ScopeGuard, in order.
Find the gaps between trade scopes before somebody prices the job around them.
ScopeGuard
Find the gaps between trade scopes before somebody prices the job around them. Included from Crew upwards.
1. Load the buyout
The trades on the job, their scopes and their budgets.
2. Read the gaps
Work no trade has claimed, work two have claimed twice, and the coverage figure for each trade — with a trade nobody has priced saying so rather than reading as complete.
3. Close them
A clarification per gap, drafted against the trade it belongs to.
What it will not do
- It reads the scopes it is given. A trade that was never scoped cannot have its gaps found.
- Coverage is against the published checklist, not against the drawings.
On a paid plan every program opens a sample set, so none of this has to be read before trying it. A free trial starts blank, on your own drawings. See the sample project.
In detail
In short
Twenty trades quote your job and twenty estimators write their own exclusions. Somewhere in that pile is work the specification requires that nobody priced, work two trades are both carrying, and wording that will be read against you in an adjudication. ScopeGuard reads all of it against the master documents and puts a number on every gap — while you can still change the award. Opens in a browser on any machine. From $299 a month.
Fifteen letters. One afternoon. Days before award.
What comes back from the trades is prose — inclusions, exclusions and qualifications written by fifteen different estimators, each protecting their own number. Most GCs review it by hand, with a checklist, under time pressure.
The failure is never dramatic. It is one sentence in the concrete letter that puts winter heating on somebody else, on a job whose structure runs November to March. Nobody notices until the first cold morning, and by then the price is a change order instead of a negotiation.
Silence is the expensive part. The most costly item on a buyout is the one no letter mentions at all, so ScopeGuard reports absence as loudly as it reports a bad clause.
Winter concreting — heating, hoarding, curing blankets, accelerator — excluded
Vaughan Forming & Concrete Ltd has explicitly excluded work that Section 03 30 00 — Cast-in-Place Concrete requires. Unless another trade has carried it, this work has no price behind it. Read from your specification at page 2.
Master: Section 03 30 00 — Cast-in-Place Concrete · CSA A23.1 cold weather requirements; GTA Dec–Mar Proposal: 04-CONC_Vaughan_Forming_scope.pdf p. 1 — Exclusion Confidence 95% · Basis: 1,450 m³ × $96.00 — cold-weather premium (Toronto 2026 rate table)
One finding out of 128 on the demo buyout. Nothing here is written by a model.
Eight ways a buyout leaks.
Every finding carries a plain-language explanation, a citation on both sides — the clause in your master documents and the page and line in the sub's letter — a confidence score, and a cost exposure range with its stated basis. The counts below are what the demo buyout actually produced.
Required by the master documents, and the sub explicitly took it out. The fire pump, the winter heating, the fire alarm verification written as "by others".
Wording that will not survive a dispute — "as required", "standard", "allowance", "to suit". Twenty patterns that reliably become an argument.
Two or more trades carrying the same work. You pay for it twice, or you negotiate one of them down and give the other a reason to claim.
Work in the gap between two trades. Carpentry says blocking is drywall's; drywall says it is carpentry's. Twenty-two of these recur on jobs anywhere.
Priced against a superseded document. Four proposals on the demo job never acknowledged Addendum 3.
A missing or unusable rate, or a price with a validity window that closes before your award date.
Required by the master documents and simply absent from the letter, not excluded, not mentioned. The one nobody catches by reading.
Falls short of your own buyout standard — your blocking rule, your access-panel rule, your hoisting rule. You write these; they are checked on every letter after that.
128 open findings, 78 of them high severity, $3,198,498 of mid-case exposure against a $14,250,000 contract — 22.4%. Sixteen documents, one afternoon of review compressed into the time it takes to make coffee.
What it does, screen by screen
The figures in this section are the shipping application running the demo buyout that ships with it: Lakeshore Yards, a 38-unit residential infill at 118 Jefferson Avenue in Toronto — $14.25M, twenty trade packages, fourteen proposals received, a full project manual and three addenda. The figures come from a fixture so you can check them against what you see here; the reader, the extractor and the cross-reference engine are the ones your own documents go through.
Where the money is leaking, ranked.
One screen answers the question you actually have: which trades to worry about, and how much. Exposure is shown per trade as a share of that trade's own quoted price — fire protection at 120% of its own quote is not a rounding error, it is a bid that priced a different job.
- A single buyout risk score with the reasoning printed beside it
- Trades still outstanding are listed separately and kept out of the exposure figure — an unreviewed trade is unmeasured risk, and unmeasured is not zero
- "Fix before you award" pulls the highest-exposure findings to the top of the page
Cited on both sides, or it isn't a finding.
Every entry names the clause in your master documents that requires the work and the page and line in the sub's letter that deals with it. Below that: what to do about it, in a sentence you could paste into a letter. Ranked by severity, then by exposure.
- Every dollar figure states its basis — quantity, rate, and the share of that rate the requirement represents
- The range is an argument, not a hedge: the low end is buying it cooperatively now, the high end is buying it as a change order after mobilisation
- Mark any finding real or noise — that verdict persists and sharpens the next project
- Filter by trade, severity or category; export the whole register to CSV
Your spec on the left, their words on the right.
The review a good estimator does in their head, written down. Every master requirement for the trade, its treatment in the letter — included, excluded, qualified or never addressed, and where a treatment costs money, the sub's own sentence quoted verbatim beside it.
- Nothing paraphrased. The words you negotiate against are the words they wrote
- Requirements with no finding are shown as carried, so the clean part of a letter is visible too
- Exports as a per-trade scope report, or as revised scope language ready for the subcontract
A scope letter is not a data format. It becomes one.
Drag the proposals onto the window — PDF, Word, Excel, CSV or a saved email. Each is read, classified by trade and split into inclusions, exclusions, qualifications, unit rates and alternates, with the page and an extraction confidence on every line.
- Addenda acknowledged is tracked per letter — 3 of 3, or the gap that is about to cost you
- Revisions stack: import Rev 2 and the register updates in place
- Every extracted line is editable. The engine reads; you rule
One click turns findings into the letter you'd have written.
Every open finding on a trade, numbered, each with the clause that requires the work and the exact question to put to the sub. Paste their reply back in and ScopeGuard reads what it actually did, which items were withdrawn, which were priced, and which were answered with more prose.
- Draft for one trade, or for every trade with open findings at once
- Your standard scope language — blocking, firestopping, hoisting, winter, access panels, cutting and patching — is a library you edit once and it is used from then on
- Export the letters as PDF, and the final accepted scopes for the project file
It learns your losses
Load your change-order history and the failures that actually cost you money get checked against every future letter. Alongside it: eighteen documented regional dispute hotspots, and your market's price movement by division, because a 30-day price on plumbing running 11.9% year-over-year is a buyout risk worth pricing.
The evidence stays with the job
Every imported document is copied into the project's own folder. A finding that cites page 4 of a proposal is worth nothing once that proposal has moved off somebody's desktop, so it doesn't. Master package, addenda and every proposal, kept where the citations still resolve.
A range you can't audit is a range you won't use.
Nothing is invented
Exposure comes from your market's subcontract rate table applied to your project's own quantities. Severity comes from requirement criticality and your own dispute record. Confidence comes from the match score. MaraponeAI explains, prioritises and drafts — it never decides a number.
Every figure states its basis
Each finding prints the quantity, the rate, the share of that rate the requirement represents, and the stage multiplier behind its number. You can check the arithmetic in front of the sub, which is the only reason a number is any use in a negotiation.
Silence is a finding
The most expensive item on a buyout is the one no letter mentions. Absence is reported as loudly as a bad clause, and trades that have sent nothing at all are listed by name, with their budget, so the hole is visible on the page.
You overrule the engine
Every finding can be dismissed, every extracted line edited, every rate overridden. Dismissals and overrides persist and compound: the second project is measurably quieter than the first because you taught it what noise looks like on your jobs.
A named city, not a continent.
Scope disputes are local. The grey zones, the standard exclusions, the wording that gets used because it has worked on jobs in your market before — all of it is jurisdiction specific, which is why a tool sold as covering "North America" and nothing more precise is telling you it knows none of it.
ScopeGuard runs in any Canadian province, US state or European country, because the scope taxonomy is CSI MasterFormat and the engine reads whichever jurisdiction pack you point it at. The numbers below are the Toronto pack — the finished one, and the standard your market's pack is built to during your engagement. See where each market actually stands before you buy on the strength of your own. Every file is plain JSON, readable and editable, with its source printed on it.
Worked example — the Toronto & Ontario pack, shipped and in production
Trade packages. 20 packages with 215 master scope requirements mapped to CSI MasterFormat, each with its typical share of contract value and a Toronto-specific note on where that trade normally leaks.
Interfaces. 22 interface conditions — blocking, access panels, firestopping, backing, hoisting, cutting and patching, and the rest of the grey zones between two trades that recur on jobs in any market.
Language. 20 language patterns that detect qualifications which will not hold up, plus the inclusion, exclusion and qualification cues that make extraction reliable on letters nobody formatted for a machine.
Dispute record. 18 documented regional dispute hotspots with frequency, typical exposure and the wording that prevents each one — checked against every scope you import.
Statute & contract. The statutory layer for your jurisdiction — shown here as the Ontario Construction Act holdback, prompt payment and adjudication, CCDC 2 with CCA 1 subcontract, and the 2024 Ontario Building Code.
Municipal. Toronto Green Standard v4, Green Roof By-law Ch. 492, Noise By-law Ch. 591, Sewer Use Ch. 681, Backflow Ch. 851, Tree By-law Ch. 813, O. Reg. 406/19 excess soil, O. Reg. 278/05 asbestos, and the ESA, TSSA and AODA gates.
Rates. Toronto 2026 subcontract rate table, revision 2026.1, in CAD excluding HST — the basis every exposure figure is calculated from, and editable line by line.
Market. Q4 2025 Ontario non-residential price movement by division, so a firm price shorter than the gap to your award date is flagged for what it is rather than assumed to hold.
The spine is CSI
All 215 requirements are indexed by CSI MasterFormat across 21 divisions — the taxonomy a US or Canadian estimator already reads. The Ontario layer is 59 local_ref entries and one note per trade, in named fields. Roughly three quarters of the knowledge base — hoisting responsibility, cut-and-cap, submittal cycles, as-builts — is disputed the same way on any job, anywhere.
Six steps, and the fifth one is the product.
Create the project
Name, contract value, stage, award date. Every trade package is added with its budget apportioned from the contract value, or imported from the estimator's division rollup.
Import the master package
Specifications, drawing list, addenda. Spec sections are matched to trade requirements so every finding cites your own documents.
Drop in the scope letters
Drag them onto the window. Each is read, structured and classified by trade automatically, no template, no naming convention, no data entry.
Review the findings
Ranked by severity then exposure. Mark each one real or noise; that verdict trains the engine for the next project.
Send the clarifications
This is the step that saves the money. Every open finding becomes a numbered question with the clause behind it — asked while the sub still wants the job.
Export and award
Buyout risk summary (PDF), a living register (Excel, seven sheets), a scope report per trade, the clarification letters, and the final accepted scopes for the project file.
The model never decides a finding.
A tool that hands you 128 findings has to be right about them, because the first hallucinated exclusion you take to a subcontractor is the last time anyone in your office opens the app. So the cross-reference engine is deterministic: requirement matching, treatment classification, overlap and interface detection, exposure and severity all run on rules and your own data.
MaraponeAI explains, prioritises and drafts. Every figure it quotes traces back to a clause, a page and a rate. Ask it show me everything over $50k on mechanical and the instruction is parsed by rule and previewed before it changes anything on screen.
The engine underneath is deterministic: every finding traces to a clause, a page and a rate you can check. Our own model sits on top of that, adding the free-form narrative, and it never gets to decide what was found.
Your workspace, and only yours
Your subcontractors' pricing is the most commercially sensitive data in your office, so it is read in a workspace scoped to your account by our own model, no third-party AI vendor receives it, no other firm's work is pooled with it, and nothing is trained on it.
Fully functional without a model
Every finding, explanation and clarification is available from the deterministic engine alone. Without a model you lose free-form narrative — you do not lose a single finding or a single dollar figure.
The buyout record survives the subscription
Documents, findings, clarification letters and the scope language you have edited over time live in your workspace and export as Excel and PDF whenever you want them. A buyout argued two years ago should still be readable when it comes back, whether or not you are a customer.
The knowledge base is yours
Trade scopes, the interface matrix, the language patterns, the regional intelligence and the rate table are plain JSON in the app's resources folder. Edit them, extend them, keep them. Every rule is visible in the program and exports to CSV, so what it checked against is auditable.
A finished program, open in a browser tab.
One flat price per firm. No per-file charges, no metering, no implementation fee. The working program the minute you sign in, and you can cancel from your dashboard the month it stops earning its keep.
One finding on the demo buyout — winter concreting excluded by the concrete trade, on a structure running December to March — carried $118,320 to $201,840 of exposure. It was one of 128 findings on one job.
The Crew plan is $299 a month and runs on every job after that one. That is the whole value argument: you are not buying software hours, you are buying the finding you would otherwise pay for as a change order.
- The complete program the moment you sign in — nothing to install, no licence key, no machine to provision
- Unlimited projects, buyouts, documents and machines in your company. No per-seat or per-project fee, ever
- All 215 master requirements, 22 interface conditions, 8 finding types and the 2026 rate table for your jurisdiction
- The whole knowledge base as editable JSON — your trade scopes, your language, your rates
- Every improvement the day it ships, no version to buy, no upgrade to schedule
- Read by our own model in your own workspace, no third-party AI vendor, no training on your work
- Cancel any month from your dashboard, no long-term contract — see the plans
- Fine-tuning on your historical buyouts, your subcontract templates and your house language
- A rate table rebuilt for a market other than the one you bought, or for self-perform divisions you price differently
- Wiring into Procore, Sage, Bluebeam or your own project database
- Bespoke development beyond the eight programs — that is quoted as project work
Buying the product first doesn’t lock you out of a build later — most teams start here, then commission the parts that are specific to them. See custom builds →
Or all four pre-award programs on the Crew plan: $299/mo — the Blueprint Auditor, SpecChecker, the AI Estimator alongside this one, plus one setup call that runs a real job of yours from the drawing set to the awarded subcontract. Need the Bid Leveler and the post-award suite too? The Pro plan is $499/mo for all eight programs — FloatGuard, TradeVet and CostLine included.