Strategic Go / No-Go Review — Commercial Memory Layer

Date: 16 June 2026
Evidence base: strategic-review-pack.zip, including all 12 competitor dossiers, _CROSS-COMPETITOR.md, _LANDSCAPE-SYNTHESIS.md, _landscape/claims-delay/note.md, _landscape/commercial-qs/note.md, construction-sop.md, and external public market/statutory sources for UK, US, Canada and Australia.

One-line VERDICT

VERDICT — SERVICES-FIRST, with Australia as the preferred first beachhead. Build this first as a productised commercial-recovery service with software underneath, not as standalone venture SaaS. The critical uncertainty is not whether AI can draft claims; it is whether contractors will trust you with adversarial commercial judgement, pay against recovered money, and grant rights to aggregate outcome data.

Bluntly: the geo correction strengthens the services-first verdict. Australia gives the cleanest evidence-to-cash loop because statutory Security of Payment adjudication turns substantiated commercial events into a near-term payment weapon. The US is the biggest market, but it is a worse first market because Magra is already closest to the recovery wedge, Procore/Raken/Trunk own much of the data surface, and litigation/change-order culture makes sales and proof cycles longer.

Plain-language primer — read this before the analysis

This section exists so someone with zero construction-claims background can understand the decision above. The rest of the memo assumes the jargon; this part does not. If a term below feels obvious to you, skip ahead.

What “commercial recovery” actually means

It is not insurance (no premiums, no risk pool, no payout-on-disaster) and not debt collection (not chasing an agreed-but-overdue invoice).

It is: getting back money a builder already earned but never collected, because the paperwork needed to claim it was never done properly.

A builder signs a contract for a fixed scope, in fixed conditions, by a fixed date, for a fixed price. Then reality drifts from that deal — the client changes the design, blocks site access, issues drawings late, adds scope. Each drift costs the builder real time and money. The builder earned that extra money. But to actually collect it he must (1) notice the event inside a contractual deadline, (2) prove it with contemporaneous evidence, (3) quantify it, and (4) file a claim. Builders are bad at this — crews build, they do not file paperwork — so they leak money they were legally owed. Recovering that leak is the product.

Why the builder is owed the money (the part that confuses everyone)

The builder is not paid “because the plan changed.” He is paid because the contract he signed already promises it — when the change is the client’s doing. Standard construction contracts (AS 4000 in AU, NEC/JCT in the UK, AIA in the US) have built-in machinery for exactly this:

• Variation — client changes scope → client pays for the extra work the builder never agreed to at the original price.
• Extension of Time (EOT) — client-caused delay → builder gets more days, and often the cost of standing idle (crew, cranes and site sheds cost money per week, even while producing nothing).
• Disruption / lost productivity — work the client made harder than priced → builder recovers the extra labour hours.

The whole thing hinges on one word: causation — who caused it?

This is why the Commercial Event schema later in this corpus has a causation field. The entire product is proving the cause was the client’s, not the builder’s. If the builder caused it, there is nothing to recover — that is not what we chase.

Is this legal territory? Yes — and that is the central constraint

It touches law, so the line matters. Three different activities, only some of which we do:

1. Assembling evidence + drafting the claim — this is commercial / quantity-surveying (QS) work. Claims consultants and QSs do it every day with no law licence. This is our lane.
2. Filing a statutory payment claim / SoP adjudication — in Australia this is designed to work without lawyers (Victoria literally markets Security of Payment as recovering payment “without needing lawyers”). The builder files it himself. This is a core reason AU was chosen.
3. Giving legal advice / representing in a dispute / litigation — this is regulated legal practice. We must never cross here.

So the play is parked deliberately in the QS/commercial lane, not the lawyer lane. We help substantiate and assemble a right the builder already holds under his contract; we stop short of legal advice or representation. That is why the model is “services-first with human review,” why human sign-off is hard-coded into Notice, EntitlementPosition and ClaimRecoveryPack, and why Stage 0 requires AU construction counsel to confirm the boundary before anything material is built.

What “services-first” actually means (a common misread)

“Services” here does not mean “we consult and advise them how to do it.” It means we produce the finished recovery pack for them — done-for-you, as opposed to selling self-serve software and walking away. Software runs underneath the whole time; the customer simply does not operate it alone yet.

The distinction is done-for-you service vs do-it-yourself SaaS — not advice vs software.

The thing people get backwards: the app comes LAST, not first

The intuitive picture is: builder starts a project → installs our app → tracks everything live → when something goes wrong, we react. That is the end state (Stages 3–4, 12–24 months out), not the start.

The start (Stage 1) is deliberately more manual and uglier:

Stage 1 (0–3 mo):   builder hands us a messy pile from an already-broken
                    (or already-finished) project. Humans + light software
                    reconstruct the commercial events, draft a recovery pack,
                    a QS reviews entitlement & quantum, the builder files it.
                    NO app. NO live tracking.

Stage 2 (3–6 mo):   same loop, made repeatable, with AU-state rule-packs.

Stage 3 (6–12 mo):  NOW build the app — live capture + integrations with
                    BuildPass / Procore / Raken / email / WhatsApp. Catch
                    events as they happen instead of digging them up later.

Stage 4 (12–24 mo): software does most of the pack assembly itself.

Why start ugly? Because if we build the slick capture app first, we have built a diary app nobody pays serious money for — BuildPass, Raken, Procore and WhatsApp already own capture. Stage 1 proves the only thing that matters: can we turn a mess into recovered cash? We learn which evidence actually wins claims first, then build the tool to capture it. Build the capture app last, not first. (The memo’s blunt version: optimising the sequence for demos “loses the money.”)

Who reviews the work

The primary reviewer is a QS / commercial-claims expert, not a lawyer — because the hard judgement (is this entitlement valid? is the quantum right?) is quantity-surveying, not legal. A lawyer is used only for the legal boundary (are we allowed to do this; do success fees create a conflict), not to review every claim.

Why does this opportunity even exist — why is there no dominant player?

First, correct the premise: the field is not empty. Magra (US), Gather (UK) and several delay/quantum micro-tools already exist. The real gap is that no scaled, trusted, jurisdiction-specific recovery platform with cross-firm outcome data owns the market yet. And it is not a new law — AU Security of Payment dates to ~1999. So this is not “nobody noticed.” It is hard in all the ways that scare off fast-moving software founders:

1. It does not scale like SaaS. The core judgement is welded to one jurisdiction’s law and contract forms. AS 4000 (AU) ≠ NEC/JCT (UK) ≠ AIA (US) — each is a near-rebuild, not a translation. The thing that makes it defensible (a legal/regime moat) is the same thing that makes it hard to scale cheaply.
2. It is judgement-heavy, adversarial and liability-laden. It is a contested position against a counterparty who fights back. An unreviewed AI answer is not a product, it is a liability generator. So it cannot be clean self-serve SaaS — it needs humans in the loop, which looks like consulting and repels pure-software founders and investors.
3. Trust is earned slowly. A commercial director will not bet recovered cash on a cold AI output from a startup. Trust comes from outcomes, one matter at a time, and cannot be shortcut with a demo.
4. The real moat (cross-firm benchmarks) has a cold-start problem. It needs ~50+ projects and 500+ events before a benchmark means anything — and firms must grant rights to sensitive cost/margin data, which they refuse. Rate QS, Gauge and BenchIt all exist and stay single-firm; they market privacy as the feature. Whoever solves the trust + governance to pool data wins. Nobody has.
5. Inputs are a mess. Every project is “bespoke archaeology” — scattered diaries, emails, photos, cost ledgers, no standard. If it takes more than ~40 expert hours per diagnostic, the unit economics die. Turning that chaos into cheap, repeatable output is the actual hard engineering/AI bet.
6. The founder profile is rare. It needs someone who is both construction-commercial / QS / claims native and able to build AI/software. Those two rarely live in one team.
7. Incumbents sit upstream and do not want this layer. Procore/BuildPass/Raken own low-liability capture and stop before the high-liability recovery judgement. And narrow tools get absorbed (Pype→Autodesk, Payapps→Autodesk, Document Crunch→Trimble) — so the only durable position is owning the recovery P&L line plus the outcome data, which nobody has built yet.

The net: this is not virgin land nobody found. It is land that is hard to farm — everyone planted one corner (US notices, UK detection) and quit before the hard part (quantum + recovery outcomes + cross-firm data). The bet is that whoever survives the slow trust-and-data grind owns the whole field. The moat exists because it is hard, not despite it. That is the entire reason for services-first, Australia-first, human-reviewed: it is the sequence built to survive the hard parts long enough to build the moat nobody else has.

Market reality & sizing

The market is not “construction software”. It is jurisdiction-specific commercial recovery.

The product is not portable in the way a field-photo app is portable. The product’s core judgement layer depends on contract form, notice regime, adjudication/payment law, and the accepted quantum methods in that market.

Tender baseline
   ↓
Contract obligations / notice regime
   ↓
Commercial Event
   ↓
Entitlement position
   ↓
Quantum: time + cost + causation
   ↓
Recovery route: payment claim / adjudication / final account / claim / litigation
   ↓
Outcome data → future pricing and benchmarking

That means the correct strategy is single-beachhead density first, not shallow multi-country launch. The cross-firm historical-cost moat only works after enough comparable projects, contract forms, event categories, cost structures and recovery outcomes accumulate inside one local market.

Beachhead ranking: UK vs US vs Canada vs Australia

Scoring below is strategic, not just market size. I weight the criteria as follows because the first market must prove recovery economics before it proves global TAM:

Willingness-to-pay / ACV                      30%
Claims + entitlement regime / recoverability  30%
Competitive density                           20%
Cold-start ease                               20%

Geo notes

Australia — recommended first beachhead

Australia has the best first-market shape because the statutory payment-dispute route is operational, short-cycle and evidence-heavy. Security of Payment regimes vary by state, but the market has a clear “pay now, argue later” logic. Queensland’s QBCC annual report frames adjudication as a timely, efficient and cost-effective way to resolve payment disputes, and its 2024–25 figures show real monetary flow: 324 valid adjudication applications, A$516.7m in claim value, 214 decisions released, A$957.0m claim value in decisions released, and A$81.9m awarded under adjudication.¹ Victoria describes Security of Payment as a fast and inexpensive process for recovering progress payments/performance securities without needing lawyers.² NSW publishes quarterly adjudication performance reports, which is useful because public volume data exists at the state level.³

The macro base is large enough. ABS reports construction employment of 1.291m and construction industry value added of A$177.1bn in 2023–24.⁴ Infrastructure Australia reports 452,820 construction businesses as of June 2024, with 98% under 20 employees and 55% turning over less than A$200k.⁵ That looks fragmented, but the fragmentation helps the service wedge: the target is not the 98%; it is the thin layer of medium commercial builders, fit-out contractors, civil contractors and specialist subcontractors with enough payment leakage to pay.

Why AU beats the UK first: the UK has a more familiar QS language, but AU gives a cleaner “commercial event → payment claim/adjudication → recovered cash” loop. That is exactly what a services-first business needs to prove.

Main AU threat: BuildPass. The dossier shows BuildPass has shipped more real field/admin AI than most peers, is AU/NZ native, has ~400 companies, A$ pricing, free worker/sub access, and a credible MCP/agent posture. It does not do claims/quantum, but if you beachhead in Australia, BuildPass becomes a likely integration source, partner, or fast follower.

UK — second market, not default first

The UK remains attractive. ONS reports £140.7bn of Great Britain construction new work in 2024, £71.7bn of new orders, and 370,770 VAT/PAYE registered construction firms.⁶ King’s College London reports UK adjudication referrals hit 2,264 in the most recent surveyed year, the highest on record, with inadequate contract administration and lack of competence as leading dispute causes.⁷

The UK has the best language fit for a QS/commercial product: NEC4 compensation events, JCT instructions, EOT, loss and expense, final account, CVR, measured mile, prelims/prolongation. But the corpus now shows the UK is not empty: Gather Insights is the real UK neighbour, selling to the exact QS/commercial buyer, detecting NEC/JCT compensation events and drafting clause-referenced notices. Gather stops before quantum and recovery, but it has the front half of the UK motion.

Read: UK is an excellent expansion market once the event-to-recovery workflow is proven. It is not automatically better than AU as a first market.

US — biggest prize, worst first market

The US is the obvious TAM answer and the wrong starting answer. Census reports US construction spending at a seasonally adjusted annual rate of $2.17tn in April 2026, including $729.8bn private nonresidential and $532.7bn public construction.⁸ Census County Business Patterns reported roughly 780k construction employer establishments in 2021.⁹ AAA’s construction dispute page shows the scale of dispute culture, including a $514m largest construction claim in 2025 and 1,170+ construction panelists.¹⁰ Arcadis’ 2025 North America disputes data reports average dispute value of $60.1m and average duration of 12.5 months.¹¹

That supports high ACV. A US contractor might plausibly pay $50k–$150k/year if you reduce claims-consultant spend or improve change-order recovery. But the US is crowded on every input surface: Procore owns the SoR; Raken owns daily logs for smaller contractors; Trunk Tools owns AI document Q&A for enterprise GCs; Document Crunch/Trimble owns contract risk; SmartPM owns schedule analytics; Magra is the closest direct notice-to-recovery competitor.

Read: the US is the scale market, not the learning market. Enter after you have a proven recovery engine, outcome data and a wedge not reducible to “Magra but thinner”.

Canada — interesting, but not first

Canada has an attractive prompt-payment/adjudication direction, but it is province-by-province and less mature than AU/UK. ISED reports 415,406 Canadian construction establishments, 98.9% with 0–99 employees, and average SME revenue of C$539.6k.¹² Ontario’s ODACC annual report shows growth but still a smaller formal adjudication system: 324 adjudications commenced and 143 determinations rendered in FY2025; the report says adjudication is fast, efficient and cost-effective, with determinations normally due within 30 days of supporting documents and payment due within 10 days of a payment determination.¹³ Alberta’s prompt-payment rules establish statutory payment timelines and adjudication for construction payment disputes.¹⁴

Read: Canada is viable after the AU/UK pattern works, especially Ontario/Alberta. It is not the best first market because the local buyer density and cultural maturity are less obvious, and the market is smaller than the US without being as clean as AU.

Recommended beachhead sizing — Australia

This is bottom-up and intentionally conservative. The headline construction market is irrelevant; almost all firms are too small, residential, low-dispute, or commercially immature for a A$20k–A$80k product/service.

Macro filter:

Australian construction businesses, June 2024:        452,820
Less than 20 employees:                                  98%
Implied 20+ employee businesses:                       ~9,050

That 9,050 is the first plausible filter. It still includes residential, low-complexity, local trade businesses, plant-heavy firms and businesses with no serious commercial function.

Targetable beachhead filter:

20+ employee construction businesses:                  ~9,050
Relevant commercial/civil/fit-out/specialist slice: 35–45%
                                                      = 3,200–4,100
Firms with enough change/payment leakage + admin maturity: 40–60%
                                                      = 1,300–2,450
---------------------------------------------------------------
Core AU SAM for first beachhead:                       1,500–2,300 firms

ACV reality in Australia:

Practical beachhead math:

The honest AU first-market ceiling is probably A$15m–A$30m revenue before adjacent expansion. The theoretical SAM is larger, but it assumes penetration you have no right to assume. This is enough to build a strong specialist company and enough to seed a venture story if AU proof transfers into the UK, Canada and then the US. It is not enough to justify a pure venture SaaS raise on day one.

Expansion order

1. Australia
   Security of Payment, clear recovery route, less direct recovery-tool competition.

2. UK
   Mature adjudication + QS buyer + NEC/JCT commercial-event logic.
   Watch Gather. Enter with quantum/recovery, not notice detection.

3. Canada
   Ontario/Alberta first. Prompt-payment adjudication gives structural fit,
   but the market is provincial and less mature.

4. US
   Largest TAM and highest ACV. Enter only with proof, data and a Magra counter-position.

Market not to enter first: US. The US is where you go to scale once the motion works. Starting there burns capital against Magra, Procore, Document Crunch/Trimble, Trunk Tools, SmartPM and longer dispute cycles before you have proof that your event-to-recovery engine produces money.

Competitor table — dossier-grounded

Coverage shorthand: A1 tender/bid, A2 estimating/takeoff, A4 field/daily, A8 PM system of record, A9 schedule, A10 progress, A11 RFIs/submittals/docs, A13 reality capture, A14 cost/commercial, A15 change/claims/recovery, A19 risk/insurance, A21 historical cost. Scores are from the corpus heatmaps and notes, not reconstructed from vendor homepages.

The claims-delay note also found ClaimDD/ClaimEOT, Masin AI and Delay Claim Builder as genuine or near-genuine recovery micro-entrants. They are not in the user-requested table list, but strategically they matter: area 15 is not empty. It is occupied by small tools, with Magra and ClaimDD the cleanest quantum examples. That changes the thesis from “no one recovers money” to “no credible, funded, cross-firm-data recovery platform owns the market yet.”

What the table changes

1. The US is more contested than the original brief implied. Magra is not a curiosity; it is the closest functional analogue. Trunk Tools and Document Crunch/Trimble also make any US AI-over-docs wedge dangerous.
2. Australia is not empty either, but the threat is upstream capture, not recovery. BuildPass is strong in AU capture/admin AI. Payapps/Autodesk is strong in progress-claim workflow. Neither owns quantum/recovery.
3. The UK is not a blank QS beachhead. Gather owns the front half of NEC/JCT compensation-event notices. A UK entry must start where Gather stops: quantum, delay/disruption, recovery and outcome data.
4. Area 21 is still the cleanest moat, but only after trust. Rate QS, Gauge and BenchIt prove demand for historical-cost intelligence while proving why cross-firm pooling is not solved: vendors market privacy as the feature.

Vehicle recommendation

Recommended vehicle: productised recovery service first; software underneath

The evidence supports a consulting / managed-service / claims-services play that uses software internally, with a path to SaaS only after repeatability is proven.

Do not position as generic claims consultancy. That caps upside. But also do not pretend the first product can be self-serve SaaS. The buyer will not trust automated entitlement and quantum outputs in an adversarial payment/dispute process unless there is source-linked evidence, professional review and accountable judgement.

The first vehicle should look like this:

Commercial Recovery Audit / Event Memory Service

Inputs:
  contract + scope + programme + drawings + RFIs + emails + diaries + photos
  payment claims + variations + cost ledgers + subcontract records

Output:
  Commercial Event register
  entitlement / notice status
  quantum basis
  recovery probability
  payment-claim / adjudication / final-account pack

Commercial model:
  fixed diagnostic fee + annual account retainer + optional success-linked upside

Why not pure SaaS now

Pure SaaS dies early for three reasons:

1. The field-capture wedge is already commoditised. Raken, BuildPass, Fieldwire, PlanRadar, Procore and diaries/WhatsApp all capture events. The buyer will not pay serious ACV for another capture UI.
2. The highest-value output is judgement-heavy. Entitlement, causation, notice compliance, disruption and quantum need commercial/legal review. An unreviewed AI answer is not a product; it is a liability generator.
3. The moat depends on data permissions that buyers will not grant upfront. Cross-firm benchmarking requires trust, anonymisation, contract terms and proof that sharing data returns value. You earn that through outcome work, not a cold SaaS signup.

Why not a lifestyle point tool

A narrow event-log or notice-drafting app can become a decent £1m–£5m ARR business in one market. But the dossier table shows it will be surrounded immediately: Gather on notices, Magra on recovery, BuildPass/Raken on capture, Kreo/Civils.ai on tender reads, Rate QS/Gauge/BenchIt on single-firm cost history. The defensible company is not the UI. It is the commercial memory graph plus recovery outcomes plus trusted cross-firm cost/event benchmarks.

What this could become

The venture case only becomes credible if all four become true:

1. Recovery economics repeat across 30–50 projects.
2. Commercial Event schema survives messy real records without huge manual work.
3. Buyers sign data-use terms for anonymised outcome benchmarking.
4. Expansion into UK/Canada/US does not require rebuilding the product from zero.

Until then, call it what it is: a strong specialist services-first company with a possible SaaS/data moat later. That is a good business. It is not yet a venture SaaS company.

Commercial Event schema — keep this as the core object

The corpus is right to centre the Commercial Event. Keep it. Every competitor either captures upstream evidence without commercial consequence, or manages downstream cost/claims without a compounding event memory.

Commercial Event
├── Identity
│   ├── project / package / subcontract / location / WBS
│   ├── event type: instruction, variation, delay, disruption, access denial,
│   │   design change, acceleration, resequencing, scope gap, late information
│   └── source system IDs: diary, RFI, email, drawing revision, payment claim
│
├── Contract basis
│   ├── contract form: AS 4000 / AS 2124 / AS 4902 / bespoke, later NEC/JCT/AIA
│   ├── clauses: notice, variation, delay, EOT, L&E/prolongation, payment
│   ├── notice deadline and actual notice status
│   └── barred / preserved / arguable status
│
├── Evidence bundle
│   ├── contemporaneous site record: diary, photos, sign-ons, access records
│   ├── correspondence: email, WhatsApp/export, meeting minutes, RFIs
│   ├── design record: drawings, revisions, instructions, submittals
│   ├── programme record: baseline, updates, critical path, impacted activities
│   └── cost record: labour, plant, materials, prelims, subcontractor costs
│
├── Commercial judgement
│   ├── causation: employer/client/sub/design/weather/neutral/self
│   ├── entitlement theory
│   ├── time impact: EOT days, criticality, concurrency, mitigation
│   ├── quantum method: direct cost, prelim/prolongation, disruption,
│   │   measured mile, overhead, acceleration, loss of productivity
│   └── confidence / evidentiary gaps / reviewer sign-off
│
└── Recovery outcome
    ├── submitted value
    ├── certified / agreed / adjudicated / settled / rejected value
    ├── cycle time
    ├── counterparty objections
    └── benchmarkable features for future pricing and recovery probability

In Australia, the first implementation should be SoPA-aware: payment claim references, payment schedule response, adjudication application status, adjudicated amount, state regime, and response deadlines should be first-class fields. For the UK, add NEC4/JCT compensation event, final account, EOT and loss-and-expense fields. For Canada, add prompt-payment/adjudication province fields. For the US, add AIA change-order/claim notice, change directives, delay claim/litigation posture, Eichleay/Hudson-style overhead and state-specific constraints.

Risk register per bet — with kill criteria

Bet 1 — Field Evidence capture

What works: it is the natural demo wedge. People immediately understand “capture the thing on site before it becomes a fight”. It also creates the raw material for the Commercial Event object.

Where it dies: as a product category, capture is already crowded and low-WTP. BuildPass, Raken, Fieldwire, PlanRadar, Procore, OpenSpace, WhatsApp and email already capture enough. The issue is not absence of records; it is that records do not become money.

Verdict: do not build a standalone field app. Build thin event capture only where it improves recovery diagnostics, and ingest from existing tools wherever possible.

Bet 2 — Tender Intelligence

What works: it is clean to sell pre-award: identify contract risks, notice obligations, scope gaps, pricing exclusions and commercial assumptions before the job starts. It also creates the baseline against which later events can be measured.

Where it dies: Civils.ai and Kreo already own major pieces of tender/takeoff reading. Tender intelligence can become a clever preconstruction assistant that never creates the recovery data loop. If it stops at “nice risk summary,” it is not the company.

Verdict: Tender Intelligence is a useful software wedge, but not the first strategic proof. Build it as the baseline component of the recovery service, not as a standalone product.

Bet 3 — Claims / entitlement recovery — Area 15

What works: this is where money is. Magra, ClaimDD/ClaimEOT, Masin AI and Delay Claim Builder prove recovery tooling exists and can be productised. SoPA/adjudication makes Australia especially testable because substantiation can lead to near-term payment outcomes.

Where it dies: trust, liability and professional judgement. If your outputs are not credible enough for a commercial director, adjudicator, claims consultant or construction lawyer to rely on, you are a drafting toy. If every claim requires bespoke consulting, you are a consultancy with software props.

Verdict: this is the first serious money wedge. It must be service-wrapped until trust and repeatability are proven.

Bet 4 — Cross-firm historical-cost benchmarking — Area 21

What works: this is the best long-term moat. The corpus shows Gauge, Rate QS and BenchIt validate demand for historical-cost intelligence, while all three stay single-firm. That is not a modelling gap; it is a trust/legal/commercial-design gap.

Where it dies: data rights. Contractors may want benchmarks but refuse to contribute clean outcome data. They may also fear that anonymised data will reveal rates, margins, subcontractor pricing or negotiating positions. Rate QS marketing private data as a feature is not incidental; it is the objection in product form.

Verdict: do not build this first. Design for it from day one. Sell it only after private benchmarks and recovery outcomes create earned trust.

Sequenced roadmap with staged go/no-go gates

The correct order is not “capture app → tender tool → claims tool → benchmark marketplace.” That sequence optimises for demos and loses the money.

The better order is:

Stage 0: Validate AU beachhead and legal route
Stage 1: Productised Recovery Diagnostic
Stage 2: Tender Baseline / Commercial Memory setup
Stage 3: Live Commercial Event capture via integrations + thin UI
Stage 4: Repeatable Recovery Workbench
Stage 5: Private benchmarks → permissioned cross-firm benchmarks

Stage 0 — 30-day beachhead validation

Goal: confirm Australia is genuinely the best beachhead, not just analytically attractive.

Work: interview commercial managers, QSs, contract administrators, adjudicators, claims consultants and construction lawyers in NSW, Victoria and Queensland. Get redacted project records. Test whether Commercial Event reconstruction produces credible payment/adjudication packs.

Pass gate:

- 20+ senior buyer/expert interviews completed.
- 5+ firms provide redacted project packs.
- 3+ firms agree to paid diagnostic terms in principle.
- AU counsel confirms compliant service boundaries and data-rights structure.

Fail gate: no paid diagnostic intent, no access to records, or legal advice says the service cannot operate without becoming a regulated legal/adjudication practice.

Stage 1 — Productised Recovery Diagnostic

Build first: a service-led diagnostic that ingests messy historical/live project data and outputs a Commercial Event register plus recovery pack.

Why first: it tests the hardest thing immediately: can you turn records into recoverable money? It also creates the labelled event/outcome data needed for product later.

Offer:

A$5k–A$15k per project diagnostic
or A$25k–A$50k annual retainer for multiple project reviews
optional success-linked upside only after legal review

Output: event register, evidence gaps, notice/payment status, quantum hypothesis, recovery probability, payment-claim/adjudication/final-account pack.

Pass gate:

- 5 paid diagnostics sold.
- At least 3 diagnostic outputs used in a live recovery/payment/final-account process.
- Average delivery time below 25 human hours by pilot 5.
- At least one credible recovery, accelerated payment or preserved entitlement.

Fail gate: buyers like the idea but will not pay; every project is bespoke archaeology; outputs are rejected as not commercially credible.

Stage 2 — Tender Baseline / Commercial Memory setup

Build second: a pre-award/post-award baseline that extracts contract obligations, scope gaps, rates, assumptions, programme constraints, notice deadlines and risk allowances.

Why second: Tender Intelligence becomes valuable when it seeds the live Commercial Event ledger. Built standalone, it competes with Civils.ai, Kreo and generic AI. Built as baseline memory, it becomes proprietary context for recovery.

Pass gate:

- 5 firms pay A$10k+ per project or A$25k+ annualised for baseline setup.
- 70%+ of live events tie back to baseline assumptions, rates, clauses or scope positions.
- Commercial teams use baseline in CVR/final-account meetings, not just precon.

Fail gate: buyers classify it as precon productivity and refuse to connect it to live recovery.

Stage 3 — Live Commercial Event capture via integrations and thin UI

Build third: event capture that reads from existing systems: BuildPass, Procore, Raken, PlanRadar, Fieldwire, email, WhatsApp exports, Sablono/OpenSpace where present, spreadsheets and cost systems. Add thin mobile/web capture only for missing commercial signals.

Why third: capture is necessary but not sellable alone. By Stage 3, you know which event fields actually mattered in recovery and can avoid building a generic diary product.

Pass gate:

- 60%+ of material events are captured within 48 hours on active pilot projects.
- 75%+ of events have at least one contemporaneous evidence item.
- Commercial lead reviews event register weekly without founder prompting.
- At least 2 recovery packs use live-captured evidence rather than reconstructed history.

Fail gate: field teams refuse workflow, evidence is too weak, or commercial teams keep treating it as admin rather than money.

Stage 4 — Recovery Workbench

Build fourth: software that makes the service repeatable: entitlement clause mapping, notice deadline radar, quantum calculators, evidence-gap scoring, measured-mile support, prolongation/prelims, disruption, overhead, programme-impact summaries, reviewer sign-off and pack generation.

Why fourth: by now you know which manual steps repeat. Automate those. Do not automate the imaginary workflow you wished existed.

Pass gate:

- 30+ paid recovery matters processed.
- 50%+ of event-to-pack workflow automated or template-driven.
- Gross margin on diagnostics/recovery support exceeds 55%.
- At least 10 referenceable outcomes: recovered cash, accelerated payment,
  preserved entitlement, reduced write-off, or improved negotiated settlement.

Fail gate: gross margin stays consultancy-like, outputs require senior expert rewriting, or competitors beat you on credible recovery case studies.

Stage 5 — Private benchmarks, then cross-firm benchmarks

Build last: first private single-firm benchmarks, then opt-in cross-firm pooled benchmarks with strict governance.

Why last: area 21 is the moat, but premature benchmarking is bullshit. You need enough normalised events and outcomes before a benchmark means anything.

Pass gate for private benchmarks:

- 50+ projects normalised.
- 500+ Commercial Events classified.
- Buyers use private benchmarks in tender pricing, CVR reviews or recovery probability.

Pass gate for cross-firm benchmarks:

- 50+ firms or a narrower consortium agree to anonymised pooling.
- Competition/privacy counsel approves governance.
- No benchmark exposes identifiable rates, margins, counterparties or projects.
- Buyers pay separately for cross-firm insight.

Fail gate: data permission below 30%, counsel blocks pooling, or customers only want private single-firm benchmarking. In that case, remain a services/software company; do not pretend the data moat exists.

The 3–5 things to validate in the next 30 days before committing capital

1. Australia buyer willingness-to-pay against real recovery. Run 20 interviews and ask for money, not opinions. Target commercial managers, contract administrators, QSs, specialist subcontractor owners, claims consultants and adjudicators. The test is whether they will pay A$5k–A$15k for a recovery diagnostic on a real project.

2. Access to messy project records. Secure at least 5 redacted project packs with contract, programme, RFIs, diaries, emails, payment claims, variations and cost records. If firms will not share records under NDA for a paid diagnostic, the product cannot start.

3. Legal/professional boundary in AU states. Get construction counsel in NSW/VIC/QLD to review the service model, success-fee structure, adjudication support boundary, data-use rights and competition/privacy design. Do this before building anything material.

4. BuildPass/Payapps/Procore data reality. Validate whether target firms already use BuildPass, Payapps, Procore, Aconex, Excel/email or nothing. The integration roadmap should be based on actual AU mid-market stacks, not the global competitor map.

5. Recovery outcome proof from three historical matters. Back-test the Commercial Event schema on three completed/lost/settled disputes. The standard is not “nice summary”; it is whether a commercial director or claims consultant says the pack would have improved recovery, speed or negotiation posture.

Where your research is wrong or over-confident

1. The original UK assumption was too strong. The contract regime is the product. NEC/JCT is not a minor localisation layer; neither are AIA, SoPA or Canadian prompt payment. A single-market first strategy is not timid. It is the only sensible way to build a dense outcome-data moat.

2. Area 15 is more contested than the synthesis language still sometimes implies. The claims-delay note is clear: Magra, ClaimDD/ClaimEOT, Masin AI and Delay Claim Builder cross or approach genuine recovery. The opportunity is not “no one quantifies claims.” The opportunity is no scaled, trusted, jurisdiction-specific recovery platform with cross-firm outcome data.

3. Area 21 being empty is not enough. It may be empty because buyers do not trust it, not because nobody thought of it. Rate QS marketing private-data-only as a feature is a warning label. Cross-firm benchmarking is the moat only if you solve governance and buyer psychology.

4. Tender Intelligence is over-loved as a first paid wedge. It is sellable, but it risks becoming a precon productivity tool. Civils.ai already reads tender packages and obligations. Kreo already handles AI takeoff. Tender only matters strategically if it creates the baseline for later Commercial Events and recovery.

5. The dossiers’ review intelligence is valuable but biased toward non-commercial users. Raken, Fieldwire, PlanRadar and BuildPass reviews mostly reflect field/admin/safety/QA pain. They prove adoption and workflow gaps, but they do not directly prove QS/commercial willingness-to-pay for recovery.

6. Platform absorption risk is not theoretical. Pype→Autodesk, Payapps→Autodesk, Document Crunch→Trimble, Disperse→OpenSpace, and the Trunk Tools/Procore API conflict all say the same thing: a narrow AI layer on someone else’s workflow gets absorbed, blocked or copied. The only way around that is to own the recovery P&L line and the outcome data.

7. Australia-first raises the BuildPass risk materially. The previous UK memo treated BuildPass as an adjacent AU example. If AU is first, BuildPass becomes a local strategic factor. Do not build a field-capture app in its path.

Source notes

Internal corpus used

• dossiers/_CROSS-COMPETITOR.md
• dossiers/_LANDSCAPE-SYNTHESIS.md
• dossiers/_MARKET-PROBLEM-MAP.md
• dossiers/_OPPORTUNITY-LENS.md
• dossiers/raken/dossier.md
• dossiers/procore/dossier.md
• dossiers/fieldwire/dossier.md
• dossiers/planradar/dossier.md
• dossiers/document-crunch/dossier.md
• dossiers/openspace/dossier.md
• dossiers/trunk-tools/dossier.md
• dossiers/civils-ai/dossier.md
• dossiers/kreo/dossier.md
• dossiers/pype/dossier.md
• dossiers/sablono/dossier.md
• dossiers/buildpass/dossier.md
• dossiers/_landscape/claims-delay/note.md
• dossiers/_landscape/commercial-qs/note.md
• 03-research-construction-sop/output/construction-sop.md

External public sources