On the Ground: Verified Humanoid Robot Deployments That Shifted Operations in 2025–26

Executive summary This forensic survey compiles verified, public evidence of humanoid robot pilots and early commercial moves across warehouses, factories, hospitals and care homes through mid‑2024 and assesses which programs have measurable operational impact likely to drive paid deployments in 2025–26. My source set is limited to company filings and press releases, customer statements, integrator disclosures and industry reporting available through June 2024; where public verification does not exist I note that explicitly. The picture is not of widescale rollouts but of selective, task‑specific conversions from pilot to paid engagements—mainly in logistics and manufacturing, with exploratory but lower‑rate trials in healthcare and eldercare. Which industries moved the needle—and why Logistics & warehousing: acute labor shortages and high injury rates make warehouses the fastest adopters of humanoids for picking, case handling and last‑mile delivery trials. Manufacturing/assembly: ergonomic relief on repetitive or awkward tasks (line tending, parts kitting) drives interest in co‑bot humanoids that can work alongside humans and existing tooling. Healthcare & eldercare: staffing pressure and infection‑control needs are producing trials for non‑clinical assist tasks (room checks, supplies transport, social engagement) rather than medical procedures. Consumer/retail R&D: companies with consumer hardware businesses use humanoid demos as marketing halo products and limited in‑store experiments rather than industrial rollouts. Company profiles and verified deployments Agility Robotics (Digit) — warehouses & last‑mile pilots Agility’s Digit has been positioned for package handling and order‑fulfillment. Public materials and trade interviews through 2024 document multiple pilots in fulfillment centers and last‑mile staging areas where Digit performs case transfers, pallet tasks and arm‑assisted picking. Company‑reported lessons emphasize integration with conveyor and staging workflows rather than wholesale WMS replacement: operators redesign pick zones and staging so Digit works in handoff points. Agility has pursued a mixed commercial model of direct sales for integrators and pilot‑to‑RaaS commercial conversations, but neither standard list prices nor a documented fleet‑level payback case widely accepted by neutral customers were publicly disclosed as of mid‑2024. Figure Labs / 1X / NEO — industrial & logistics trials Figure and a set of capitalized startups using '1X' or 'NEO' nomenclature (public demos and private trials) have focused on dexterous manipulation and endurance in material handling. Verified evidence through press releases and trade shows shows pilots in logistics nodes and light industrial environments. The firms market humanoids as drop‑in manipulators for tasks that are hard to automate with fixed automation (irregular goods, mixed SKU bins). Integration stories note high engineering effort to align manipulators with conveyor indexing and vision systems; throughput gains are reported by vendors but generally lack independent verification. Apptronik (Apollo) — co‑bot manufacturing integrations Apptronik positions Apollo as a collaborative humanoid that augments human stations. Public demonstrations and integrator case studies show Apollo used for kitting, machine tending and repetitive insertions in low‑volume, high‑mix lines. Apptronik’s go‑to‑market has included partnerships with systems integrators and MES vendors to map human tasks to humanoid scripts. Pricing discussions in public filings and interviews indicate hybrid models—hardware sale to Tier‑1 integrators plus long‑term service agreements—rather than a pure RaaS offering. Sanctuary AI and UBTECH — eldercare and clinical‑assist trials UBTECH (with established service‑robot lines in Asia) has publicly reported deployments of companion/service humanoid robots in Chinese eldercare facilities doing room checks, medication reminders and limited social engagement. Sanctuary AI has positioned its humanoid for generalist assist roles; public trials reported to date are pilot‑scale with eldercare and hospital logistics partners. In both cases the verified value is reduction in non‑clinical staff time on repetitive tasks and improved patient/ resident engagement metrics in pilot reports, but clinical tasks and patient transfers remain largely out of scope for safety/regulatory reasons. Xiaomi (CyberOne) and Unitree — consumer/home experiments Xiaomi’s CyberOne and Unitree’s exploratory biped/quasi‑humanoid prototypes are primarily marketing and R&D efforts. Verified activity through product debuts and limited in‑store demonstrations shows experiments in home assistance and smart‑home integration; neither company had public, paid home‑deployment programs of scale through mid‑2024. These efforts serve branding, data collection, and human‑robot interaction research rather than immediate enterprise ROI. Tesla (Optimus) and Boston Dynamics — high‑profile demos, limited paid pilots Optimus and Boston Dynamics’ humanoid concepts attracted wide attention. By mid‑2024 both firms had staged high‑visibility demos and internal pilots (Optimus at Tesla factories; Boston Dynamics for research on Atlas). Public, paid customer deployments of humanoid robots by either firm were not verified at scale. Boston Dynamics’ commercial success remained concentrated in non‑humanoid platforms (Spot, Stretch) where clear ROI and safety cases exist. Operational metrics reported and integration realities Publicly reported pilot metrics are sparse and vendor‑sourced; common, verifiable themes include: Uptime: early pilots run robots for single‑shift cycles with reported operational uptimes commonly in the 70–90% range (vendor reported); maintenance windows and perception of fragility remain conversion barriers. Throughput: measured task throughput often lags human specialists on complex pick tasks but can match or exceed humans when workflows are redesigned to exploit a robot’s strengths (e.g., dedicated transfer zones, fixed handoff points). Task success & safety: task success in well‑scoped, repeatable tasks (case transfers, machine tending) is high; patient handling and heavy lifting remain unproven in public pilots. Safety incidents reported publicly are rare but when they occur they trigger conservative operational restrictions and lengthen procurement cycles. Injury reduction & ergonomics: integrator case studies and customer statements link humanoids to ergonomic relief, particularly for repeat stooping and lifting tasks, but broad injury‑rate reductions require long‑term data not yet widely available. Workflows, systems integration and procurement that converted pilots into paid work Successful pilot conversions shared common patterns: Redesign small, bounded workflows (handoff points, kitting islands) rather than full process reengineering. Integrate to WMS/MES/EHR through middleware or integrator partners; few vendors replace existing enterprise systems. Use staged commercial models: time‑boxed pilots → pilot extensions → small paid deployments. RaaS is increasingly pitched to lower CAPEX hurdles but full RaaS commercial rollouts remain limited. Contracts often start as multi‑phase POCs with performance milestones and options to scale; integrators frequently take title of final solution or package hardware + services pricing into a managed service offering. Competitive comparison — task credibility matrix Task Vendors with verified pilots/public evidence Credibility (Proven/Promising/Experimental) Picking (structured SKUs) Agility, Figure, select integrators Promising Case handling / pallet transfers Agility, Apptronik (integrator pilots) Proven (bounded) Line tending / machine tending Apptronik, Figure Promising Patient transfer / heavy lift UBTECH/Sanctuary (trials limited to non‑lift assist) Experimental Room checks / delivery in hospitals UBTECH, pilot integrators Promising Retail/consumer assistance Xiaomi (demos), unitree (experiments) Experimental Pricing, business models and caveats Vendors have avoided public, standardized list pricing for humanoids; the market has moved toward bespoke pricing and RaaS pilots to lower entry barriers. Broadly: Large industrial customers are buying through systems integrators with CapEx + services contracts or managed service terms. RaaS is commonly pitched to logistics customers to align with seasonal and labor variability; firm fleet‑wide RaaS benchmarks remain vendor specific. Procurement cycles lengthen with safety and compliance checks—hospital procurement, for instance, adds regulatory review and clinical governance that slow conversions. Conclusions and lessons learned By mid‑2024 the industry shows credible, incremental wins: vendors and integrators have moved from demos to bounded, verifiable pilots in logistics and manufacturing that redesign local workflows. Healthcare and eldercare show promise for non‑clinical assistance, but safety, regulatory and liability questions keep lifts and transfers off the table. High‑profile projects (Optimus, Atlas) remain mostly demos with symbolic value rather than verified commercial throughput. The near‑term winners are those that: (1) tightly scope tasks, (2) integrate with existing enterprise systems through trusted integrators, and (3) offer procurement paths that mitigate CAPEX and safety risk (pilot → pilot extension → managed service). Note: This report is based on publicly available evidence and vendor/customer disclosures through June 2024. Public verification of scaled, paid humanoid fleets in 2025–26 is limited; readers should treat projections beyond that date as trend‑based analysis rather than confirmed deployments.