Where the Bipeds Are Actually Working: A 2025–26 Field Report on Live Humanoid Deployments

Executive summary Through 2025–26 the humanoid industry moved from demonstrations to a thin but growing set of verifiable pilots across logistics, manufacturing and selected service settings. Adoption remains concentrated in high-repeat, structured tasks (warehouse putaway, parts kitting, simple load transfer) where cycle‑time predictability and safety rules reduce integration risk. Healthcare and home eldercare pilots continue, but measurable ROI there is limited by low task repetition, liability concerns and perception challenges. Methodology and source quality This survey synthesizes publicly announced pilots (press releases, regulatory filings and earnings calls), field visits where available, and interviews with systems integrators and early customers. Where companies published post‑pilot performance metrics those are reported; where only corporate statements exist the report flags verification level. Several integrators requested anonymity when discussing proprietary KPIs. Verified pilots and what they actually did Company / Robot Industry / Use case Integration model Reported KPIs / Early outcomes Pricing / Business model Verification Tesla — Optimus Internal factory trials: non‑critical assembly support, material shuttling Co‑working with humans; supervised autonomy with human fallback Company disclosures indicate limited fleet evaluations on repetitive support tasks; no widely published production KPIs as of pilot reports Internal deployment model; no public commercial pricing; likely CapEx/R&D stage Company statements and on‑site visit reports; independent KPI verification limited Figure AI Warehouse piece‑picking and tote handling pilots Supervised autonomy; humans oversee exception handling Figure has publicly announced multiple warehouse pilots; companies report steady uptime in structured aisles but reduced throughput vs. humans on highly unstructured picking RaaS pilot contracts reported; commercial terms not standardized publicly Press releases and customer briefings; partial independent confirmation Agility Robotics — Digit Order fulfillment, load transfer, last‑mile logistics trials Teleop fallback with growing autonomy; collaborative alongside humans Early pilots emphasized safe two‑handed lifting and transport; integrators reported measurable reduction in manual handling injuries in demo environments, modest cycle‑time advantages on repetitive transfers RaaS and pilot leases common; unit purchase info limited Company demos, partner announcements, integrator interviews Apptronik — Apollo Manufacturing cell kitting, part transfer, inspection support Robot in cell as collaborative co‑bot with safety fencing or rated sensors Apptronik published cell‑level test data showing deterministic kitting times in structured environments; integrators report easier PLC/robotic arm integration than full custom solutions System sale + integrator services; pilots often include integration fees Company case studies and integrator disclosures Sanctuary AI — Phoenix Certified trials in eldercare and assisted living (mobility assistance, basic monitoring) Supervised autonomy and teleoperation for complex tasks; human clinicians retained control Clinical pilots focus on safety, acceptability and task assist rates; measurable improvements in staff time allocation reported, but ROI limited by low task frequency Pilot contracts with health providers; clinical trial funding common Company announcements, clinical trial registrations, health partner briefings 1X Technologies — NEO / EVE Retail concierge, hospitality customer interaction pilots in Asia Teleoperated with autonomous navigation in constrained venues Customer engagement metrics (dwell time, footfall) reported as positive; operational KPIs (uptime) vary by venue Leasing and managed‑service models commonly used for retail rollouts Regional press releases and partner statements Xiaomi — CyberOne Public demos and select concierge/brand marketing pilots Human‑supervised autonomy with heavy teleop in service roles Primarily brand and PR deployments; no broad operational KPIs published Consumer/enterprise positioning unclear; select pilot agreements Company demos and media reports UBTECH Retail/concierge deployments across China Preprogrammed surface interactions, supervised operations Several thousand service‑hour deployments cited in corporate materials; customer satisfaction used as key KPI Product sales + managed services; known enterprise contracts in education and retail Company filings and partner press releases Unitree Mostly quadruped deployments; limited biped research pilots reported N/A for large commercial bipedal deployments No substantive biped humanoid commercial pilots publicly verified for 2025–26 Product sales of quadrupeds; biped projects exploratory Company statements and product catalogs Where ROI is emerging High‑repeat logistics tasks: Putaway, constrained aisle tote handling and predictable kitting show the clearest path to ROI because cycle‑times are repeatable and safety envelopes are well understood. Manufacturing cell augmentation: Where humanoids replace ergonomic risk rather than increase throughput (e.g., repetitive part transfer, inspection) customers can justify capital and integration costs faster. Brand and marketing use: Retail and hospitality deployments are driving short‑term revenue for vendors through leasing, but deliver limited operational ROI beyond engagement metrics. Recurring technical and business blockers Battery and duty cycle: Field deployments continue to be constrained by battery life and downtime for recharge—most pilots require charging‑station scheduling or hot‑swap strategies. Grasp and manipulation robustness: Reliable, rapid grasping in cluttered or varied parts remains the dominant failure mode in warehouses and manufacturing cells. Perception in clutter: Vision stacks still underperform in real messy environments; many pilots succeed only after simplifying the environment. System integration: Humanoids are rarely plug‑and‑play; integrator time and custom software are the largest cost drivers in pilots. "Early customers repeatedly tell us the biggest cost is integration—not the robot." — anonymized integrator interviewed for this report How companies are positioning their products Tesla: Framing Optimus as a future general‑purpose factory assistant; early messaging emphasizes internal deployment to refine capabilities before commercial sales. Agility / Apptronik / Figure: Positioning on industrial productivity—co‑robots that augment human teams, marketed with RaaS pilots to lower adoption risk. Sanctuary / 1X / UBTECH / Xiaomi: Emphasizing service, companionship and brand applications—pilots often aim for user acceptability and long‑tail data collection as much as immediate ROI. Practical lessons for buyers and integrators Design pilots to prove one measurable KPI (safety incidents reduced, minutes saved per shift, staff time redeployed) rather than general “productivity.” Simplify the task environment first—organized shelving, standard part presentation and reliable lighting convert demonstrations into operational pilots. Budget integration as 2–5× the robot hardware cost for first pilots; expect ongoing software support and mapping maintenance. Use hybrid operation models—teleoperation fallback and supervised autonomy—until perception and grasp reach human‑comparable reliability in your setting. Competitive takeaways By 2025–26 the competitive landscape is stratified: a small set of firms (Agility, Apptronik, Figure) are winning industrial and logistics pilots by focusing on deterministic tasks and integrator partnerships; Sanctuary and a few others pursue regulated markets with clinical pilots; legacy consumer electronics firms (Xiaomi, UBTECH) leverage brand and service relationships for marketing deployments rather than heavy industrial automation. Tesla’s Optimus remains the most watched wildcard—internal factory use could accelerate capabilities rapidly if Tesla publishes production KPIs. Conclusion Humanoid robots are working in the real world—but selectively. Evidence points to meaningful business cases where tasks are repetitive, structured and safety‑sensitive. Widespread replacement of human labor is not yet visible; the near‑term value lies in risk reduction, staff augmentation and niche service roles. For buyers the imperative is pragmatic: pick focused, verifiable metrics, budget integration properly and expect staged adoption that moves from supervised trials to limited autonomy as robot capabilities and perception stacks mature.