← All research

Synthesis & method

Cross-competitor working notes — coverage heatmap + opportunity matrix

Accumulates the MATRIX and COVERAGE lines each dossier agent returns. Source dossiers in dossiers/<slug>/. Method: _RESEARCH-METHOD · lens: _OPPORTUNITY-LENS · grid: _MARKET-PROBLEM-MAP. Read this with the build brief: _LANDSCAPE-SYNTHESIS (the decision and the wedge). The two finished heatmaps live below.

Opportunity matrix (lens scores — working notes, NOT printed in briefs)

CompetitorCounter-positionAI-reimaginationWedgePrizeArchetypeVerdictWalk-away risk
Raken60757080AGOdistribution-owner (Procore) bundles commercial-AI before we own the data loop
Procore20802085DAVOIDit IS the distribution-owner — bundles the wedge free on data it holds
Fieldwire55726270CFLANKHilti funds a fast move into the money gap, or API stays enterprise-locked
PlanRadar45787280CFLANKPlanRadar extends AI Assistant into commercial docs / acquires a claims tool first
Document Crunch35807085CFLANKTrimble (acquired DC Apr 2026) assembles contract+project data into entitlement intelligence first
Civils.ai35807080CFLANKa funded move bolts historical-cost benchmarking onto a takeoff engine before our data loop exists
Trunk Tools45807080BFLANKa platform owner (Procore) cuts the API and ships a rival agent — or Trunk Tools itself bundles a claims agent — before our data loop locks
OpenSpace20151535CFLANKProcore (the SoR beneath it), not OpenSpace, bundles commercial AI first; their AI is computer-vision, not LLM, so not our fight
Kreo45803055DFLANKa takeoff peer ships a real cross-firm cost-benchmark dataset before our data loop exists (its current benchmarking is single-firm + announced-only)
Pype (Autodesk)22782045DAVOIDthe completed proof of the risk — Autodesk bought this exact AI doc tool and made the neutral feed a captive feature
ContraVault · micro55725070BFLANKfunded tender specialist (~$3.7M) extends its Claims Assistant down-funnel before we own a benchmarking loop
ConstructionDailyReport.ai · micro45806865BGOalready named the capture→claims move ($299 Claims tier, claim-package export); could add entitlement/benchmarking first
Voice Log Pro · micro20553025BAVOIDtoo thin to threaten — a feature, not a platform

Micro rows are the ultra-early cohort (_micro-entrants/cohort); their GO/AVOID means proximity-to-our-lane, not target priority. ConstructionDailyReport.ai’s $299 Claims tier is the single most thesis-relevant find — someone is executing our exact capture→claims up-funnel at micro scale.

Archetypes: A wedge-and-expand (GO) · B reimagine-the-job (build-new) · C build-on-top/ride-along (FLANK) · D avoid head-on. Raken row reconstructed from its dossier read (scores were in earlier working notes, not formally logged).

Coverage heatmap (0-100 vs the 21 areas, map order)

Columns 1-21 = the canonical _MARKET-PROBLEM-MAP order: 1 Bid/Tender · 2 Estimating/Takeoff · 3 Prequal/Procurement · 4 Field/Daily · 5 Time/Labour · 6 Safety · 7 Quality/QA-QC · 8 PM(SoR) · 9 Scheduling · 10 Progress · 11 RFI/Submittal/Doc · 12 BIM · 13 Reality-capture · 14 Cost-mgmt · 15 Change/Claims · 16 Accounting/Payroll · 17 O&M/Handover · 18 Equipment/Material · 19 Insurance/Risk · 20 Comms/Client · 21 Historical-cost

Competitor123456789101112131415161718192021
Raken00101008570403015604505352555540255010
Procore6545608570808510065709575559070605055709055
Fieldwire055852035806545456045530305202510655
PlanRadar001060104090454035655045555653015655
Document Crunch2510150025102055600030355100803025
Civils.ai55702500035155060535105002525155
Trunk Tools352020250545403015803501010010010555
OpenSpace00040020455106545551001050202010350
Kreo10950000000015300600001052020
Pype (Autodesk)012250004025088815055062018825
ContraVault · micro8520350000151005000540000302520
ConstructionDailyReport.ai · micro000903560101015201005555002020255
Voice Log Pro · micro000551510500550100300001550

What the full set shows (13 competitors)

The two thesis columns are empty across the entire field.

Three structural patterns, each now evidenced rather than assumed:

  1. Everyone reads documents; nobody recovers money. The whole AI-native frontier (Document Crunch, Trunk Tools, Civils.ai, ConstructionDailyReport.ai, Pype) is document reading/extraction/coordination. The value chain stops at “here is the risk / here is the log / here is the notice.” Converting that into recovered money — the adversarial, P&L-bearing step — is where all of them, by their own product boundaries, stop.
  2. The build-on-top trap is real, twice proven. Pype was an independent AI submittal/spec tool absorbed by Autodesk (neutral feed → captive feature, no open API left). Trunk Tools had its Procore API access revoked in Sept 2025 while Procore shipped a rival Agent Builder. Building a doc-AI feature on a platform’s API and buyer is rented land. Procore (AVOID), Pype (AVOID) and OpenSpace (rides Procore) are the don’t-fight-here rows.
  3. Defensibility is accountability + a proprietary data loop, not the model. Document Crunch’s own founder argues the moat is accountability + 7 years of benchmark data + legal-buyer trust, not the LLM — sharpest exactly where output enters an adversarial process. That is the recovery wedge’s natural moat.

Target board (sorted for OUR play)

TierCompetitorsMove
Build alongside (hosts)Raken, Fieldwire, PlanRadarOpen APIs, strong field/QA capture, money layer empty. Consume their logs/RFIs/defects as input; own the commercial/claims layer on top, for the QS/commercial buyer they don’t serve.
Neighbours on the document axisDocument Crunch, Trunk Tools, Civils.ai, KreoAI-native, do reading/takeoff well, stop before recovery + cross-firm benchmarking. Don’t rebuild them; flank one axis over into the money. Potential data partners.
Avoid head-onProcore, Pype (Autodesk), OpenSpaceDistribution-owners or already-absorbed. Procore is the master walk-away risk; Pype is the completed proof of it.
Micro-watchConstructionDailyReport.ai, ContraVault, Voice Log ProConstructionDailyReport.ai already named the capture→claims up-funnel at micro scale ($299 Claims tier) — watch and out-execute on depth; ContraVault touches claims as correspondence only; Voice Log Pro ignore.

Net read for the build: the wedge the whole field leaves open is money recovery (change / variation / claims / entitlement) plus cross-firm historical-cost benchmarking, sold to the commercial / QS office (a buyer the platforms don’t reach), built alongside the open-API field tools rather than on a platform’s API. That is the one position that is simultaneously high-prize, unoccupied, and defensible against the bundling risk that has already claimed Pype and squeezed Trunk Tools.

(Set complete: 10 full dossiers — Raken, Procore, Fieldwire, PlanRadar, Document Crunch, OpenSpace, Trunk Tools, Civils.ai, Kreo, Pype — plus the 3-tool micro-entrant cohort. Heatmap = the coverage table above.)


Tier B/C landscape sweep (scan depth — dossiers/_landscape/)

Lighter scan of the wider field, grouped by category. Each tool still scored on both axes so it lands in the heatmap. Notes: _landscape/reality-capture/note · _landscape/estimating-takeoff/note · _landscape/field-daily/note · _landscape/platforms-sor/note · _landscape/materials-handover/note.

Opportunity (lens)

CompetitorCatCntAIWdgPrzArchVerdictNote
DroneDeployreality30252040CFLANKmost consumable evidence feed (public API; frames output as payment-claim proof)
Buildotsreality25251545CFLANKclosed enterprise ($166M); data needs partnership
Dispersereality15201030DAVOIDacquired by OpenSpace Oct 2025 — bundled away
XYZ Realityreality30202040DAVOIDAR-hardware moat; evidence gated behind headsets
Doxelreality35452555CFLANKnearest to our lane — ships earned-value, uses LLMs; watch
STACKestim40703055DAVOIDcloud takeoff+estimating incumbent; cost only via bought-in 1build catalog, not a proprietary loop
Togal.AIestim45803055DFLANKAI-native CV takeoff (Kreo analog); single-firm “predictive costing” announced-only
Bluebeam Revuestim25552045DAVOIDNemetschek-owned legacy markup incumbent; bundles Bluebeam Max AI on its install base
PlanSwiftestim35602545DAVOIDConstructConnect-owned legacy desktop; “Takeoff Boost” AI bolt-on
Countfireestim40503035DAVOIDnarrow single-trade (electrical/M&E) symbol-counter; furthest from commercial cost
SiteLogsfield10101510DAVOID$0 ad-supported micro-utility, not a competitor
EasySiteLogfield20403520BAVOIDcapture already free ($0/$19); no money layer to defend
BuildPassfield40724560BFLANKonly field tool climbing up-funnel + shipping real agentic AI (Agent + MCP into Claude); still stops before claims
SmartBarrelfield35303045CAVOIDtime/labour hardware; payroll-integration moat, CV-not-LLM
Fonnfield30504055CFLANKacquired by The Access Group (ERP roll-up) — 2nd build-on-top proof; UK fit-out
Matrakmaterials30552045CFLANKdisruption-claim evidence (delivery/out-of-sequence); only one here with shipped AI (AI Takeoff→BOQ)
Sablonomaterials35602560CFLANKbest delay-claim data source (actual-vs-plan on P6/Asta/MSP); QS buyer overlap
Zutecmaterials30401540DAVOIDcompliance handover / golden thread; off our line
Operancematerials25351535DAVOIDpure O&M asset handover; off our line
Catenda Hubmaterials30401540DAVOIDBIM coordination + handover; off our line
Autodesk Construction Cloudplatform20822085DAVOIDproven absorber (bought Pype); owns distribution + spec corpus + agent platform
Trimble Construction Oneplatform22722580DAVOIDowns ERP+PM+estimating; Viewpoint Finance Assistant could aim a finance agent at the commercial layer
Oracle Aconex+Primaveraplatform20702080DAVOIDstrongest audit corpus; already ships cross-project predictive AI (Advisor for Safety) on captive data
Buildertrendplatform25452045DAVOIDresidential SMB distribution; thin AI; different buyer

Coverage (0-100, map order 1-21)

Competitor123456789101112131415161718192021
DroneDeploy0503005550153075104010020200253025355
Buildots00020255303055901045851510051015405
Disperse000250530540805359020505510250
XYZ Reality0001505701520451080751015030525250
Doxel0002500402060955509045100515153010
STACK5951510100520051010055000502515
Togal.AI590000001000510020000002015
Bluebeam Revu10750505301505602502500500550
PlanSwift5855050510005503500050100
Countfire0700000550050050005050
SiteLogs000551550055500000015050
EasySiteLog00080201000055000500010150
BuildPass001080457560504530550015100154030500
SmartBarrel00045903052503500035105003015305
Fonn0015752555506535405055201010253020600
Matrak05540200035203060103502510059510405
Sablono00035105703075902015015150251510455
Zutec0010200357530510702055100951555305
Operance005502510250040400050953045250
Catenda Hub0001005454010206595305100351010500
Autodesk Construction Cloud706055805575809060609090507555456045658050
Trimble Construction One357555706555608580557085458545904560507040
Oracle Aconex+Primavera302555503080709095609555308550406035808055
Buildertrend3035408055404575705040557045552030258515

What the landscape sweep adds (25 more tools, ~38 total)

Bottom line after the sweep: the picture from the first 13 held — then a direct hunt of the two thesis areas (below) sharpened it and forced one correction. Money recovery (area 15) is NOT empty — a real but sub-scale cohort (Magra et al.) does it. What stays empty is cross-firm historical cost (area 21) and the integrated loop no single tool closes. Refined thesis: _LANDSCAPE-SYNTHESIS.


Thesis-area deep dive — the wedge hunted directly

Two extra scans aimed straight at the wedge. Full per-tool profiles + 21-area vectors in _landscape/claims-delay/note and _landscape/commercial-qs/note; decision columns below. Area 21 caveat: the benchmarking entrants score on the area but are SINGLE-firm — true cross-firm pooling is still occupied by no one. Sablono and BuildPass were promoted from sweep rows to full dossiers.

Claims / delay / recovery (14 tools) — area 21 is ≤10 for every one

ToolClassArea 15VerdictNote
nPlana · predict10AVOIDhuge schedule dataset; forecasts risk, not recovery
Nodes & Linksa · predict20AVOIDproject-controls AI; stops at analysis
SmartPMb · delay30FLANKschedule analytics / delay
ClaimLogicb · delay40FLANKclean P6 delay engine (claimlogic.app)
ForensicPMb · delay40FLANKMCP-native forensic engine
ScheduleLensb · delay40FLANKLLM delay narrative
Steelray Delay Analyzerb · delay40AVOIDpriced expert-witness install base (~$3,990/yr)
Systechb · delay40AVOIDconsultancy delay-analysis software
ClaimMaster.aic · claims-mgmt60FLANKdefensible claims; stops short of quantum
Astorac · evidence55FLANKclaim evidence layer (astora.app)
Magrad · RECOVERY85FLANKnearest competitor — Eichleay/Hudson/measured-mile, ~$36K/yr, owner+contractor (magra.app)
ClaimDD / ClaimEOTd · RECOVERY85FLANKone-person LLC, two apps; computes quantum
Masin AId · RECOVERY80FLANKconsultancy spin-out, strong team (masin.ai)
Delay Claim Builderd · RECOVERY80FLANKlive; automates cost-quantum

Class: a = schedule-risk prediction · b = delay analysis · c = claims/evidence mgmt · d = genuine quantum/recovery. Four reach (d); all four are 2024-25, unfunded/sub-scale; none pools cross-firm cost.

Commercial / QS (12 tools) — our buyer’s existing software

ToolBuildsA14A15A21VerdictNote
CausewayQS commercial suite801520AVOIDowns the QS desk; legacy, no real AI
Eque2ERP-attached commercial801510AVOIDinvoice OCR only
RedSkyERP commercial801210AVOIDERP roll-up
Cleopatra EnterpriseEPC cost control903525AVOIDchange mgmt, not entitlement
RIB CostXestimating + commercial552025AVOIDestimating base
RIB Candycivils EVM + commercial752515AVOIDEVM; stops before entitlement
Planyardcloud CVR challenger852515FLANKcloud-native CVR
Payapps (Autodesk)subcontractor valuations60205AVOIDabsorbed by Autodesk (4th absorption)
Gather Insightscomp-event detect + notices355510FLANKfront-half mover — NEC/JCT events + notices, our QS buyer, UK; stops before quantum
GaugeAI benchmarking (single-firm)60555FLANKsingle-firm only
Rate QSAI rate library (single-firm)65555FLANKmarkets “private, not cross-firm” as a feature
BenchItAI cost benchmarking (single-firm)60550FLANKsingle-firm pools

Read: the QS suites run the money workflow to variation/CVR/final-account and stop (the variation↔entitlement seam); the recovery happens in a separate vendor category (the claims engines); the benchmarking entrants are all single-firm by design. The recovery workflow is contested but sub-scale; cross-firm cost and the integrated loop are open. Nearest competitors to watch: Magra (recovery/quantum), Gather Insights (front half, our buyer, UK), BuildPass (capture + agentic AI).