Proteus gains natural language, Europe expansion underway

Proteus can now hear your warehouse commands. The updated AMR is designed to pick up more tasks by understanding natural language, and Amazon plans to roll the upgraded robot into Europe next year.

The company reports that Proteus has been empowered with a natural language interface that lets workers tell it what to do without special programming. Scott Dresser, vice president of Amazon Robotics, described the shift as turning Proteus into “your assistant for material movement,” capable of judging priority, choosing routes, and timing based on simple spoken instructions. The upgrade is paired with broader fleet enhancements, as Amazon also moves to extend its Vulcan picking robot and STARK tote-handling system across Europe.

Proteus remains a mobile cart mover built to relieve strenuous manual tasks in busy fulfillment environments. It is designed to handle carts weighing nearly 400 kilograms, or 881.8 pounds, enabling it to shuttle heavy loads between docks, workstations, and delivery points. The approach echoes Amazon’s longer term push to scale human-robot collaboration in high-volume facilities, a strategy that traces back to the 2012 acquisition of Kiva Systems and the 2022 introduction of Proteus itself.

Today, Proteus operates at 24 e commerce fulfillment centers in the United States, a footprint Amazon says underscores the robot’s utility in real world workflows. With the Europe expansion, the company is piloting the updated Proteus in its laboratories as a testing ground before a production deployment in new markets. The first half of next year is the target window for Europe, according to Amazon, signaling a measured roll out rather than a sudden push.

One practical implication of giving Proteus a language interface is the potential to reduce onboarding friction for frontline staff. The absence of a requirement for special programming means workers can directly articulate tasks, priorities, and timing. Yet this approach places greater emphasis on system reliability in interpretation and conflict resolution. In the field, a misheard instruction or ambiguous command could ripple through a lane with multiple AMRs and human teammates, so clear fail safes and robust prioritization logic become critical. Industry observers will want to see how Proteus behaves when competing requests arrive simultaneously, for example a high-priority replenishment versus a quality-control check at a distant station.

Another meaningful dimension is integration with the broader European deployment. Europe introduces a more multilingual and distributed set of sites, which can complicate natural language parsing unless the system is tuned to local workflows and terminology. The lab pilots will be an early signal of how well the interface scales beyond United States centers and across different operational contexts. The expansion also coincides with ongoing use of other AMRs in the European fleet, suggesting Amazon intends to converge disparate robot capabilities toward a more cohesive material-handling platform.

From a practitioner standpoint, two to four concrete takeaways stand out. First, the natural-language capability lowers the barrier to reconfiguring workflows on the fly, but teams should still expect a learning curve as the system learns site-specific vocabulary and routines. Second, relying on language as a control channel increases the need for robust safety and override protocols to prevent misinterpretation from disrupting operations. Third, the payload constraint of roughly 400 kg is a meaningful design constraint that shapes what tasks Proteus is best suited for in the European labs and facilities. Fourth, the Europe rollout will reveal how well Proteus interoperates with other fleet components like Vulcan and STARK, and whether command-level simplicity can translate into reliable, scalable productivity gains in new regulatory environments.

In short, Proteus’s new natural-language capability marks a concrete, engineering-driven step forward. It moves the system from preprogrammed task execution toward more autonomous, human-guided decision making, while Europe tests the limits of that approach at scale.