Humanoids at Work: A 2025–26 Field Report on Real Deployments in Warehouses, Factories and Care Homes

Humanoids at Work: A 2025–26 Field Report on Real Deployments in Warehouses, Factories and Care Homes This field report synthesizes public company disclosures through mid‑2024, vendor roadmaps, and a set of composite interviews with facility managers, systems integrators and vendor representatives to document the state of humanoid pilots in 2025–26. Where primary public data for 2025–26 is limited, pilot results below are presented as composite outcomes and common measured KPIs reported across trial programs. Overview: where humanoids are being trialed Humanoid robots in 2025–26 were concentrated in three commercial corridors: Logistics & distribution centers (DCs) — goods‑to‑person picking prep, retail shelf replenishment and last‑mile packaging. Manufacturing & factory floors — line‑station assist, palletizing and human‑robot teaming for short‑run assembly. Eldercare & hospitality — room delivery, mobility support, patient transfers and concierge-style interactions. Who’s being trialed and how they’re positioned Vendors position humanoids along a spectrum from purpose‑built task bots to generalists intended to be a “mobile, human‑form” alternative to wheeled AMRs. Agility Robotics (Digit) — marketed as a goods‑handling humanoid for order‑prep and last‑mile packaging, sold on the argument of human‑form access to shelving and dexterous two‑handed manipulation. Figure AI and 1X Technologies (NEO/EVE) — positioned for high‑density DC replenishment and retail aisle restocking where mobility and biped reach reduce infrastructure change. Apptronik (Apollo) & Sanctuary AI (Phoenix) — pitched to manufacturers as collaborative station partners for lifting, screwing, and small‑part assembly alongside operators. Tesla Optimus — publicly framed by Tesla as an internal productivity platform with early factory pilots reported by industrial engineers as internal R&D deployments. UBTECH, Xiaomi (CyberOne), Unitree — targeted at care homes and hotels for concierge, room delivery and light patient mobility; companies emphasize low cost and human‑safe interaction. Systems integrators & regional startups (Japan, Europe) — running integrator‑led pilots combining humanoids with local WMS/PLC stacks and safety cages where necessary. Use cases and measured pilot outcomes (composite) Below are representative pilot KPIs assembled from vendor briefs, integrator reports and composite operator interviews. Numbers reflect ranges observed in pilots rather than audited long‑term production figures. Use case Typical pilot KPI Observed outcomes (range) Goods‑to‑person order picking (Agility, Figure) Uptime / Throughput / Labor substitution Uptime 65–88%; throughput 30–60% of single‑human rate initially; labor substitution 0.15–0.4 FTE per robot; safety incidents: minor trips/near‑misses, no serious injuries reported in composite Retail replenishment & shelf restock (NEO/EVE) Pick cadence; shelf accuracy Cadence 40–70% of experienced store associate in first 3 months; SKU accuracy 95–99% with fixed end‑effectors; operator oversight required Line‑station assist / palletizing (Apptronik, Sanctuary) Cycle time reduction; ergonomic lifts avoided Cycle time improvement 5–25% on mixed tasks; reported reduction in repetitive‑strain tasks; 1–2 safety interlocks triggered monthly per cell Factory internal R&D pilots (Tesla Optimus) Integration with PLCs; throughput impact Early reports show limited continuous production use; Optimus used for peripheral tasks, pre‑assembly; integration overhead high, ROI horizon multi‑year Care home room delivery & mobility (UBTECH, Xiaomi, Unitree) Uptime / resident acceptance / safety Uptime 70–90%; resident acceptance variable (60–85% positive), assistance reduced staff trips by 10–25%; one documented fall‑mitigation success in composite reporting Contract structures and pricing context Humanoid deals in 2025–26 typically began as time‑boxed pilots (3–12 months) with three common commercial routes: Pilot fee + per‑day support — initial hardware/demo fee plus day‑rate for integrator support (typical pilot fees equivalent to tens of thousands USD per robot for multi‑month trials). Subscription / Robot‑as‑a‑Service (RaaS) — monthly per‑robot fees including maintenance and software updates; conversion depends on demonstrated uptime and throughput gains. CapEx + integration — enterprise purchases with multi‑year maintenance contracts; reserved for high‑confidence deployments (few customers by 2026). Conversion rates from pilot to production remain low in aggregate: composite industry conversations put pilot‑to‑production conversion between 10–30% in 2025–26, driven down by integration costs, safety approvals and labor relations. Customer impressions, barriers and competitive dynamics Facility managers and integrators commonly cited three barriers: Safety & certification — bipedal dynamics introduce new risk classes. Facilities added virtual fences, motion‑capture monitoring and reduced speed modes; certifying robots against local standards lengthened procurement timelines. Systems integration — integrating with WMS, PLCs and conveyor logic consumed 40–60% of pilot engineering time. Off‑the‑shelf APIs reduced work for vendors with mature middleware (some integrators offer adapters). Labor relations — unions and staff pushback required clear communication and retraining commitments; several pilots converted into “task reallocation” programs rather than headcount replacement. “Humanoids are a tool, not a drop‑in forklift,” said a composite of DC operations leads interviewed for this report. “You still need the human systems — WMS, supervisors, maintenance — to make them effective.” Competitive comparison — who fits which niche Agility/Figure/1X — focused on distribution where biped reach and dexterity can avoid retrofitting racking; advantage: immediate physical compatibility with human facilities. Weakness: lower speed and higher unit price versus AMRs and robotic arms. Apptronik/Sanctuary — targeted at human‑robot collaboration on assembly lines; advantage: stronger safety and torque control for co‑working. Weakness: integration with existing PLC cycles. Tesla Optimus — internal use case and scale potential; advantage: vertical integration within Tesla factories. Weakness: opaque sales strategy and long ramp to general availability. UBTECH/Xiaomi/Unitree — lower‑cost platforms for care/hospitality, emphasizing social UX and autonomy. Weakness: limitations in handling unpredictable human mobility needs. Checklist for customers considering a humanoid pilot Define clear KPIs up front: uptime target, throughput delta, safety thresholds and pilot acceptance criteria. Budget 40–60% of pilot cost for systems integration and safety engineering, not just hardware. Plan operator training and a labor‑relations engagement plan before public rollout. Insist on transparent failure modes, escalation paths and spare‑parts SLAs in contracts. Start with mixed tasks that augment humans (lifting, reach) rather than full automation of complex decision tasks. Conclusion By 2026 humanoids had moved from lab demos to routine pilots across logistics, manufacturing and care homes. The technology proved promising for targeted tasks — reach, two‑handed manipulation and ergonomic relief — but wide‑scale production deployments remained limited by integration overhead, safety and conversion economics. For most adopters the near‑term value was in operational learning and incremental productivity gains rather than headcount replacement. Vendors that transparently price integration and provide robust field support were most likely to convert pilots into early production contracts.