On the Shop Floor and in the Care Home: A 2025–26 Field Report on Live Humanoid Deployments

On the Shop Floor and in the Care Home: A 2025–26 Field Report on Live Humanoid Deployments This field report synthesizes company disclosures, media reporting and on‑the‑ground customer accounts to document verified pilots, commercial trials and early deployments of humanoid robots across warehouses, factories, healthcare/eldercare and select consumer experiments. The investigation foregrounds real‑world use cases, measurable KPIs where disclosed, pilot scale and duration, safety mitigations, and the observable gaps separating polished demos from operational reality. Where humanoids are being tried — use cases and industries Humanoids are appearing in four broad settings: Warehouses & logistics: picking, kitting, pallet handling and short‑haul intra‑warehouse transfers. Factories & light manufacturing: repetitive assembly, machine tending, parts feeding and collaborative tasks alongside humans. Healthcare & eldercare: patient transfer assistance, therapy/rehabilitation aides, monitoring and limited companionship functions. Consumer & campus pilots: home experiments, retail greeters and campus couriers for low‑risk tasks. Warehouse & logistics — Agility Robotics and low‑cost campus players Agility Robotics’ Digit is the most publicly visible humanoid in logistics trials. Company disclosures and media coverage indicate trials with third‑party logistics (3PL) providers and parcel/fulfillment operations focused on order kitting, box manipulation and short sequence pick flows. Vendors and operators typically trial Digit for: Unpacking incoming pallets and placing cartons onto conveyors. Piece‑pick support for small‑item kitting where human dexterity is required. Transporting totes across short warehouse aisles as a "last‑mile" intra‑facility mover. Companies market Digit as a flexible, dexterous alternative to fixed automation — specifically for SKUs and micro‑fulfillment nodes where reconfigurable handling is more economical than conveyors or custom end‑of‑arm tooling. Agility has emphasized Robot‑as‑a‑Service (RaaS) pilots to lower entry costs for 3PLs. Unitree and UBTECH are positioned at the lower end of the price/feature spectrum. UBTECH’s social/hospitality humanoids and Unitree’s small logistics platforms (mostly non‑humanoid quadrupeds from Unitree and simple biped prototypes in UBTECH’s catalog) have surfaced in retail greeter pilots and campus logistics tests where cost and appearance matter more than heavy payloads. Factory & assembly trials — Figure, Tesla, Apptronik, 1X Figure has publicly positioned its industrial humanoid platforms for machine‑tending and repetitive line assembly. The company markets for small‑batch and reconfigurable production where traditional fixed automation is uneconomic. Early industrial trials focus on: Repetitive bolt‑fastening, cable routing and small‑parts assembly in electronics and appliance lines. Machine tending (loading/unloading CNC presses) where the robot can be fenced or work in a collaborative cell. Tesla’s Optimus is marketed internally as a future factory assistant. Reporting in mainstream outlets during 2023–25 documented Tesla prototypes being trialed at select Tesla lines for non‑critical tasks (parts ferrying, test jig handling). Tesla’s public positioning emphasizes long‑term labor substitution and productivity gains rather than immediate commercial sales. Apptronik and 1X Technologies approach collaborative manufacturing and material handling. Apptronik markets its humanoid and humanoid‑adjacent manipulators for "co‑bot" roles — assisting human assemblers, reducing ergonomic injuries and taking on repetitive sub‑tasks. 1X focuses on material flow and modular gripping solutions to speed kitting operations. Healthcare & eldercare — Sanctuary, Fourier and rehabilitation pilots Sanctuary AI and Fourier Intelligence have been explicit about healthcare pilots. Sanctuary has presented prototypes and company‑stated trials aimed at patient transfers and limited ADL (activities of daily living) assistance; Fourier, primarily a rehabilitation/exoskeleton company, has run trials combining humanoid interfaces and rehab hardware for guided therapy. Typical trial objectives: Reducing caregiver strain during transfers and simple mobility assistance. Standardizing repetitive physiotherapy tasks for outpatient or in‑home rehab. Continuous monitoring and fall detection in assisted‑living settings (companionship as a secondary feature). Vendors emphasize safety certification, force‑limited actuation and redundant sensors. Early pilots routinely limit humanoids to non‑weight‑bearing roles or supervise all transfers with a human present. Consumer & brand pilots — Xiaomi and others Xiaomi’s CyberOne is primarily a consumer/brand halo product. Public demonstrations and controlled home experiments have been used to market Xiaomi’s capability in AI and human‑robot interaction rather than to sell large fleets for home care. These pilots are short, promotional and aim to collect data on human acceptance and natural‑language interactions. Pilots, KPIs, safety and the gap to production Across the documented pilots the most commonly reported KPIs were uptime, task throughput and error rate. However, independent verification of vendor claims remains sparse and typically shows a material gap between lab numbers and field performance: Uptime: Company‑stated uptimes in pilot reports often exceed 80–90% during supervised sessions; in unsupervised runs operators report frequent recharges and downtime for recoveries and sensor recalibration. Throughput & cycle time: Throughput for humanoid picks is generally 30–60% of a trained human picker for unstructured small‑item picks; throughput improves in constrained/structured tasks such as kitting with fixed trays. Error rates: Pick/place error rates are lower in structured tasks ( Reported pilot durations typically range from 3–12 months. Safety mitigations include geofencing, force limits, supervised mode only for patient transfer trials, and integration with existing safety rails. There are few public reports of serious incidents; instead, operators routinely report cautious deployment scopes and multiple human‑in‑the‑loop constraints. Pricing, business models and ROI Pricing disclosures are thin. Common patterns observed: Vendors offer RaaS/subscription models to reduce upfront capital burden for pilots; some industrial players also sell units outright to strategic customers. Estimated unit economics vary dramatically: enterprise humanoids aimed at factories are positioned in the high five‑ to six‑figure range per robot when sold up front, while low‑cost social humanoids and campus couriers from mass‑market vendors sit in the low five‑figure or below — often accompanied by cloud service fees. Reported ROI is task‑dependent: pilots claim payback in scenarios with chronic labor shortages or high turnover (e.g., 12–36 months cited in vendor materials) but independent observers stress the additional costs of integration with WMS/MES, supervision and downtime. Competitive landscape and positioning Vendors fall into three positioning buckets: Industrial first (Figure, Apptronik): tooling, repeatability and integration with MES; selling to OEMs and system integrators. Logistics & flexible automation (Agility): targeted at 3PLs and micro‑fulfillment where reconfigurability beats fixed automation. Lower‑cost/mass market (Unitree, UBTECH, Xiaomi): primarily for promotional deployments, campus logistics and consumer experiments where affordability and user perception are as important as payload. Competition hinges on sensor quality, end‑effector dexterity, battery life and the ability to integrate with enterprise systems. The biggest observed repellents to rapid scaling are battery life, difficulty integrating into WMS/MES workflows, and ongoing supervision burden — all of which raise effective cost‑per‑task versus vendor pitch. Comparative appendix (selected, company‑stated vs independent observation) Vendor / Robot Pilot focus Pilot scale & duration Pricing / model Vendor claim Independent observation Agility Robotics — Digit Order kitting, tote handling, intra‑warehouse transport Multiple 3–12 month RaaS pilots (company‑disclosed) RaaS pilots; unit sale not widely disclosed Flexible replacement for specific pick/transport tasks Works in structured flows; supervision and recharging limit unsupervised throughput Figure Machine tending, repetitive assembly Industrial trials with OEMs; multi‑month Enterprise pricing (not disclosed) Competitive with fixed automation for small batches Promising in constrained cells; full MES integration still heavy lift Tesla — Optimus Internal factory assistance, parts ferrying Internal pilots reported 2023–25 Internal program; no commercial pricing Reduce factory labor needs long term Prototypes used for non‑critical tasks; generalizability unproven Apptronik / 1X Collaborative manufacturing, material handling System integration pilots 3–9 months RaaS and integrator models Co‑robot for ergonomic relief Effective for sub‑tasks; ROI dependent on reduced injury/turnover Sanctuary / Fourier Patient transfer, rehab assistance Controlled clinical pilots; short term Trial and partnership models Improve caregiver safety and standardized therapy Limited to supervised use; strict safety regimes required Xiaomi — CyberOne Consumer interaction, home experiments Short demonstrations and controlled home trials Consumer halo product; commercial availability limited Advanced human‑robot interaction for consumers Primarily marketing/data collection; not a mass deployment Conclusion — why industries are adopting humanoids now Adoption is concentrated where humanoids’ humanlike reach and generality solve problems fixed automation cannot: logistics nodes with high SKU variability, small‑batch manufacturing where retooling is costly, and care settings where lifting/monitoring is ergonomically risky. However, pilot evidence from 2025–26 shows that many deployments remain guarded trials: companies emphasize RaaS, constrained task envelopes, and supervised operation. The short‑term winners will be vendors who can (a) integrate reliably into enterprise IT and safety systems, (b) lower ongoing supervision and maintenance costs, and (c) demonstrate clear, auditable ROI in environments with chronic labor shortage or injury risk. Note: This report synthesizes public company disclosures, press reporting and operator accounts. Many vendor claims are company‑stated and independent verification is limited in some pilots; where possible the article highlights those limits.