Standard-work OS for warehouse teams to prove which humanoid tasks are safe, economic, and rollout-ready across sites.

1. More than $300 million of multi-year orders and a 30-plus customer pipeline show warehouse buyers are already committing budget, so they need a repeatable method to decide what gets rolled out next.
2. Nine live customer sites across named operators mean humanoids have moved beyond demo floors, making cross-site standard-work decisions immediate rather than theoretical.
3. Safety is still the main blocker even after 65,000-plus operating hours, so the scarce workflow is not robot control but operator-side evidence and approval for each new task and facility.
4. Agility is investing in Arc and an integrated software and safety stack, which means OEM-side tooling will exist; the open wedge is the neutral layer that helps buyers govern workflows across sites and vendors.

Catalyst. Agility's SPAC financing, $300 million-plus order base, and planned scaling to broader deployments mean operators must choose rollout-ready workflows now, even though safety is still the primary blocker.

The product sits above OEM fleet software and below the warehouse operator's rollout committee. It ingests task histories from WMS or WES systems, intervention and failure logs from the humanoid vendor, and on-floor safety observations from site teams. It then scores each candidate workflow by repeatability, exception burden, required staffing changes, and site-specific hazards, before generating updated SOPs and an expansion packet. Customers use it to decide which next site, shift, and task should launch first instead of running every new deployment as a bespoke consulting project. Over time, the company builds the best cross-site benchmark on where humanoid deployments stall, what mitigations work, and which workflows convert into durable unit economics.

What's different. Most robotics software helps the OEM run the robot. This company helps the operator decide whether a workflow deserves replication at all, using a repeatable task-qualification model that spans sites and eventually vendors. That position lets it accumulate proprietary benchmarks on intervention rates, SOP changes, and rollout failure modes that neither a single OEM nor a consulting team can easily match.

Jobs to be done

flowchart LR
  Buyer[Warehouse network engineering lead] --> Pain[Unsure which humanoid workflow can scale safely]
  Pain --> Product[Standard-work qualification OS]
  Product --> Outcome[Faster multi-site rollout with fewer bad launches]

Executive takeaways

• The market is narrow today but already urgent for operators moving from a single humanoid pilot to repeatable multi-site rollout decisions.
• The wedge is buyer-side workflow qualification, not robot control: operators need proof that a task is safe, economic, and reproducible before they scale.
• Adjacent incumbents can run robots or warehouses, but none is purpose-built to generate an audit-ready go/no-go packet for the next facility.
• The best early product will look partly like process intelligence, partly like safety evidence management, and only secondarily like robot fleet software.

Market definition

Operator-side software that decides whether a humanoid workflow, shift, and facility are ready for replication across a warehouse or factory network. It sits above OEM fleet tools and adjacent to WMS/WES execution.

Customer and buyer

The day-one user is a network industrial-engineering or continuous-improvement team stitching together pilot telemetry, exceptions, and EHS observations before a rollout committee meeting. The buyer is the VP/Head of automation, distribution engineering, or network operations who owns the next-site decision and capital narrative.

Buying triggers

• A paid pilot is ending and the operator must defend a second-site rollout or additional capex with something more rigorous than vendor optimism. [1][54]
• EHS or operations leadership asks whether a humanoid can move beyond a defined work zone without introducing unacceptable new hazards. [58][70][71][73][75]
• Labor cost, turnover, and physically taxing tasks keep robotics budgets active, but funding approvals still require a clear path to value. [59][60][77]
• As more robots, people, and software systems share the building, orchestration and evidence become fragmented across OEM tools and warehouse systems. [10][41][47][52]

Willingness to pay

Automation programs already absorb meaningful software, integration, and commissioning line items; CPCON and GoASRS both show six-figure software/integration scopes are normal in warehouse modernization, so a qualification layer can fit inside existing program budgets if it shortens consulting cycles or prevents one bad rollout. [65][66][67]

Category dynamics

Growth signal 11% y/y U.S. industrial robot installations in 2025

Validation signals

• Agility already reports more than $300M of multi-year Digit v5 orders and a pipeline of 30+ customers.
• GXO progressed from pilot to a multi-year commercial RaaS agreement, suggesting warehouse buyers will sign formal deployment contracts.
• Digit has surpassed 100,000 totes moved in a live GXO facility, providing real throughput evidence rather than demo-floor theater.
• Figure’s BMW deployment published explicit shift, runtime, and intervention KPIs, showing the broader category is producing real operating data.
• Apptronik’s near-$1B capital base and named partnerships signal a crowded commercialization race that will create more operator demand for neutral rollout tooling.

Regulatory & technical constraints

• There is no humanoid-specific OSHA rule today; deployments inherit existing OSHA and consensus-standard obligations for robot safety.
• Legged and dynamically stable robots still sit in an emerging standards gap that requires deployment-specific safety cases.
• Current commercial use still fits best inside defined work zones or bounded material-handling tasks rather than unconstrained floor roaming.
• Trustworthy-AI and safety-architecture expectations are rising alongside industrial deployment of embodied AI.
• European expansion adds formal AI Act documentation, oversight, and operator obligations.

The nearest substitutes are OEM fleet clouds, robot-operations orchestration tools, and incumbent WMS/WES suites. Each covers part of the problem, but the open slot is the neutral layer that turns messy pilot evidence into a repeatable cross-site rollout decision.

Why incumbents do not win by default

• OEM fleet clouds. They own robot telemetry and deployment tooling, but they optimize one vendor fleet rather than a buyer’s cross-site go/no-go decision across workflows and EHS sign-off.
• Robot-orchestration platforms. These tools coordinate robots, people, and incidents across fleets, yet they stop short of producing workflow-qualification packets for rollout committees.
• WMS/WES incumbents. Incumbent warehouse suites already orchestrate tasks and resources, but they are designed to run the building, not to benchmark intervention-heavy humanoid workflows across sites and OEMs.
• Integrators and digital-twin consultants. Consulting-heavy players can assess each launch as a project, but their economics and institutional memory stay people-bound unless the workflow becomes productized software.

Humanoid deployments in warehouses are now commercially real enough that some operators must decide which pilot workflows deserve second-site rollout, yet the decision process is still fragmented across OEM dashboards, WMS exports, manual intervention logs, and EHS walkthroughs. This company sells the operator-side standard-work qualification OS that converts that fragmented evidence into a go/no-go packet for the next workflow, shift, and facility. The beachhead is North American retail, ecommerce, and manufacturing warehouse networks with 3-20 sites, one live humanoid pilot, and a fiscal mandate to expand or stop. The first sale is not generic robot software; it is a paid decision system for the quarterly pilot review where network engineering and EHS must justify more capex. The wedge is attractive because OEM fleet clouds and WMS/WES suites run operations, but neither is built to generate a neutral, audit-ready replication recommendation across sites. The plan deliberately starts with operator-owned data so the company can prove value before negotiating deeper OEM telemetry access. The largest risks are a still-small installed base, OEM feature bundling, and the possibility that buyers still treat each rollout as bespoke consulting. Research did not surface direct buyer quotes, published deployment-pricing benchmarks for this exact category, or a standard incident taxonomy, so pricing and buyer-process assumptions must be validated early.

Problem

• Multi-site warehouse operators cannot yet prove which humanoid workflow is safe, economic, and repeatable enough to copy from one pilot site into the next.
• The evidence needed for expansion decisions is fragmented across WMS/WES task history, OEM intervention logs, SOP changes, and EHS observations, so each rollout becomes an ad hoc review.

Solution

• Ingest operator workflow data, intervention logs, and safety observations to score candidate humanoid tasks by repeatability, exception burden, staffing impact, and site-specific hazards.
• Generate a committee-ready rollout packet with recommended next site, required SOP changes, explicit assumptions, and a go/no-go decision for EHS and network engineering approval.

Why we win

• The product is neutral and operator-aligned, while OEM tools are optimized for one robot fleet and warehouse suites are optimized for daily execution.
• Starting with operator-owned data reduces dependency on any one vendor and lets the company prove value before deeper telemetry integrations are required.
• Every qualified workflow adds proprietary benchmarks on intervention patterns, SOP deltas, and time-to-stable-throughput that are hard for services firms to compound.

Milestones

flowchart LR
  Wedge[Second-site qualification wedge] --> MVP[Workflow scoring and rollout packet]
  MVP --> Proof[Faster approvals and more stable launches]
  Proof --> Expansion[More sites, workflows, and OEMs per account]

Founding team

Experiment roadmap

Risk assessment

What must be true

• At least one named operator stakeholder treats second-site qualification as a budgeted problem, not just an OEM services task.
• A first packet can be built mostly from operator-owned data without waiting on proprietary OEM telemetry.
• EHS and industrial-engineering stakeholders will trust a reusable packet format enough to reduce bespoke analysis time.
• Production customers will expand from one workflow or site to at least one more within 12 months.
• Neutral benchmark data will stay more valuable than a single-vendor dashboard as the humanoid market matures.

Open diligence questions

• Who owns the final go/no-go decision for second-site rollout today, and where does the software budget sit?
• What minimum data fields are truly required to produce a packet buyers will trust?
• How often do OEMs or integrators already provide this analysis as part of deployment economics?
• Which first workflow has the fastest path from live pilot to repeatable multi-site proof?
• What evidence would make an EHS leader approve a new site without commissioning a bespoke third-party study?

Model sanity

• Revenue engine. Base-case Y3 revenue comes from 15 paying rollout programs at $300K mature ARR plus a $50K qualification fee on each new logo.
• Must go right. The company must turn design-partner work into repeatable onboarding so gross margin can move from the high-60s to the mid-70s while partner channels keep CAC near $120K.
• Model breaks if. If sales cycles drift to 6 months or margin stalls near 72%, the late-Y3 cash floor drops below zero before Series A proof is ready.
• Next-round proof. The next financing story is 8 production accounts, 2 multi-site expansions, and sub-21-day onboarding by Q4Y2, which is exactly what the seed ask is sized to reach with 6 months of buffer.

Scenarios

Sensitivity

flowchart LR
  QualifiedPipeline[Qualified pilots and partner referrals] --> Customers[Paying rollout programs]
  Customers --> Revenue[Subscription revenue + packet fees]
  Revenue --> GrossProfit[Gross profit]
  GrossProfit --> EBITDA[EBITDA after payroll and operating spend]
  EBITDA --> Cash[Ending cash]

Flags: Y1 assumes three paid logos in a still-narrow market before the team has fully validated who owns the rollout budget in every operator segment. · The base case carries zero logo churn through the explicit monthly customer schedule even though steady-state unit economics assume 2.0% monthly churn; real churn would pull forward the next raise. · Cash bottoms near $496K in late Y3, so a 1-2 quarter slip in closes or margin improvement would require slower hiring or bridge financing. · Gross-margin expansion depends on keeping implementation productized; if connector or safety work becomes bespoke, the 75% mature GM assumption will not hold.

• Narrow early market. Only a limited number of warehouse operators have live humanoid pilots today, which can slow initial customer count. Mitigation: Start with design partners already budgeting for second-site expansion and support adjacent mobile-manipulation workflows while humanoid penetration rises.
• OEM data dependency. The best qualification model needs robot intervention and failure data that some vendors may hesitate to expose. Mitigation: Begin with operator-owned WMS, ticketing, video, and SOP systems, then use buyer demand to standardize deeper telemetry adapters.
• Consulting creep. Customers may try to buy bespoke rollout analysis instead of software if every site looks unique. Mitigation: Productize around a fixed task ontology, benchmark library, and generated approval packet so professional services stay light and repeatable.