Legs on the Line: Where Humanoid Robots Are Actually Working in Warehouses, Factories and Clinics (2025–26)

Introduction — legs on the line in 2025–26 Humanoid robots moved from lab demos to the factory floor and warehouse aisle in measurable but limited ways in 2025–26. Operators in logistics, manufacturing and healthcare are running pilots and small-scale deployments that target tasks where a biped’s ability to use human-built infrastructure (stairs, pallets, shelf heights and existing tooling) provides a comparative advantage to wheeled or fixed-arm automation. This report surveys concrete use cases, company-focused case studies and the commercial realities operators face. Primary use cases and industries Order picking & e-commerce fulfillment — small-piece picking, single-item tote handling and trailer-side transfers (logistics/3PL, e-commerce). High-throughput sorting & packaging — rapid grasping, singulation and put-wall packing (sortation centers, parcel networks). Machine tending & cobot line roles — loading/unloading machines, pallet kitting, light assembly (discrete manufacturing, automotive suppliers). Patient mobility assistance, telepresence and disinfecting — assisted-living, hospitals and clinics (healthcare). Hybrid yards and micro-logistics — repetitive bin transfers in environments designed for human bodies (automotive final assembly, cold storage). Company case studies Agility Robotics — Digit in e-commerce and 3PL pilots Agility’s Digit has become the most visible commercial humanoid in logistics pilots. Agility reports pilots across North American and European third‑party logistics sites focused on box and tote transfers, trailer unloading and short-distance aisle-to-conveyor tasks. Deployment scale is predominantly multi-unit pilots (5–20 robots) rather than line-rate production. Measured outcomes: company and customer statements cite reductions in human bending and two-person lifts, with pilot KPIs showing labor-substitution equivalence for simple transfer tasks and claimed throughput uplifts in the 10–30% range on constrained workflows. Integration: pilots typically hook into warehouse management systems (WMS) from Manhattan/Blue Yonder or bespoke APIs and use systems integrators for conveyor and PLC interlocks. Business model: Agility offers sales and RaaS; reported pricing context from industry reporting places advanced humanoid hardware in the low-to-mid six-figure range per unit, with RaaS subscriptions used to spread risk. Customer feedback: operators praise the humanoid’s ability to handle human-native fixtures but flag battery endurance and predictable routing through crowded aisles as operational challenges. Figure AI and 1X Technologies (NEO/EVE) — sorting and packaging Figure AI and 1X Technologies position humanoids specifically for high‑speed pick‑and‑place and singulation, leveraging faster arms, vision stacks and optimized end‑effectors. Deployment: 1X’s NEO/EVE pilots are in parcel sort centers and cold-chain packaging trials; Figure’s systems have been trialed in third‑party sortation hubs that process high SKU variety. Outcomes: public pilot metrics emphasize cycle-time parity with human pickers on certain SKUs and error rates comparable to trained temporary staff; companies report throughput improvements when humanoids are paired with optimized conveyors and AI sortation software. Positioning: marketed as throughput-augmenting solutions, packaged with vision and motion software and often sold via integrator partnerships rather than direct to end-users. Apptronik (Apollo) and Sanctuary AI (Phoenix) — factory cobots and line automation Apptronik’s Apollo and Sanctuary’s Phoenix are explicitly pitched as collaborative humanoids for kettling, machine tending and light assembly. Early factory pilots come from automotive suppliers and electronics contract manufacturers where tooling and fixturing already suit a human form. Scale: single-cell deployments (1–4 units) integrated with PLCs (Siemens, Rockwell) and attended by human operators. Outcomes: cited reductions in repetitive-strain incidents and consistent cycle times in 24/7 shifts; operators report improved uptime versus temporary human labor during peak ramps. Go‑to‑market: both companies emphasize safety, deterministic control stacks for certification and traditional system integrator routes to customers. Tesla Optimus — trials inside Gigafactories and logistics yards Tesla’s Optimus program ran iterative trials in Fremont and other Gigafactories performing logistics-yard transfers, parts movement and simple line feeding. Public demos and internal footage showed basic material handling tasks; however, deployments are at an experimental/pilot stage rather than commercial production. Measured outcomes: company presentations highlight autonomy milestones; independent verifications of operational KPIs remain limited. Automaker insiders describe Optimus as useful for non-critical repetitive tasks during off-peak hours. Integration: internal Tesla software stacks and custom conveyors; no broad external WMS integrations reported publicly. UBTECH and Xiaomi (CyberOne) — eldercare and assisted‑living trials UBTECH and Xiaomi’s experiments with humanoid platforms have concentrated on assisted living: mobility assistance, telepresence, medication reminders and light disinfecting. Deployment scale: pilot programs in a handful of assisted‑living facilities and pilot wings of hospitals; units are typically single-site, small-scale trials. Outcomes: measured increases in resident engagement and nurse time freed for clinical tasks; however, human caregivers remain dominant and survey data show mixed acceptance, especially for physically assisting transfers. Pricing & model: firms usually offer hardware with service contracts and ongoing software updates; pricing is positioned for premium care operators, not mass-market consumer sales. Unitree and quadrupeds as complements Quadrupeds from Unitree and smaller firms are frequently deployed alongside humanoids — carrying payloads across rougher terrain, scouting aisles or transporting bins. Facility managers use quadrupeds for endurance-heavy movements and humanoids for manipulation where human interfaces exist. Competitive comparison Company / Robot Primary role Reported payload/endurance Deployment stage (2025–26) Agility — Digit Order picking, transfers ~18 kg peak claim , several-hour mission Pilots (multi-unit) Figure AI High-throughput pick/pack Arm-optimized cycles, short tasks Pilots at sort centers 1X — NEO/EVE Sortation & packaging High cycle-rate focus Pilots Apptronik — Apollo Machine tending / cobot Industrial-grade certifications in progress Single-cell factory pilots Sanctuary — Phoenix Collaborative line roles Human-safe force control Cell pilots Tesla — Optimus In-house logistics & line support Varied statements; prototypes Experimental pilots Commercial positioning and pricing context Most humanoid vendors market two propositions: (1) access to human‑native environments without expensive retrofits; and (2) labor augmentation/shift smoothing where temporary or dangerous human labor is costly. Pricing remains heterogeneous: hardware purchase prices are typically quoted in industry reporting as low-to-mid six figures, while RaaS and service-subscription models spread costs across multi-year contracts. Integrators remain a critical channel to realize ROI. "Humanoids buy us time — they let us defer fixed infrastructure investment — but they are not yet a drop-in replacement for lines designed around robots," said an operations executive at a North American 3PL in a public pilot briefing. Practical barriers and sector map Operators still face structural hurdles: Safety & certification: ISO standards and local regulations require rigorous testing and can slow deployments — especially for physical human interaction in healthcare. ROI timelines: expected payback commonly cited at 3–7 years in public discussions — shorter only where labor costs are extreme or tasks are highly repetitive. Payload & endurance limits: most humanoids remain constrained to light-to-moderate payloads and multi-hour mission times requiring charging swap strategies. Systems integration: real ROI depends on tight WMS/PLC/intralogistics integration and mature change-management on the shop floor. Sector-by-sector map (near-commercial vs experimental) Logistics (order picking, trailer unloads) — near-commercial for constrained workflows (3PL pilots expanding to limited production). Manufacturing (machine tending, kitting) — near-commercial in single-cell, high-value lines where humans and humanoids collaborate. Healthcare (patient handling, eldercare) — experimental to early-pilot; telepresence and monitoring are practical now, physical patient transfers remain rare. High-throughput sortation — pilot-stage with clear path to commercial when cycle rates and error budgets meet parcel operator thresholds. Conclusion By 2025–26 humanoids have demonstrated real operational value in a narrow band of applications that leverage their compatibility with human-built infrastructure. The most credible commercial cases are in logistics and certain factory cells — where firms like Agility, Figure AI, Apptronik and Sanctuary have established repeatable pilot patterns. Broader adoption will hinge on improved payload/endurance, faster paths to safety certification, demonstrable multi-site ROI and continued work by systems integrators to make humanoids part of predictable, auditable workflows.