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)
| Competitor | Counter-position | AI-reimagination | Wedge | Prize | Archetype | Verdict | Walk-away risk |
|---|---|---|---|---|---|---|---|
| Raken | 60 | 75 | 70 | 80 | A | GO | distribution-owner (Procore) bundles commercial-AI before we own the data loop |
| Procore | 20 | 80 | 20 | 85 | D | AVOID | it IS the distribution-owner — bundles the wedge free on data it holds |
| Fieldwire | 55 | 72 | 62 | 70 | C | FLANK | Hilti funds a fast move into the money gap, or API stays enterprise-locked |
| PlanRadar | 45 | 78 | 72 | 80 | C | FLANK | PlanRadar extends AI Assistant into commercial docs / acquires a claims tool first |
| Document Crunch | 35 | 80 | 70 | 85 | C | FLANK | Trimble (acquired DC Apr 2026) assembles contract+project data into entitlement intelligence first |
| Civils.ai | 35 | 80 | 70 | 80 | C | FLANK | a funded move bolts historical-cost benchmarking onto a takeoff engine before our data loop exists |
| Trunk Tools | 45 | 80 | 70 | 80 | B | FLANK | a platform owner (Procore) cuts the API and ships a rival agent — or Trunk Tools itself bundles a claims agent — before our data loop locks |
| OpenSpace | 20 | 15 | 15 | 35 | C | FLANK | Procore (the SoR beneath it), not OpenSpace, bundles commercial AI first; their AI is computer-vision, not LLM, so not our fight |
| Kreo | 45 | 80 | 30 | 55 | D | FLANK | a 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) | 22 | 78 | 20 | 45 | D | AVOID | the completed proof of the risk — Autodesk bought this exact AI doc tool and made the neutral feed a captive feature |
| ContraVault · micro | 55 | 72 | 50 | 70 | B | FLANK | funded tender specialist (~$3.7M) extends its Claims Assistant down-funnel before we own a benchmarking loop |
| ConstructionDailyReport.ai · micro | 45 | 80 | 68 | 65 | B | GO | already named the capture→claims move ($299 Claims tier, claim-package export); could add entitlement/benchmarking first |
| Voice Log Pro · micro | 20 | 55 | 30 | 25 | B | AVOID | too 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
| Competitor | 1 | 2 | 3 | 4 | 5 | 6 | 7 | 8 | 9 | 10 | 11 | 12 | 13 | 14 | 15 | 16 | 17 | 18 | 19 | 20 | 21 |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Raken | 0 | 0 | 10 | 100 | 85 | 70 | 40 | 30 | 15 | 60 | 45 | 0 | 5 | 35 | 25 | 55 | 5 | 40 | 25 | 50 | 10 |
| Procore | 65 | 45 | 60 | 85 | 70 | 80 | 85 | 100 | 65 | 70 | 95 | 75 | 55 | 90 | 70 | 60 | 50 | 55 | 70 | 90 | 55 |
| Fieldwire | 0 | 5 | 5 | 85 | 20 | 35 | 80 | 65 | 45 | 45 | 60 | 45 | 5 | 30 | 30 | 5 | 20 | 25 | 10 | 65 | 5 |
| PlanRadar | 0 | 0 | 10 | 60 | 10 | 40 | 90 | 45 | 40 | 35 | 65 | 50 | 45 | 5 | 5 | 5 | 65 | 30 | 15 | 65 | 5 |
| Document Crunch | 25 | 10 | 15 | 0 | 0 | 25 | 10 | 20 | 5 | 5 | 60 | 0 | 0 | 30 | 35 | 5 | 10 | 0 | 80 | 30 | 25 |
| Civils.ai | 55 | 70 | 25 | 0 | 0 | 0 | 35 | 15 | 5 | 0 | 60 | 5 | 35 | 10 | 5 | 0 | 0 | 25 | 25 | 15 | 5 |
| Trunk Tools | 35 | 20 | 20 | 25 | 0 | 5 | 45 | 40 | 30 | 15 | 80 | 35 | 0 | 10 | 10 | 0 | 10 | 0 | 10 | 55 | 5 |
| OpenSpace | 0 | 0 | 0 | 40 | 0 | 20 | 45 | 5 | 10 | 65 | 45 | 55 | 100 | 10 | 5 | 0 | 20 | 20 | 10 | 35 | 0 |
| Kreo | 10 | 95 | 0 | 0 | 0 | 0 | 0 | 0 | 0 | 0 | 15 | 30 | 0 | 60 | 0 | 0 | 0 | 10 | 5 | 20 | 20 |
| Pype (Autodesk) | 0 | 12 | 25 | 0 | 0 | 0 | 40 | 25 | 0 | 8 | 88 | 15 | 0 | 5 | 5 | 0 | 62 | 0 | 18 | 8 | 25 |
| ContraVault · micro | 85 | 20 | 35 | 0 | 0 | 0 | 0 | 15 | 10 | 0 | 50 | 0 | 0 | 5 | 40 | 0 | 0 | 0 | 30 | 25 | 20 |
| ConstructionDailyReport.ai · micro | 0 | 0 | 0 | 90 | 35 | 60 | 10 | 10 | 15 | 20 | 10 | 0 | 5 | 5 | 55 | 0 | 0 | 20 | 20 | 25 | 5 |
| Voice Log Pro · micro | 0 | 0 | 0 | 55 | 15 | 10 | 5 | 0 | 0 | 5 | 5 | 0 | 10 | 0 | 30 | 0 | 0 | 0 | 15 | 5 | 0 |
What the full set shows (13 competitors)
The two thesis columns are empty across the entire field.
- Change / Variations / Claims / Entitlement (15): highest score is Procore 70, then Document Crunch 35, ContraVault 40, Fieldwire/Raken 25-30, the rest ≤10. But not one tool recovers money. Every tool that touches area 15 stops before the money: Procore’s Contract-Review agent and Trunk Tools’ agents flag and coordinate; Document Crunch and ContraVault generate notices and organise correspondence; none assemble cost/time/causation, quantify a delay/disruption claim, or track recovery. The recovery step itself is unoccupied.
- Historical-cost / benchmarking (21): highest is Procore 55, then Document Crunch / Kreo 20-25, everything else ≤10. Critically, even Kreo’s “benchmarking” is single-firm and announced-only; no one ships a cross-firm, market-wide cost dataset. The compounding-data moat is empty at every tier.
Three structural patterns, each now evidenced rather than assumed:
- 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.
- 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.
- 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)
| Tier | Competitors | Move |
|---|---|---|
| Build alongside (hosts) | Raken, Fieldwire, PlanRadar | Open 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 axis | Document Crunch, Trunk Tools, Civils.ai, Kreo | AI-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-on | Procore, Pype (Autodesk), OpenSpace | Distribution-owners or already-absorbed. Procore is the master walk-away risk; Pype is the completed proof of it. |
| Micro-watch | ConstructionDailyReport.ai, ContraVault, Voice Log Pro | ConstructionDailyReport.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)
| Competitor | Cat | Cnt | AI | Wdg | Prz | Arch | Verdict | Note |
|---|---|---|---|---|---|---|---|---|
| DroneDeploy | reality | 30 | 25 | 20 | 40 | C | FLANK | most consumable evidence feed (public API; frames output as payment-claim proof) |
| Buildots | reality | 25 | 25 | 15 | 45 | C | FLANK | closed enterprise ($166M); data needs partnership |
| Disperse | reality | 15 | 20 | 10 | 30 | D | AVOID | acquired by OpenSpace Oct 2025 — bundled away |
| XYZ Reality | reality | 30 | 20 | 20 | 40 | D | AVOID | AR-hardware moat; evidence gated behind headsets |
| Doxel | reality | 35 | 45 | 25 | 55 | C | FLANK | nearest to our lane — ships earned-value, uses LLMs; watch |
| STACK | estim | 40 | 70 | 30 | 55 | D | AVOID | cloud takeoff+estimating incumbent; cost only via bought-in 1build catalog, not a proprietary loop |
| Togal.AI | estim | 45 | 80 | 30 | 55 | D | FLANK | AI-native CV takeoff (Kreo analog); single-firm “predictive costing” announced-only |
| Bluebeam Revu | estim | 25 | 55 | 20 | 45 | D | AVOID | Nemetschek-owned legacy markup incumbent; bundles Bluebeam Max AI on its install base |
| PlanSwift | estim | 35 | 60 | 25 | 45 | D | AVOID | ConstructConnect-owned legacy desktop; “Takeoff Boost” AI bolt-on |
| Countfire | estim | 40 | 50 | 30 | 35 | D | AVOID | narrow single-trade (electrical/M&E) symbol-counter; furthest from commercial cost |
| SiteLogs | field | 10 | 10 | 15 | 10 | D | AVOID | $0 ad-supported micro-utility, not a competitor |
| EasySiteLog | field | 20 | 40 | 35 | 20 | B | AVOID | capture already free ($0/$19); no money layer to defend |
| BuildPass | field | 40 | 72 | 45 | 60 | B | FLANK | only field tool climbing up-funnel + shipping real agentic AI (Agent + MCP into Claude); still stops before claims |
| SmartBarrel | field | 35 | 30 | 30 | 45 | C | AVOID | time/labour hardware; payroll-integration moat, CV-not-LLM |
| Fonn | field | 30 | 50 | 40 | 55 | C | FLANK | acquired by The Access Group (ERP roll-up) — 2nd build-on-top proof; UK fit-out |
| Matrak | materials | 30 | 55 | 20 | 45 | C | FLANK | disruption-claim evidence (delivery/out-of-sequence); only one here with shipped AI (AI Takeoff→BOQ) |
| Sablono | materials | 35 | 60 | 25 | 60 | C | FLANK | best delay-claim data source (actual-vs-plan on P6/Asta/MSP); QS buyer overlap |
| Zutec | materials | 30 | 40 | 15 | 40 | D | AVOID | compliance handover / golden thread; off our line |
| Operance | materials | 25 | 35 | 15 | 35 | D | AVOID | pure O&M asset handover; off our line |
| Catenda Hub | materials | 30 | 40 | 15 | 40 | D | AVOID | BIM coordination + handover; off our line |
| Autodesk Construction Cloud | platform | 20 | 82 | 20 | 85 | D | AVOID | proven absorber (bought Pype); owns distribution + spec corpus + agent platform |
| Trimble Construction One | platform | 22 | 72 | 25 | 80 | D | AVOID | owns ERP+PM+estimating; Viewpoint Finance Assistant could aim a finance agent at the commercial layer |
| Oracle Aconex+Primavera | platform | 20 | 70 | 20 | 80 | D | AVOID | strongest audit corpus; already ships cross-project predictive AI (Advisor for Safety) on captive data |
| Buildertrend | platform | 25 | 45 | 20 | 45 | D | AVOID | residential SMB distribution; thin AI; different buyer |
Coverage (0-100, map order 1-21)
| Competitor | 1 | 2 | 3 | 4 | 5 | 6 | 7 | 8 | 9 | 10 | 11 | 12 | 13 | 14 | 15 | 16 | 17 | 18 | 19 | 20 | 21 |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| DroneDeploy | 0 | 5 | 0 | 30 | 0 | 55 | 50 | 15 | 30 | 75 | 10 | 40 | 100 | 20 | 20 | 0 | 25 | 30 | 25 | 35 | 5 |
| Buildots | 0 | 0 | 0 | 20 | 25 | 5 | 30 | 30 | 55 | 90 | 10 | 45 | 85 | 15 | 10 | 0 | 5 | 10 | 15 | 40 | 5 |
| Disperse | 0 | 0 | 0 | 25 | 0 | 5 | 30 | 5 | 40 | 80 | 5 | 35 | 90 | 20 | 5 | 0 | 5 | 5 | 10 | 25 | 0 |
| XYZ Reality | 0 | 0 | 0 | 15 | 0 | 5 | 70 | 15 | 20 | 45 | 10 | 80 | 75 | 10 | 15 | 0 | 30 | 5 | 25 | 25 | 0 |
| Doxel | 0 | 0 | 0 | 25 | 0 | 0 | 40 | 20 | 60 | 95 | 5 | 50 | 90 | 45 | 10 | 0 | 5 | 15 | 15 | 30 | 10 |
| STACK | 5 | 95 | 15 | 10 | 10 | 0 | 5 | 20 | 0 | 5 | 10 | 10 | 0 | 55 | 0 | 0 | 0 | 5 | 0 | 25 | 15 |
| Togal.AI | 5 | 90 | 0 | 0 | 0 | 0 | 0 | 10 | 0 | 0 | 5 | 10 | 0 | 20 | 0 | 0 | 0 | 0 | 0 | 20 | 15 |
| Bluebeam Revu | 10 | 75 | 0 | 5 | 0 | 5 | 30 | 15 | 0 | 5 | 60 | 25 | 0 | 25 | 0 | 0 | 5 | 0 | 0 | 55 | 0 |
| PlanSwift | 5 | 85 | 5 | 0 | 5 | 0 | 5 | 10 | 0 | 0 | 5 | 5 | 0 | 35 | 0 | 0 | 0 | 5 | 0 | 10 | 0 |
| Countfire | 0 | 70 | 0 | 0 | 0 | 0 | 5 | 5 | 0 | 0 | 5 | 0 | 0 | 5 | 0 | 0 | 0 | 5 | 0 | 5 | 0 |
| SiteLogs | 0 | 0 | 0 | 55 | 15 | 5 | 0 | 0 | 5 | 5 | 5 | 0 | 0 | 0 | 0 | 0 | 0 | 15 | 0 | 5 | 0 |
| EasySiteLog | 0 | 0 | 0 | 80 | 20 | 10 | 0 | 0 | 0 | 5 | 5 | 0 | 0 | 0 | 5 | 0 | 0 | 0 | 10 | 15 | 0 |
| BuildPass | 0 | 0 | 10 | 80 | 45 | 75 | 60 | 50 | 45 | 30 | 55 | 0 | 0 | 15 | 10 | 0 | 15 | 40 | 30 | 50 | 0 |
| SmartBarrel | 0 | 0 | 0 | 45 | 90 | 30 | 5 | 25 | 0 | 35 | 0 | 0 | 0 | 35 | 10 | 50 | 0 | 30 | 15 | 30 | 5 |
| Fonn | 0 | 0 | 15 | 75 | 25 | 55 | 50 | 65 | 35 | 40 | 50 | 5 | 5 | 20 | 10 | 10 | 25 | 30 | 20 | 60 | 0 |
| Matrak | 0 | 55 | 40 | 20 | 0 | 0 | 35 | 20 | 30 | 60 | 10 | 35 | 0 | 25 | 10 | 0 | 5 | 95 | 10 | 40 | 5 |
| Sablono | 0 | 0 | 0 | 35 | 10 | 5 | 70 | 30 | 75 | 90 | 20 | 15 | 0 | 15 | 15 | 0 | 25 | 15 | 10 | 45 | 5 |
| Zutec | 0 | 0 | 10 | 20 | 0 | 35 | 75 | 30 | 5 | 10 | 70 | 20 | 5 | 5 | 10 | 0 | 95 | 15 | 55 | 30 | 5 |
| Operance | 0 | 0 | 5 | 5 | 0 | 25 | 10 | 25 | 0 | 0 | 40 | 40 | 0 | 0 | 5 | 0 | 95 | 30 | 45 | 25 | 0 |
| Catenda Hub | 0 | 0 | 0 | 10 | 0 | 5 | 45 | 40 | 10 | 20 | 65 | 95 | 30 | 5 | 10 | 0 | 35 | 10 | 10 | 50 | 0 |
| Autodesk Construction Cloud | 70 | 60 | 55 | 80 | 55 | 75 | 80 | 90 | 60 | 60 | 90 | 90 | 50 | 75 | 55 | 45 | 60 | 45 | 65 | 80 | 50 |
| Trimble Construction One | 35 | 75 | 55 | 70 | 65 | 55 | 60 | 85 | 80 | 55 | 70 | 85 | 45 | 85 | 45 | 90 | 45 | 60 | 50 | 70 | 40 |
| Oracle Aconex+Primavera | 30 | 25 | 55 | 50 | 30 | 80 | 70 | 90 | 95 | 60 | 95 | 55 | 30 | 85 | 50 | 40 | 60 | 35 | 80 | 80 | 55 |
| Buildertrend | 30 | 35 | 40 | 80 | 55 | 40 | 45 | 75 | 70 | 50 | 40 | 5 | 5 | 70 | 45 | 55 | 20 | 30 | 25 | 85 | 15 |
What the landscape sweep adds (25 more tools, ~38 total)
- Areas 15 and 21 stay empty at scale. Across the entire wider field — reality capture, estimating, field-daily, materials/handover, and the four big platforms — not one tool ships change/claims recovery (15) or cross-firm historical-cost benchmarking (21). The estimating tier confirmed it explicitly: the closest (Togal “predictive costing”, STACK via 1build) are single-firm or a bought-in catalog, never a compounding cross-firm loop.
- The build-on-top / absorption risk is now proven three times. Pype → Autodesk (captive feature), Trunk Tools’ Procore API revoked + rival agent shipped, and now Fonn → The Access Group (ERP roll-up) and Disperse → OpenSpace. Independent tool absorbed by a distribution-owner is the dominant exit, not a standalone scale-up. Reinforces: don’t be a feature on someone’s platform.
- A real partner/evidence ecosystem exists for the claims layer. The tools we would consume rather than compete with: Sablono (time-stamped actual-vs-plan on P6/Asta/MSP — the delay/EOT-claim backbone, QS buyer), Matrak (delivery/out-of-sequence = disruption evidence), DroneDeploy / Doxel (visual progress framed as payment-claim proof). These feed area-15 recovery; several are UK/EU, in our target geography. This is the supply side of our wedge.
- Platform AI is real but pointed away from the money. Autodesk Construction IQ, Trimble ProjectSight agents, Oracle Advisor for Safety are all GA and substantive — on risk-prediction, document extraction, summarisation. None target recovery or cross-firm cost. Oracle’s Advisor for Safety (cross-project predictive on 10,000+ project-years) is the proof-of-mechanism to watch — the same captive-data motion could later be aimed at cost/claims. That, plus the platforms’ bundling power, is the standing walk-away risk.
- One up-funnel mover to watch closely: BuildPass — the only field-capture tool already shipping agentic AI (Agent + an MCP server into Claude) and climbing toward commercial, yet it still stops before claim quantification. Closest analog to our own intended motion.
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
| Tool | Class | Area 15 | Verdict | Note |
|---|---|---|---|---|
| nPlan | a · predict | 10 | AVOID | huge schedule dataset; forecasts risk, not recovery |
| Nodes & Links | a · predict | 20 | AVOID | project-controls AI; stops at analysis |
| SmartPM | b · delay | 30 | FLANK | schedule analytics / delay |
| ClaimLogic | b · delay | 40 | FLANK | clean P6 delay engine (claimlogic.app) |
| ForensicPM | b · delay | 40 | FLANK | MCP-native forensic engine |
| ScheduleLens | b · delay | 40 | FLANK | LLM delay narrative |
| Steelray Delay Analyzer | b · delay | 40 | AVOID | priced expert-witness install base (~$3,990/yr) |
| Systech | b · delay | 40 | AVOID | consultancy delay-analysis software |
| ClaimMaster.ai | c · claims-mgmt | 60 | FLANK | defensible claims; stops short of quantum |
| Astora | c · evidence | 55 | FLANK | claim evidence layer (astora.app) |
| Magra | d · RECOVERY | 85 | FLANK | nearest competitor — Eichleay/Hudson/measured-mile, ~$36K/yr, owner+contractor (magra.app) |
| ClaimDD / ClaimEOT | d · RECOVERY | 85 | FLANK | one-person LLC, two apps; computes quantum |
| Masin AI | d · RECOVERY | 80 | FLANK | consultancy spin-out, strong team (masin.ai) |
| Delay Claim Builder | d · RECOVERY | 80 | FLANK | live; 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
| Tool | Builds | A14 | A15 | A21 | Verdict | Note |
|---|---|---|---|---|---|---|
| Causeway | QS commercial suite | 80 | 15 | 20 | AVOID | owns the QS desk; legacy, no real AI |
| Eque2 | ERP-attached commercial | 80 | 15 | 10 | AVOID | invoice OCR only |
| RedSky | ERP commercial | 80 | 12 | 10 | AVOID | ERP roll-up |
| Cleopatra Enterprise | EPC cost control | 90 | 35 | 25 | AVOID | change mgmt, not entitlement |
| RIB CostX | estimating + commercial | 55 | 20 | 25 | AVOID | estimating base |
| RIB Candy | civils EVM + commercial | 75 | 25 | 15 | AVOID | EVM; stops before entitlement |
| Planyard | cloud CVR challenger | 85 | 25 | 15 | FLANK | cloud-native CVR |
| Payapps (Autodesk) | subcontractor valuations | 60 | 20 | 5 | AVOID | absorbed by Autodesk (4th absorption) |
| Gather Insights | comp-event detect + notices | 35 | 55 | 10 | FLANK | front-half mover — NEC/JCT events + notices, our QS buyer, UK; stops before quantum |
| Gauge | AI benchmarking (single-firm) | 60 | 5 | 55 | FLANK | single-firm only |
| Rate QS | AI rate library (single-firm) | 65 | 5 | 55 | FLANK | markets “private, not cross-firm” as a feature |
| BenchIt | AI cost benchmarking (single-firm) | 60 | 5 | 50 | FLANK | single-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).