publish good news. In an October 2026 update on commercial momentum, it reported more than $300 million in committed multi-year orders for Digit 5 and more than 30 prospective customers in discussions. It said early customer access to Digit 5 is planned for the first half of 2027, with general availability by the end of 2027. Its press page shows the company completing a business combination with Churchill Capital Corp XI on 4 November 2026, a route to a public stock listing.

Orders, prospects and target dates are not deployed robots. They are worth tracking, but the measurable record so far is smaller: a handful of sites, one main task, and cumulative tote and hour counts without a baseline to compare against. For the wider picture, see what Agility has reported in our article on Digit in warehouses and the Agility Digit catalogue entry.

What this means for event organisers

Warehouse pilots and event bookings are different jobs, but the logistics record is a useful reality check.

• Expect narrow, rehearsed tasks. The best-documented humanoid in commercial use moves totes between fixed points. A robot at an event will also work best with a short, well-defined routine.
• Treat autonomy claims with care. If industrial humanoids are not yet trusted to work autonomously among workers, a robot in a crowd needs operators, barriers and clear safety distances. Our safety planning guide for public events covers this, and teleoperated or autonomous? explains what audiences actually see.
• Ask what a figure measures. When a supplier quotes hours or tasks completed, ask over what period, with how many robots and compared with what.
• Industrial models are not event models. Digit is built for warehouse work, not stage appearances. To see which humanoids are offered for events today, read which humanoid robots you can actually hire.

Frequently asked questions

Which logistics companies have publicly used humanoid robots?

The best-documented case is GXO, which has run Agility Robotics' Digit since a 2024 multi-year agreement and announced two more US sites in October 2026. Amazon has tested Digit at a research facility near Seattle, and Mercado Libre signed an agreement in December 2025 starting in San Antonio, Texas. Figure AI has published logistics speed figures without naming a customer.

What tasks do humanoid robots do in warehouses?

In the deployments that have been documented, mostly moving plastic totes. At GXO, Digit takes totes from mobile robots and places them on conveyors. Amazon's first test task was recycling empty totes. Figure has shown package handling and barcode scanning. These are repetitive tasks with standard containers and fixed pick-up and drop-off points, not general-purpose warehouse work.

How productive are humanoid robots in logistics so far?

Public data is limited. Agility says Digit moved more than 100,000 totes at one GXO site and has logged more than 65,000 operating hours across its deployments, but gives no robots-per-shift or comparison with human workers. Figure reported about 4.05 seconds per package in one logistics task. All figures are self-reported by the companies and have not been independently audited.

Are warehouse humanoid robots safe to work next to people?

Not fully yet. In February 2026, TechCrunch reported that humanoids strong enough for heavy lifting are still too unreliable to operate autonomously around people. The documented deployments involve fixed, repetitive tasks rather than open, crowded spaces. Agility says its next-generation Digit is being designed for closer work alongside humans, but that version is not yet generally available.

How this article was made: written by the humanoid4rent editorial team with AI assistance; every factual statement was checked against the sources listed below on 2026-11-05. Spotted an error? Write to info@humanoid4rent.com.