awetonomousResearchSubscribe
Menu

HUMANOIDS & ACTION SYSTEMS

Read the task behind the humanoid headline.

What manufacturer and customer reports establish, what remains uncertain, and how to frame an industrial evaluation.

AWETONOMOUS EDITORIAL · PUBLISHED November 3, 2026 · REVIEWED November 3, 2026 · SOURCE REVIEW / ORIGINAL ANALYSIS

A human-shaped robot can make a compelling demonstration. For industrial research, the useful unit of evidence is smaller: a named robot generation performing a named task in a described environment, with its assistance and limitations disclosed. A successful task is not proof of a general-purpose worker.

Two documented examples, with different evidence

CUSTOMER REPORT / JUNE 25, 2026

BMW and Figure

BMW describes Figure 02 placing sheet-metal parts in its Spartanburg body shop during 2025, and a subsequent Figure 03 project focused on logistics sequencing. The later use case involves organizing components into a trolley for onward transport. This is a customer account of specific tasks and a new project, not an independent evaluation by Awetonomous.

Keep the generations and jobs separate. Evidence from Figure 02's earlier assignment should not silently become a performance claim for Figure 03 or for an unrelated manufacturing operation.

VENDOR UPDATE / OCTOBER 26, 2026

Agility Digit and GXO

Agility reports Digit material-handling work at GXO and an agreement to expand to two additional US facilities. The announced start for those additional deployments is early 2027. As of this review, that expansion is a future plan; it should not be counted as two additional operating deployments.

The announcement supports a distinction between an existing reported application and a planned rollout. It does not establish that another site's objects, workflow, staffing, or economics will match.

Use an evidence ladder

  1. Demonstration: a selected task is shown. Ask about editing, resets, assistance, and the environment.
  2. Pilot: a bounded evaluation is described. Ask for scope, duration, acceptance criteria, and what happened afterward.
  3. Reported deployment: a customer or supplier says the system is operating. Identify the task, site, robot generation, and reporting period.
  4. Transferable evidence: documented results address the conditions that matter to your own application. This requires more than a purchase announcement.

This ladder is our editorial framework, not an industry certification. A paid agreement is evidence of a commercial relationship; it is not automatically a utilization figure or independently measured return on investment.

Questions for an industrial evaluation

Request a clear account of remote assistance, human interventions, task changes, charging, recovery, maintenance, and the work left to surrounding equipment. Ask what software and hardware configuration produced the reported result. Include integration with production IT, material flow, and the responsible safety team before discussing scale.

Our scope includes humanoid and mobile-manipulation systems when they address real physical work. It does not rank general intelligence or predict when robots will replace a workforce. We have not tested these machines, verified customer economics, or established availability for a particular country or buyer.

Sources & evidence

Official sources reviewed November 3, 2026. Vendor materials establish descriptions, not independently measured performance. Original evaluation criteria are our analysis.

  1. BMW: Figure 03 project and earlier Figure 02 task ↗
  2. Agility Robotics: October 26, 2026 commercial update ↗
Source review and original analysis. No independent equipment or site trial was conducted for this guide. Editorial standards ↗