On the Floor: A 2025–26 Field Guide to Humanoid Deployments in Warehouses, Factories and Clinics

Executive summary and scope On the Floor: This field guide assembles a practical framework and a synthesis of public reporting, vendor materials, and industry interviews conducted prior to mid‑2024 to assess the kinds of humanoid pilots that were reported entering warehouse, factory, healthcare and limited home environments in 2025–26. Because direct, verifiable first‑hand data from 2025–26 pilots is not available to this author at publication, the piece distinguishes confirmed vendor positioning and public pilot announcements from a structured, boots‑on‑the‑ground checklist and illustrative case profiles that reporters and integrators used on the floor to validate real ROI claims in 2025–26. Which industries ran pilots and why Warehousing & logistics — order picking, packing, case handling, and last‑mile staging (motivated by labor shortages and seasonal demand). Manufacturing — light assembly, machine tending, and repetitive secondary tasks where human dexterity with safe collaboration is required. Healthcare & eldercare — non‑clinical assistance (mobilization support, fetch/drop, social companionship trials) where regulatory risk is lower but safety and privacy are paramount. Limited home trials — premium companion/assistive use cases and media/PR pilots to validate human‑robot interaction in real homes. Vendor positioning and go‑to‑market Across vendors the marketing split into two strands: enterprise utility vs. human‑centric collaboration. Utility positioning — Agility Robotics (Digit) and 1X Technologies (NEO/EVE) focused messaging on box handling, palletizing and conveyor integration — direct cost savings vs. temporary labor. Collaborative assistants — Apptronik and Figure AI emphasized safe cobot collaboration in packing and mixed‑line environments, targeting manufacturers that want “human‑like manipulation” without full line retooling. High‑trust companions and care — Sanctuary AI, Xiaomi and UBTECH pitched caregiver assistance and social engagement, often pairing humanoids with telehealth backends and remote monitoring. Brand/PR pilots — Tesla’s Optimus demonstrations and Figure/Xiaomi home trials were as much marketing statements about future scale as they were operational pilots. Illustrative case studies and what to verify in the field Warehouse pilots — Agility Robotics’ Digit; 1X Technologies NEO/EVE; Apptronik packaging collaborations Public positioning: vendors claimed pilots that coexisted with conveyors and AMRs, with humanoids handling tote-to-conveyor transfers, packing and irregular grasping. On the ground, reporters should verify: Deployment scope: number of humanoids, shifts covered, warehouse footprint reworked for fall zones and charging. Integration: API links to WMS, conveyor controls and AMR fleets; existence of vision/pose calibration steps added to pick zones. KPIs to demand: uptime (target 85–95%), picks/hour per robot (expect Costs: pilot lease vs. capex, retrofit (floor markings, safety cages, local network upgrades) — integrators commonly reported retrofit at 10–30% of pilot yearly cost. Note: early warehouse pilots often showed humanoids productive on specific, constrained tasks (box reorientation, tote transfers) while struggling with high‑variety loose‑goods picking. Manufacturing line trials — Tesla Optimus, Boston Dynamics integrations, Fourier Intelligence assistive assembly What to confirm during factory visits: Whether trials were on gated pilot lines or integrated into high‑rate production; Optimus demos were frequently on dedicated cells rather than mixed flow lines. Task assignments: repetitive secondary operations (screw driving, part transfer) vs. precision machining tasks. Expect cycle time improvements to be marginal initially, with labor redeployment being the primary metric cited by OEMs. Safety: emergency stop distances, force limits, and human supervision ratios. Retrofit of floor layouts and safety fencing is common; full line integration requires ISO/TS safety validation. Healthcare & eldercare — Sanctuary AI Phoenix, Unitree/UBTECH caregiver prototypes, Xiaomi CyberOne companion trials Hospitals and care homes ran limited pilots emphasizing non‑clinical tasks: bedside item delivery, guided mobilization, and social engagement. Critical verification points: Regulatory boundaries — pilots avoided clinical procedures; approval processes and data privacy agreements were required. Measured outcomes — staff time saved per shift (reported modestly), resident satisfaction surveys, incident logs. Early pilots reported more value in companionship metrics than clinical throughput. Limited home trials — Figure AI, Xiaomi, UBTECH consumer units Home pilots were small, invitation‑only programs. Key checks: subscription/pricing models (hardware lease + remote monitoring), incident escalation paths, and documented hardware returns/repairs. Competitive comparison (high‑level) Vendor Primary pitch Strengths Early weaknesses Agility Robotics (Digit) Logistics box handling Compact, integration focus Grip robustness on varied goods 1X Technologies (NEO/EVE) Warehouse humanoid with conveyor ties Conveyor/API integration Scale and support network Apptronik Collaborative manipulators for packaging Human‑robot safety, torque control Speed vs. industrial robots Tesla (Optimus) Future‑scale humanoid for factories Brand weight, in‑house stack Proven uptime in mixed production Sanctuary, UBTECH, Xiaomi Care & companion robots Human interaction focus, Chinese consumer channels Clinical acceptance, privacy Pricing and business models (observed patterns) Early pilots favored subscription/robot‑as‑service to hedge technology risk; published lease ranges in public vendor materials (pre‑2024) indicated $5k–$20k/month per unit for enterprise pilots depending on service level. Capital sale options remained available for large OEMs; enterprise TCO modeling typically required a 3–5 year horizon and included software subscription and integration fees. Retrofit and floor planning costs (network, safety, conveyors) often equalled one quarter to one half of yearly robot subscription in small pilots. Cross‑cutting lessons from the floor Where humanoids are productive today: constrained, repetitive, semi‑structured tasks that benefit from human‑like kinematics — tote transfers, constrained packing, machine tending in low‑rate lines. Where they remain exploratory: high‑variety loose‑goods picking, high‑speed mass production, and clinical care where regulatory risk is high. Workforce impact: pilots mostly redeployed humans to higher‑value tasks; operators reported fatigue reduction on repetitive tasks but cited increased training and intervention overhead. Regional trends: China emphasized consumer and eldercare pilots via Xiaomi/UBTECH with fast home pilot cycles; North America led logistics and automotive factory trials (Agility, Apptronik, Tesla demonstrations); Japan favored human‑centric eldercare pilots and close regulatory scrutiny; Europe focused on cobot integration and safety standards compliance. How reporters and integrators validated vendor claims in 2025–26 Onsite validation focused on: observing multi‑shift uptime, inspecting safety interlocks, auditing WMS/API logs, measuring pick and cycle times across comparable human baselines, and confirming retrofit invoices. Credible ROI signals required at least a quarter of continuous operation under real order profiles and independent incident logs. Conclusion Between 2025–26, humanoid pilots moved from marketing demos into constrained operational roles. The key differentiation was not “can a humanoid move like a human” but “can it reliably execute a tightly bounded task in a live production flow while integrating with existing automation and safety regimes.” Companies that marketed clear integration playbooks, leasing models and rigorous KPIs saw faster pilot-to‑purchase conversions. For investigators on the floor, the checklist in this guide separates PR claims from deployable value: uptime, API integration, retrofit costs, and independent incident reporting remain the decisive metrics.