On the Floor: Inside a Dozen Live Humanoid Pilots Transforming Warehouses, Factories and Care Homes (2025–26)

Executive summary Humanoid robots moved from lab demos into pragmatic pilots in 2023–24, and companies pushed expanded trials through 2025–26. What was once a parade of technology showcases is now a set of focused, short‑run deployments in logistics, factories and care settings. This field‑style report synthesizes public company disclosures, media reports and vendor materials to catalog verified and company‑reported pilots (Agility Robotics, Figure, Apptronik, Sanctuary, Xiaomi, UBTECH, Tesla) and draws lessons about where humanoids add measurable value today, what remains experimental, and the procurement realities procurement teams must confront. Catalog of notable pilots (what companies say they are running) Company / Robot Industry & Use Case Status (public) Business model / Pricing context Agility Robotics — Digit Warehousing / material handling: tote carrying, human‑robot handoffs, parcel movement. Multiple short pilots in distribution centers; company‑reported customer tests with 3PLs and retailers. Hardware sales + pilot programs; pilots reported as multi‑month engagements. Per‑unit pricing not publicly standardized; pilots typically under RaaS or trial contracts. Figure — Figure‑1X / 01 Logistics: end‑to‑end picking, shelf‑to‑line transport in fulfillment centers. Company reports 1X pilot programs with fulfillment customers in North America and Asia. Figure has promoted occupancy‑style commercial agreements; pricing reported as premium for early adopters. Apptronik & Sanctuary Factory automation: kitting, line support, parts presentation in electronics/aerospace assembly. Trials announced and demoed with manufacturing partners; Apptronik emphasizes human‑adjacent workflows. Hardware + integration services; some pilots use time‑limited leases. Xiaomi — CyberOne Residential assistance, monitoring and consumer demonstrations; positioning for personal help and smart‑home integration. Brand pilots and smart‑home integration demos; limited public commercial deployments. Positioned as premium consumer product; Xiaomi talks about unit sales rather than enterprise RaaS. UBTECH Eldercare / assistance: monitoring, reminders, basic mobility assistance and social engagement in care homes. Multiple deployment announcements in China; pilot installations in senior care facilities and pilot programs with local governments. Product sales to institutions + subscription software for updates and services. Tesla — Optimus (reported) Internal factory trials: repetitive material movement, light assembly support (reported internally by Tesla). Elon Musk and Tesla have publicly stated internal trials; independent verification remains limited. Tesla aims for low unit costs; business model unspecified for external customers. Where humanoids are already adding measurable value Across the cataloged pilots there are recurring, measurable benefits where humanoids map to human workflows without heavy workspace changes: Carrying and shifting loads between human stations — In logistics pilots, bipedal mobility lets robots navigate human‑centric aisles and hand items to workers, reducing worker walking time and partial FTE substitution for repetitive movement. Kitting & line support — In factory trials, humanoids that can present the correct part to an operator cut pick‑to‑light walk time and reduced cognitive load in mixed‑model assembly. Patient mobilization and monitoring (early stage) — In eldercare pilots, robots provide reminders, passive monitoring and limited physical assistance (supporting, not replacing, human caregivers). Performance metrics reported Published pilot metrics are still sparse and vendor‑specific, but several trends emerge from reported results: Short‑run uptime during pilot days typically ranges 80–95% (downtime due to perception/environment mismatch rather than hardware failure). Task completion rates vary widely: 60–90% for single‑type carrying tasks in structured environments, lower for open picking. Human‑robot handoff times in trials often approach human‑to‑human times for simple tote exchanges once ergonomics are tuned (~1–3 seconds extra per handoff). Safety incidents publicly reported are rare; most companies emphasize supervised operation and human‑in‑loop intervention during pilots. Known issues are falls in cluttered spaces and perception errors causing brief stoppages. How vendors position their robots — and why it matters Vendors cluster into three marketing archetypes: Enterprise integrators (Agility, Figure, Apptronik) — Emphasize modular workflows, RaaS pilots, and integration with WMS/PLC to reduce worker walking time and improve throughput. Consumer & lifestyle (Xiaomi) — Present humanoids as smart‑home companions and personal assistants, prioritizing perception and social features over heavy payload. Industrial low‑cost disruptors (Tesla aspiration) — Promise scale via in‑house manufacturing and low per‑unit cost; commercialization path remains speculative. Competitive analysis: form factor and payload matter Bipedal humanoids split along payload and dexterity axes. Key competitive observations: Light, mobile humanoids (10–20 kg payload) — Best for last‑mile movement, kitting, and human handoffs in cluttered human spaces (current sweet spot for Agility/early Figure models). Higher‑payload, torso‑capable systems — Needed for heavy industrial material handling; more complex, costlier and less common in current pilots. Form factor flexibility — Robots with human‑like hands (more dexterity) are more costly to integrate but enable broader tasks; simpler end‑effectors speed uptime but limit use cases. Taxonomy of verified real use cases Picking — Partial picks in fulfillment centers where vision is constrained and parts are uniform. Kitting — Supplying operators with the next part in mixed assemblies. Line support — Moving subassemblies between stations, buffering work in progress. Patient mobilization — Assisted standing/transfers in supervised configurations; limited to non‑critical tasks today. Companionship/Monitoring — Social presence, reminders and fall detection in senior living pilots. Lessons from pilots: humanoids succeed where the robot can slot into an existing human workflow with little environment rework, and where the task is bounded and repetitive. Procurement checklist for pilot teams Define a single, bounded use case with clear KPIs (walk time saved, picks/hour, uptime target). Require a staged integration plan: perception tuning, safety validation, and human‑robot ergonomics testing. Insist on data access: logs, failure modes and intervention records for the pilot period. Budget for systems integration — early pilots typically double base robot costs once integration and software are included. Plan human oversight: designate trained operators for supervision and rapid intervention. Negotiate trial terms with exit criteria and defined handover if scaling is approved. Timeline to scaled deployment — a frank assessment For narrowly defined logistics and kitting tasks in structured environments, scaled deployments (multi‑site rollouts) are plausible in a 3–6 year window (2028–2032) if vendors standardize integration and lower the per‑unit TCO. For broad factory automation replacing skilled human dexterity, or for unsupervised patient handling in healthcare, meaningful scale remains 5–10+ years out. Consumer/home humanoids face an even longer path due to cost, safety and user acceptance hurdles. Humanoids are no longer purely speculative. The pilots documented by vendors and selectively reported in the field show concrete wins where form factor matters. The remaining gap is predictable integration work — perception tuning, ergonomics, and convincing finance teams of repeatable ROI. Procurement teams that approach pilots as systems projects, not gadget buys, will be best placed to move from promising tests to scaled deployments.