Dashboard: Dashboard · grid: _MARKET-PROBLEM-MAP · lens: _OPPORTUNITY-LENS · cross-competitor: _CROSS-COMPETITOR · worked tone example: kreo/dossier
This is the cost side of pre-construction: turning drawings into quantities, then quantities into a price. It is the same lane Kreo occupies, and the conclusion from the Kreo brief carries across the whole category — it is adjacent to our wedge (recover money from variations; reuse cost history across firms), not the same as it. These tools price the job before it starts; our thesis recovers money and learns from jobs after they run. The single point of overlap is area 21 (historical-cost benchmarking), and the scan below confirms it is unoccupied here in shipped form, just as it was for Kreo.
The category splits cleanly into two generations. AI-native (Togal.AI, and Kreo already covered) lead with computer-vision auto-measure as the product. Legacy desktop (PlanSwift, Bluebeam Revu) were built as manual on-screen-takeoff tools and are now bolting AI symbol-detection onto a 2D markup engine. STACK sits in the middle — a cloud takeoff-and-estimating platform, AI-assisted but not AI-first. Countfire is a narrow vertical (electrical/MEP symbol counting). None of them is our buyer’s tool: every one sells to the estimator / quantity surveyor / sub in the pre-contract commercial team, and every one stops at the priced estimate or the raw quantity.
Per-tool mini-profiles
STACK (stackct.com) — cloud takeoff + estimating, AI-assisted
STACK is a cloud preconstruction platform: digital quantity/labor/material takeoff plus assembly-based estimating and proposals, sold per-seat to GCs, subs and specialty contractors. Pricing is published and high for the segment — roughly $1,999/yr (Start, 1 full user) to $4,999/yr (Grow, 3 users) on the platform tier, with the takeoff-and-estimate product around $2,599–$2,999 per user/yr (per-seat cost falls as seats are added); an AI-bearing “Premium”-class tier carries GPT-powered project chat and AI-assisted autocount. Crucially STACK is the only tool here that reaches real cost data — but via a third-party integration with 1build (an API to ~68M live national/regional cost data points), i.e. a commercial cost catalog plugged in, not a cross-firm benchmark built from its own customers’ project outcomes. It also advertises “AI-powered material waste prediction” using current material pricing. Reviews praise takeoff speed and deep assembly libraries; complaints are the learning curve, weak sub-bid-solicitation, and year-over-year price hikes.
Togal.AI (togal.ai) — AI-native takeoff, the closest analog to Kreo
Togal is an AI-first takeoff tool: upload architectural drawings, computer vision auto-detects, measures and labels spaces/elements (“up to 97–98% accuracy”), plus Togal.CHAT, an LLM panel to “talk to” the plans. Buyer is the estimator / GC / sub, strongest on multifamily, retail and commercial; it is explicitly not a full estimating platform — no built-in CSI cost database, no assembly pricing, no sub-bid management (an independent 2026 review states this directly: “Togal gives you quantities… pricing is your problem to solve with other tools”). Pricing is published per-seat: $299/user/month (Growth, billed yearly), with a custom Business tier for 4+ users and Enterprise. It integrates out to Procore, Bluebeam, Sage and Destini (Procore Marketplace listing), so quantities export to those tools for pricing. Togal’s own marketing claims “predictive costing” that “learns from every job” — but read precisely this is single-firm (your own past projects) and reads as announced/marketing language, contradicted by the independent review that finds no cost database shipped. This is the same pattern as Kreo’s “intelligent benchmarking”: single-firm, announced, not a cross-firm market dataset.
Bluebeam Revu (bluebeam.com) — legacy 2D markup engine, now AI-on-top
Revu is the long-standing AEC PDF markup/measurement/collaboration tool; takeoff is one use of its measurement engine, done manually — the estimator draws markups to measure lengths/areas/volumes, then pushes them live into Excel via Quantity Link (bidirectional PDF-to-spreadsheet) where unit costs are applied by hand. Its newer assists — VisualSearch (count matching symbols) and Dynamic Fill (auto-trace irregular areas) — are rule-based, not ML; a 2026 Bluebeam Max AI subscription ($590/user/yr intro) layers AI drawing-review/workflow automation and “AI-driven symbol detection” on top. Per-license pricing is published (Basics ~$260, Core ~$330, Complete ~$440/user/yr; older Standard/CAD/eXtreme $349–$599). No cost database, no benchmarking — it stops at quantities and hands off to Excel. Buyer spans estimators, architects, field. Owned by Nemetschek.
PlanSwift (planswift.com, by ConstructConnect) — legacy desktop takeoff, AI bolt-on
PlanSwift is a Windows-desktop drag-and-drop takeoff-and-estimating tool for GCs and specialty trades (concrete, electrical, HVAC, plumbing, +10 more), priced as a low-cost license (~$1,595–$2,000/yr range cited). It now ships “Takeoff Boost,” an AI suite — auto-measure areas/lengths, auto-count symbols across a sheet set, auto-scale, auto-bookmark — so its AI claims are real but recent and grafted onto a legacy core. It has a Plugins page (extensibility exists, technical depth not verified). Reviews like the intuitive interface and templates; recurring complaints are crashes/lag on large files and no Mac. Owned by ConstructConnect (Roper/Thoma Bravo lineage). No cost-benchmarking or cross-firm cost data found; downstream costing is the user’s spreadsheet.
Countfire (countfire.com) — narrow vertical, automated symbol counting
Countfire is a cloud takeoff tool tightly focused on automated symbol counting for electrical and M&E estimators — count thousands of symbols across many PDFs in seconds, auto-count revised drawings on design change, plus length/area/pipework measures. Its automation reads as rule-based pattern/symbol recognition rather than ML (it does not heavily brand itself “AI”). Pricing is custom/quote-based (entry estimates ~$99–$200/month cited); ROI is pitched as estimator-days saved (UK framing, £/day). Buyer is the electrical/M&E sub-estimator — a single trade, not a platform. No cost database, no benchmarking, no API/integration surfaced. The narrowest tool in the set and the furthest from any commercial-cost layer.
Comparison
| Tool | Builds | Buyer | AI angle (shipped vs announced) | Open API? | Touches historical-cost (21)? | Read |
|---|---|---|---|---|---|---|
| STACK | Cloud takeoff + assembly estimating + proposals | GC / sub / specialty estimator | AI-assisted: GPT project chat + AI autocount (shipped, gated to top tier); “material waste prediction” | Partial — integrations incl. 1build cost-data API; not an open dev platform | Third-party cost catalog only (1build national data plugged in) — not cross-firm benchmarking from its own users | NEIGHBOUR — cost-side, but cost data is bought-in, not a compounding loop |
| Togal.AI | AI-native takeoff (auto-measure/label) + plan-chat LLM | Estimator / GC / sub (multifamily, commercial) | AI-first CV auto-takeoff + Togal.CHAT (shipped). “Predictive costing from past jobs” = single-firm, announced/marketing | Integrates OUT (Procore, Bluebeam, Sage, Destini); no open public API found | No (shipped) — markets single-firm “learns from every job”; independent review finds no cost DB. Same as Kreo’s pattern | NEIGHBOUR — closest analog to Kreo; flank, don’t rebuild |
| Bluebeam Revu | 2D PDF markup + manual measure + Excel link (Quantity Link) | Estimator / architect / field (AEC-wide) | Manual measure; VisualSearch/Dynamic Fill rule-based; Bluebeam Max adds AI review/symbol-detect (2026) | App ecosystem; no open takeoff data API surfaced | No — stops at quantities → Excel for costing | LEGACY desktop incumbent; not AI-native, not our buyer |
| PlanSwift | Desktop drag-drop takeoff + estimating | GC / specialty trades (Windows-only) | “Takeoff Boost” AI: auto-measure / auto-count / auto-scale (shipped, recent bolt-on) | Plugins page exists; depth unverified | No | LEGACY desktop (ConstructConnect-owned); AI grafted on |
| Countfire | Automated symbol-counting takeoff | Electrical / M&E sub-estimator | Automated symbol counting (rule-based; not branded AI) | None surfaced | No | Narrow single-trade tool; furthest from commercial-cost layer |
App-store note: none of the five has a genuine mobile app — the app-store pulls returned only unrelated games/flight-trackers (these are web/desktop/Windows products). That is a real finding, not a data gap: estimating/takeoff is a desktop-screen job, not a field-phone job.
What this category tells us
- The cross-firm cost-benchmark moat (area 21) is empty here — confirmed, not assumed. Not one of the five ships a cross-firm, market-wide cost dataset built from its own customers’ project outcomes. The two that come closest both fall short in the exact way Kreo did: Togal markets single-firm “predictive costing that learns from every job” (announced/marketing, and an independent review finds no shipped cost DB), and STACK reaches real cost data only by plugging in 1build’s third-party national catalog — bought-in reference pricing, not a compounding proprietary loop. Single-firm reuse or a licensed catalog is reuse-within-an-account / commodity data, not the cross-firm intelligence our thesis is built on. The moat column is unoccupied across the entire estimating-takeoff category.
- Area 15 (change/variations/claims) is entirely absent. Every tool stops at the priced estimate or the raw quantity. None tracks the job once won, sees actual vs estimate, or has any notion of a variation or claim. This whole category lives strictly pre-contract; the money-recovery wedge is off-axis for all of them.
- AI-native vs legacy is a clean split, and computer-vision takeoff is commoditising. Togal (and Kreo) are AI-first CV engines; Bluebeam and PlanSwift are legacy desktop tools bolting symbol-detection and “AI” suites onto a manual-measure core in 2026; STACK is cloud-but-AI-assisted; Countfire is rule-based. Auto-measure/auto-count is now table stakes being added everywhere — which confirms the Kreo read: do not build a cheaper AI takeoff, the automation IS the incumbent and the capability is racing to commodity.
- The “talk-vs-ship gap” on cost intelligence is the consistent tell. Across the category, vendors talk historical/predictive cost (Togal’s “learns from every job,” STACK’s “material waste prediction,” echoing Kreo’s “intelligent benchmarking”) but what ships is either single-firm or a licensed third-party catalog. The strategically interesting claim — cross-firm benchmarking — is announced everywhere and shipped nowhere. That is a wide-open window on our moat, told by the market itself.
- Distance to our wedge is large but the relationship is friendly. These are cost-side neighbours, not targets. Don’t rebuild any of them. A contractor who has priced a job in STACK/Togal/Kreo has produced a clean, structured estimate — the perfect baseline to measure variations and actuals against. The posture is consume-their-output / partner-on-data (most integrate OUT to Procore/Sage/Excel already), not build-on-their-API. Togal is the closest analog to Kreo and the same verdict applies; STACK is the one to watch because its 1build tie shows appetite to move toward cost intelligence, though via a catalog rather than a cross-firm loop.