Dashboard: Dashboard · market grid: _MARKET-PROBLEM-MAP · opportunity lens: _OPPORTUNITY-LENS · cross-competitor: _CROSS-COMPETITOR · format exemplar: _micro-entrants/cohort · the OpenSpace dossier sits in this category: openspace/dossier
This category turns the physical jobsite into structured data — drone/360/AR capture of what is actually built — and (for most of these tools) runs computer vision over that capture to measure progress against the plan, the BIM model, and the schedule. The AI here is almost entirely computer vision, not LLM: the job is “look at the site, tell me what is built, where it deviates, and how far behind it is,” not language generation. The strategic point for us: the output of this whole category is potential delay / disruption evidence — objective, timestamped, third-party-credible ground truth of what was (and was not) built when. That is exactly the raw material an entitlement / claims layer consumes. None of these tools cross into the money layer themselves; they stop at progress %, valuation support, and “here is the deviation.”
Per-tool mini-profiles
DroneDeploy — The broadest reality-capture platform: one login over drone (aerial), 360 walkthroughs, fixed site cams, and ground robots (Boston Dynamics Spot), producing 2D maps and 3D models across the project lifecycle, sold across construction, energy, mining, ag, utilities, roofing, insurance. Its AI is vision-language CV (“trained on billions of sq ft”) shipped as Progress AI (schedule tracking, earthworks quantities), Safety AI (hazard detection), and quality/defect detection (rust, leaks, debris); notably it markets payment / claims verification — “if they claim 100% done but it’s actually 80%, we can prove it.” Pricing is published and self-serve-ish: Individual ~$329/mo (annual) / $499 monthly, Advanced ~$599/mo, Enterprise custom — unusually transparent for this set. Openness is the standout: a genuinely public developer platform (GraphQL API + legacy REST map-processing API + API keys + an actual third-party App Market with Procore/BIM integrations). Distance to our wedge: medium-far on the money axis (it verifies and even validates payment claims, but does not assemble entitlement / quantify a delay claim), but closest of this set as a data source we could consume — its capture and progress output is exactly claim evidence, and its open API means we could read it.
Buildots — Helmet-mounted 360 camera + a construction-intelligence platform that uses computer vision to turn each site walk into “ground truth,” then compares actual progress to the BIM model and the schedule, flagging deviations and predicting delays before they hit (headline claim: up to 50% delay reduction, ~2-3 months saved). Buyer is enterprise: top-tier GCs and owners (Turner, JE Dunn, Mortenson, Multiplex, Kier, VINCI, Bouygues; UK-strong). Well-funded — ~$166M total, $45M Series D in 2025 at ~$300M valuation; ~£6.35M FY24 revenue, projecting triple-digit growth; it also operates Genda (workforce intelligence). Pricing unpublished (enterprise sales-led); public API not verified — it ingests BIM/schedule/site into “one coordinated system” but no open developer API was found. Distance to our wedge: far on money (pure progress/risk detection, no claims/cost), but its deviation+schedule-impact data is directly relevant as delay evidence — though as a closed enterprise platform it is a harder data source to consume than DroneDeploy.
Disperse — Was an independent UK CV progress-tracking firm (360 capture → work-in-place vs plan → productivity analytics, milestone-based). Acquired by OpenSpace on 28 Oct 2025 (terms undisclosed), following a June 2025 partnership; it now powers “OpenSpace Track” / OpenSpace Progress Tracking. The disperse.io domain now 301-redirects to openspace.ai. So Disperse is no longer a standalone target — it is OpenSpace’s progress-tracking engine, and OpenSpace itself sits on Procore as the system of record (see openspace/dossier). AI angle: CV plus expert human verification to validate work-in-place for billing and flag schedule risk early. Distance to our wedge: same as OpenSpace — far from money, billing-validation only, and now inside a distribution-adjacent platform.
XYZ Reality — The odd one out: not a CV progress engine but AR hardware. The Atom is a safety-certified hard-hat AR headset (onboard Intel i7/16GB/1TB) that overlays the BIM model on the live site at millimetre accuracy, so trades build directly from the model and verify works in place. It is build-verification / quality / rework-avoidance, not automated % complete: the headline outcomes are rework cost avoided ($135M+ portfolio; $4.14M on one data center) and days saved. Buyer is mission-critical, mostly data center GCs and MEP/structural specialists (Mace, PM Group, Interstates). Revit/Autodesk BIM360 integration via a custom plugin; broader open API not verified; pricing undisclosed (hardware + platform, enterprise). Its own marketing explicitly cites “evidence capture for claims/delays” and objective accuracy-to-tolerance validation. Distance to our wedge: far (AR build-aid + QA), but it generates timestamped, mm-accurate as-built verification that is unusually strong dispute evidence on quality/rework — a niche evidence source, gated behind proprietary hardware.
Doxel — CV progress tracking with the deepest push toward the money side of this set. 360-video capture (hard-hat) → CV quantifies work-in-place per trade/system/stage → a “rules-of-credit engine” converts that into accurate % complete → color-coded 3D planned-vs-actual, production rates, and out-of-sequence/rework detection. Crucially it ships automated earned-value tracking (“instantly calculate cost vs. progress,” supports monthly valuations / pay apps, claims up to 16% cash-outflow reduction) and uses LLMs for the scheduling layer (map schedule to actual progress) — so it is CV-primary but not CV-only. Buyer is both owners and GCs (Layton, Sundt, DPR, QTS, Scripps, Stream Data Centers). Funded ~$56.5M (a16z, Insight); ~$3.9M ARR FY24 — well-funded relative to revenue, i.e. still early commercially. Pricing and public API not disclosed. Distance to our wedge: nearest of this category — it touches cost (earned value) and forecasts delays — but it stops at progress-linked cost/valuation; it does not assemble a delay/disruption entitlement narrative, quantify a claim adversarially, or reuse cross-firm historical cost. It validates the up-funnel direction without occupying our step.
Comparison
| Tool | Builds | Buyer | AI angle | Open API? | Distance to our wedge | Read |
|---|---|---|---|---|---|---|
| DroneDeploy | Multi-modal reality capture (drone/360/site-cam/robot) → maps, 3D, progress/safety/quality AI | GCs, owners, data centers + many verticals (energy, mining, ag) | CV / vision-language (Progress, Safety, quality, payment-claim verification). Not LLM | Yes — public GraphQL + REST + App Market (Procore/BIM) | Medium-far on money; closest consumable evidence source (open API + claim-verification framing) | CONSUMABLE-DATA (ride-along data source, open API) |
| Buildots | Helmet-360 → CV progress vs BIM + schedule, deviation + delay prediction | Enterprise GCs / owners (Turner, VINCI, Kier; UK-strong) | CV vs BIM/schedule; predictive delay. Not LLM | Not verified (closed enterprise; no dev API found) | Far on money; strong delay/deviation evidence but harder to consume | NEIGHBOUR-closed (well-funded, gated) |
| Disperse | (Now OpenSpace Track) 360 → CV+human progress vs plan, billing validation | GCs / owners (via OpenSpace) | CV + expert verification. Not LLM | Via OpenSpace (Procore/Autodesk integrations) | Far on money; now inside OpenSpace | ABSORBED (acquired by OpenSpace 28 Oct 2025) |
| XYZ Reality | Atom AR hard-hat headset: mm-accurate BIM overlay → build verification, rework avoidance | Mission-critical / data center GCs + MEP/structural | AR hardware (+ accuracy validation); not CV auto-progress, not LLM | Revit/BIM360 plugin; broad API not verified | Far; niche mm-accurate as-built/QA evidence, gated behind hardware | AR-OUTLIER (hardware, not a data feed) |
| Doxel | 360-video → CV work-in-place per trade → % complete, earned value, production rates, delay forecast | Owners + GCs (DPR, Sundt, QTS, Stream) | CV-primary + LLM for scheduling; rules-of-credit % complete | Not disclosed | Nearest — touches cost (earned value) + delay forecast, stops before claims/entitlement | NEAREST-NEIGHBOUR (reaches toward money, stops short) |
What this category tells us
- Our wedge (money recovery, area 15) is touched at the edges but never occupied. DroneDeploy markets payment-claim verification, XYZ Reality cites “evidence capture for claims,” and Doxel ships earned-value + delay forecasting — but every one stops at “here is the objective progress / valuation / deviation.” None assemble a delay/disruption entitlement narrative, quantify a claim adversarially, track recovery, or reuse cross-firm historical cost (area 21 is essentially empty across all five). The recovery step stays unoccupied — consistent with the full-set finding that “everyone reads/measures; nobody recovers money.”
- The AI frontier here is computer vision, not LLM — which is good for us, not a competitor. Four of five are CV-primary (Doxel adds LLMs only for the scheduling map). This category is solving a different problem (what is built, % complete) than our LLM-shaped entitlement-narrative problem. That makes them upstream evidence producers, not rivals on our axis — their timestamped progress/deviation output is exactly the claim evidence our layer would consume.
- DroneDeploy is the single most consumable data source in the category — uniquely it has a genuinely public developer API (GraphQL/REST/App Market) and already frames its output as payment-claim proof. If we want third-party objective progress evidence under a claim, this is the cleanest “build-on-top / ride-along” data partner (Archetype C). Buildots, Doxel, XYZ Reality are closed/enterprise/hardware-gated and harder to read.
- Acquisition / distribution-owner risk is live and already realised here. Disperse was acquired by OpenSpace (Oct 2025) and folded into OpenSpace Track; OpenSpace in turn rides Procore as the system of record. So the progress-tracking layer is consolidating into the capture platforms, which sit on the SoR incumbents. This is the same bundling pattern that claimed Pype — capture/progress is being rolled up, leaving the money layer above it still open and still the contested ground.
- Doxel is the one to watch for up-funnel drift. It is the only tool here that has crossed into cost (earned value / valuations) and uses LLMs at all. It is the nearest neighbour to our lane from the reality-capture side; if anyone in this category bolts a claims/entitlement step onto progress data, Doxel is the most likely. Still early commercially (~$3.9M ARR on ~$56.5M raised), so the window is open, not closed.
Sources & method
Per-tool light pulls (exa search/answer + App Store) under raw_<tool>/raw/; vendor pages via WebFetch (dronedeploy.com, buildots.com, doxel.ai, xyzreality.com/atom, openspace.ai); funding/acquisition/pricing via web search (AEC Magazine / PRNewswire / Construction Dive on the OpenSpace–Disperse deal; DroneDeploy pricing; Buildots and Doxel funding; DroneDeploy developer docs for API openness). App-store note: DroneDeploy (Flight App, 4.57, real but it is the drone-flight controller, not the analytics product), Buildots and Doxel have real but unrated companion-capture apps; Disperse and XYZ Reality returned no genuine app match (unrelated “Learn Building Construction” and the “Atomas” game) — expected, as both are enterprise/hardware with no consumer app. No absence stated without a check; “not verified” used where the developer surface could not be confirmed.