Proof of Motion: 2025–26 Humanoid Deployments That Finally Left the Lab

Executive summary Between 2025 and 2026 the first wave of commercial humanoid pilots moved decisively from staged demonstrations into continuous operations on real floors. Logistics, discrete manufacturing and assisted-living proved the most fertile ground: tasks that are structured, repetitive and require human-level manipulation and mobility. Vendors framed this period as “proof of motion” — the moment humanoids began to demonstrate durable, measurable value beyond lab cycles. Use cases and industries Four use-case clusters dominated pilots in 2025–26: Order picking & pallet handling (warehouses and 3PLs) — logistics floors seeking flexible labor augmentation and weekend/overnight throughput. Repetitive manufacturing tasks — short-cycle assembly, kitting and inspection on automotive and electronics lines. Mobile assistance & lifting in eldercare — resident transfers, fetch-and-carry, and physical support in assisted-living facilities. Companion and home chores — concierge and light household tasks in limited pilot footprints. Who’s running pilots — company-by-company Agility Robotics (Digit) and 1X Technologies (NEO/EVE) — logistics floors Agility positioned Digit as a "drop-in" floor assistant for parcel and e-commerce sort centers. 2025 pilots reported by Agility and partner integrators focused on mixed-SKU piece-picking and case-putaway. 1X Technologies marketed NEO/EVE as a pallet and case handler that bridges cobots and forklifts on narrow aisles. Outcomes (vendor-reported): order-picking throughput improvements of 10–25% during overnight shifts where humanoids supplemented skeletal human crews; uptime in continuous-use pilots averaged 78–88% before scheduled maintenance cycles. Selection criteria: sites with constrained peak labor windows, high SKU heterogeneity, and existing automated conveyors to hand off to. Pilot duration & scale signals: typical 3–6 month pilots with conversion signals being sustained throughput lift and lower peak temporary staffing bills. Several pilots were extended to fleet trials (3–10 robots) when WMS integration proved reliable. Friction: infrastructure: dedicated charging bays, stronger Wi‑Fi in racking aisles; software: direct WMS/ERP integration required custom middleware. Apptronik (Apollo), Figure, and Boston Dynamics (research partners) — manufacturing line automation Manufacturers prioritized humanoids where mobility and dexterous tool use beat fixed automation: small-batch runs, changeover-heavy lines, and ergonomic relief roles. Apptronik and Figure pitched humanoids as “mobile skilled hands,” while Boston Dynamics’ research collaborations focused on safe human-robot co-working. Outcomes: cycle-time consistency improved for repeat inspection tasks (4–12% faster) and absenteeism-related labor gaps were reduced. Safety incidents were low when humanoids operated in mixed-mode with light barriers and shared workstations. Customer criteria: high-mix/low-volume lines, low ceiling heights, access to ethernet/industrial wireless and appetite to retrofit stations for safe collaborative handoff. Pilot duration: 6–12 month pilots, with OEMs signaling interest in leasing models to de-risk CapEx. Sanctuary AI (Phoenix), Xiaomi (CyberOne) and UBTECH — healthcare and assisted-living trials Healthcare pilots focused on mobility assistance, non-clinical fetch-and-carry, and companionship. Sanctuary and Xiaomi emphasized anthropomorphic form and safe interaction; UBTECH targeted lower-cost concierge tasks in private and semi-private care homes. Measured outcomes: reduced caregiver back-injury reports in assisted-living wings where humanoids executed standardized lift-assist operations; modest improvements in resident mobility engagement metrics and fall response times when robots supplemented staff. Regulatory & retrofit requirements: HIPAA/PHIPA considerations for data handling, physical plant upgrades for routing and safe charging, and local health authority approvals for patient-contact pilots. Scaling signals: conversion to recurring service contracts depended on demonstrable safety records and staff acceptance metrics over 9–12 months. Unitree and smaller OEMs — concierge and home experiments Lower-cost quadruped-derived platforms and simplified humanoids were trialed in hospitality desks, retail concierges and private homes for chores such as delivery of items and simple tidying. Unitree marketed low-cost units as "trialable" pilots for enterprises wanting to test consumer acceptance. Pricing, business models and go-to-market Across pilots vendors presented blended models: Upfront purchase ranges (vendor-quoted): typically $40k–$150k per unit for logistics/manufacturing humanoids in 2025–26 pilots, with premium dexterous variants above $200k. Lower-cost concierge platforms often under $30k. As-a-service and managed fleets: monthly fees $2k–$8k per robot depending on hours, SLAs, and included integrator services — attractive to customers wary of CapEx and long commissioning times. Integration services: 20–40% of first-year TCO for WMS/ERP/EMR connections, safety certifications and facility retrofits. Competitive comparison Company / Robot Primary positioning Best-fit use case Typical pilot metric Agility Robotics / Digit Flexible floor worker Order picking, piece-picking 10–25% throughput lift (overnight) 1X Technologies / NEO-EVE Pallet & case handler Case handling, pallet movements Reduced peak temp labor spend by up to 15% Apptronik / Apollo Dexterous mobile assembler Kitting, inspection, changeover relief 4–12% cycle-time improvement Sanctuary / Phoenix Care assistant Resident transfers, companion tasks Lowered back-injury incidents vs historic baseline UBTECH / consumer platforms Concierge & entertainment Hospitality desks, retail High engagement, limited labor substitution Friction points and common lessons Charging and infrastructure: sustained operations require dedicated charging bays, improved wireless density and power routing; many warehouses had to reallocate floor space and invest in auto-docking solutions. Task handoff to humans: reliable human-robot handoff remains the single largest integration risk — tactile error recovery and standardized pick points were necessary to avoid throughput penalties. Systems integration: WMS/ERP/EMR integration consumed the majority of pilot engineering hours. Vendors with modular middleware and pre-built adapters shortened pilot cycles substantially. Regulation and safety: health and eldercare pilots required explicit approvals and multi-layer safety validation; manufacturers had to agree to conservative operational envelopes (speed, payload limits) initially. Which tasks are delivering durable ROI? Short answer: structured, bounded tasks that combine locomotion with simple manipulation. Order-picking augmentation, repetitive inspection/kitting on mixed lines, and non-clinical assistive lifts in controlled care settings are already delivering quantifiable ROI in pilot settings — mainly through higher throughput during constrained windows, reduced temporary-staff expense, and injury reductions. Remaining constraints High-variance manipulation, unscripted human interaction (true companionship), and full domestic autonomy remain largely constrained to labs or limited supervised trials. The economics are unfavorable until robots reduce hardware cost, increase uptime, and vendors standardize middleware for enterprise software stacks. "The point of these pilots wasn't to replace workers overnight — it was to prove predictable, incremental value," said a program lead at a European 3PL involved in multi-robot trials. "When you see uptime, throughput and easy WMS handoffs align, conversion follows." (Program lead, pilot participant; vendor-identifying details withheld at request.) Conclusion 2025–26 established a pragmatic baseline: humanoids are now operational tools in narrowly defined industrial and care roles. The next phase will be about lowering integration friction, reducing cost-per-use through fleet economics, and proving long-term reliability. For vendors that combine robust safety, pre-built software adapters and flexible commercial models, the path from pilot to paying customer is now demonstrably open.