VERA · For Benoît Toussaint · October 2026

Blinded CapitalThe Multi-Trillion-Dollar Trust Gap

Arrow keys, swipe, or the buttons below

The problem, up close

Capital decisions depend on claims about real-world conditions.

However, the evidence supporting those claims is fragmented, unstandardized, difficult to rely on across institutions, not machine-readable, and highly exposed to the principal-agent problem.

As capital becomes increasingly automated in the agentic AI era, this creates critical gaps between physical reality and financial decisions.

Where it's widest

In construction.

Capital goes out in stages against an asset that does not exist yet, on progress evidence the lender cannot read and the next lender cannot reuse. Here, the gap is not an accident or a fraud. It is how every project gets paid for.

Released, in tranches Built Funded, not built

Where it hurts

Six symptoms, one cause.

Tap a card for the detail. Pick a party to see each card from their side.

See it as

01Manual Verification Rp 1.5–3.5m appraisal fee per unit in Indonesia, then re-linked by hand

Verification is paid for twice.

You pay the appraiser, then pay a reviewer to rebuild what comes back.

Your tranche waits in the review queue, and you bridge the wait.

You choose what to check and write it up your own way, then find the missed measurement next week, when going back costs a day.

Once in the appraiser's fee, Rp 1.5–3.5 million per unit in Indonesia, and again in the hours a reviewer spends re-linking drawings, specification, quantities and site report by hand. Each professional writes in their own format, so the next reviewer starts from scratch. Disbursement is slow because checking fragmented documents is manual work, not because anyone is slow.

The cost never shows as a line in the accounts, which is why nobody fixes it.

Sources: KJPP appraisal fee range · field interviews on the manual re-linking

02Liquidity Chokeholds 100 days a subcontractor claim can wait, in Singapore

Dead capital and the 100-day cash trap

A late tranche at the top reaches you as unmet payroll, then a stopped site.

You fund the gap between paying subcontractors and drawing, at your cost of capital, at every milestone.

Your book holds credit released against progress nobody can read quickly.

Cash waits on paper. In Singapore, subcontractor claims sometimes wait 100 days to be paid; in Indonesia, draws wait on manual review. Every one of those days is working capital funded by whoever is furthest from the bank, at the worst price.

Slower building, dearer capital, and a stock of credit disbursed against progress nobody can read: IDR 571tn in Indonesia, US$456bn on US bank balance sheets.

Sources: OCBC, Visa and Doxa press release, 2 Sept 2026 · Bank Indonesia SSKI · FDIC Quarterly Banking Profile, Q4 2025

03Unbacked Risks 73% of Indonesian developer funding is their own money

The unverified word

You fund ahead of what is built and find the gap at default, with nothing to attach.

You underwrite a risk you cannot observe.

You fund about 73% of the build from your own money, and carry it while the buyer holds theirs.

You self-fund to the next payment and can lose it on a technicality: withheld retention, an informal variation, a called guarantee you then reimburse.

You carry personal liability and indemnity on every finding you sign, for years.

You hold a claim, not a title. If the project stops, you recover what you can prove.

Progress is claimed in a report written by the party being paid, and everyone carries risk on it. The lender and the surety fund ahead of what is built and find the gap at default. The developer fronts most of the build from its own capital, 73% in Indonesia, and waits on the tranche. The contractor self-funds to the next payment and can lose it on a technicality. The professional who signs carries the liability for years. Construction firms are 27% of all Australian company insolvencies.

Everyone carries risk on a word nobody can check, and an honest claim is as hard to prove as a false one.

Sources: ASIC insolvency data, 2023–24 · Bank Indonesia residential property survey, Q2 2026 · contractor practices from Indonesian legal commentary

04Misaligned Agents 5 parties one fact, no version they all accept

Every party protects itself, and the weakest pays.

You front-load an invoice to fund payroll, and every claim you make becomes suspect, the honest ones too.

You hold retention because releasing it costs cash you don't have, and it reads as bad faith.

You hold the tranche until you're sure, and the site slows while you wait.

Paid by one side, doubted by the other, whatever you find.

You withhold what you can't check. When it fails, you have the least leverage.

You set credit-growth targets, then hold banks to account for the book.

Each party's move is rational on its own. The builder front-loads an invoice to fund payroll. The lender holds the tranche until it is sure. The verifier, paid by one side, is doubted by the other. The investor or buyer withholds what it can't check. The regulator sets credit-growth targets for banks, then holds them to account for the book. With no fact they all accept, every defensive move looks like bad faith, and the argument is settled by whoever holds the money or can afford to keep arguing.

The loss lands on whoever has the least leverage, usually the subcontractor or the buyer, not on whoever was wrong.

Sources: front-loading and pay-when-paid from Indonesian legal commentary · RPJMN credit targets, Perpres 12/2025

05Dispute Drag 12.5 months average dispute in North America, Arcadis 2025

Is it finished? Nobody wrote down what finished means.

Your variation fails on process, not merit, and your retention waits on an aesthetic objection.

Completion arguments cost weeks, and the programme absorbs them.

With photographs and a group chat, you are a creditor with an argument.

Guarantees are called or frozen mid-dispute, with no quick way to test the claim.

You're the one both sides blame.

Most payment arguments are one argument: was the work done, and what did "done" mean? The variation agreed on a phone call, the retention held on an aesthetic pretext, the handover left unsigned. Tribunals read paperwork, not intent, and arbitration is paid upfront by parties who are often already unpaid.

The build stops while the money argues.

Source: Arcadis, Construction Disputes Report 2025: average North American dispute US$60.1m, about 12.5 months to resolve, with failure to understand or comply with contract obligations among the leading causes

06Accountability Void 38% vs 18% of regulators, and of industry, who put AI harm on the institution

An AI model cannot carry liability

You answer for a decision your agent made on a document it misread.

You hold the institution to account; industry's most common answer is case by case.

Your tranche goes out early, or is held without cause.

You pay interest on progress that doesn't exist.

You have to prove what your report actually said.

Agents are arriving in financial firms: 52% of those surveyed are piloting or running agentic AI. When one misreads a milestone in an unstructured PDF, the error doesn't stay in the bank. The developer's tranche goes out early or is held without cause. The buyer pays interest on progress that doesn't exist. The professional has to prove what their report actually said. The bank answers the regulator and may go after the vendor. And nobody has agreed who should carry it: 38% of regulators put primary responsibility on the institution, and only 18% of industry agree.

A model can't be held to account, so everyone near it is.

Source: Cambridge Centre for Alternative Finance, 2026 Global AI in Financial Services Report, p.8 (52%), p.10 and Fig. 6.8 (38% vs 18%), p.95 (recourse to vendors)

Infrastructure

The infrastructure problem

While automation promises market efficiency, it compounds systemic risk by executing decisions on unverified, non-standardized inputs at an unprecedented velocity, creating systemic failures in what risk theorists call "tightly coupled" systems.

Without a reliable source of truth for physical reality, the agentic financial stack is forced to buy trust on a per-transaction basis from intermediaries whose interests it cannot see, leaving capital completely blind to structural changes in the material world.

Source: Basel Committee, Principles for the sound management of third-party risk, Dec 2025

A rule of evidence

Dig your well before you're thirsty.

Evidence of physical reality can only be captured while the reality is there. By the time anyone needs it, at the default, the dispute or the regulator's question, the pour is covered, the site has moved on and the people who saw it have left.

The answer

Connecting physical reality to capital deployment is now a critical requirement wherever money moves against a claimed state of the world.

What VERA isVERA provides the infrastructure, standards and protocol through which construction progress is independently evidenced and assessed by qualified parties — connecting physical reality to capital deployment.

Four parts, each closing a symptom

The Payment Condition Standard

What has to be true at each payment point, and what counts as proof. Published free.

Closes05Dispute Drag

05Dispute Drag"Finished" is written down before the work starts, as criteria, each with its own evidence requirement.

The protocol

Turns the contract, the bill of quantities and the drawings into conditions a qualified professional can test, and captures the evidence at the event, on the professional's own device.

Closes01Manual Verification

01Manual VerificationThe link between drawing, specification, quantities and site is built once, at capture, instead of rebuilt by hand by every reviewer.

The Record

The professional's finding in a structure a credit officer can read and a machine can process: condition, evidence, basis, finding, a named licensed author and their declared interest. A confidential, tamper-evident Record.

Closes03Unbacked Risks06Accountability Void

03Unbacked RisksThe finding is made by a qualified professional outside the claimant's system, so an honest claim can finally be proven.

06Accountability VoidEvery finding carries a named, licensed human. The bank's agent reads it; the officer decides; someone is accountable.

The registry

Where every Record resolves, so every party relying on it gets the same answer, current and checkable.

Closes04Misaligned Agents02Liquidity Chokeholds

04Misaligned AgentsFive parties read one Record instead of commissioning five versions of the same site.

02Liquidity ChokeholdsA Record the next reader can resolve is evidence a lender, a surety or a financier can act on without starting again.

Tap a part to see how it closes each symptom.

What all six shared was a missing source of truth. A Record is evidenced once and relied on many times, instead of trust being bought per transaction.

Where it sits

Everyone keeps their role.

The professional makes the finding and carries it. The bank, surety or escrow agent decides. The money moves between the parties exactly as it does now. VERA never inspects a site, never makes a finding and never touches the money: it supplies what all of them work to, so a developer, an engineer and a credit officer finally read the same thing.

See it as

Contract

the parties, as today

Payment conditions

VERA: the Standard

Evidence rules

VERA: what counts as proof

Finding

a licensed professional, never VERA

Record

VERA: the format

Registry

VERA: where it resolves

Decision

bank, surety or escrow agent, never VERA

Pick a party to see where it sits.

For the lender · DecisionYou decide, as today, on a Record you can read in one screen.

For the developer · ContractYour progress is evidenced once, and the next lender can rely on it too.

For the contractor · ContractYour work is evidenced when it's done, not reconstructed at the final account.

For the professional · FindingYou make the finding, in your scope, under your licence. VERA never does.

For the buyer · RecordYou can see the Record for your own unit.

For the surety · DecisionYou read the same Record the bank reads, instead of underwriting blind.

For the regulator · RegistryAggregate benchmarks only. No personal data in the registry.

The money

moves between the parties, as it does now. VERA is not on that line. For a payments person: a message standard, not a rail.

The professional's licence is the regulated object. VERA checks it against the state register; it never grants one.

What is being built

Infrastructure, not a product.

A substrate is stable, slow and deliberately boring; the products on top of it are many, fast and competitive. UCP 600, GS1 and ISO 20022 each shipped a standard, a way of using it and a place where things resolve, then let everyone else build the engines. VERA is three things, and only three.

Tap a part

what The Standard

The Payment Condition Standard: what has to be true at each payment point, what counts as proof, and the states a finding can take. Published free, versioned, held by a steering group rather than owned by a vendor.

Conditions, then criteria, then evidence classes, then a finding in one of six states: pass, fail, partial, not evidenced, not testable, not applicable. Who found it, on what basis, and who paid them are fields on the Record, not footnotes.

Construction reality is not binary. Keeping "we looked and it failed" apart from "we could not look" and "nobody supplied it" is what prevents the silent pass.

how The Protocol

How a Record is made and handled: derived from the contract, captured at the event on the professional's own device, signed there, resolved, superseded and never edited, disputed under an administered path, relied on under a named framework version.

The engine proposes the criteria from the contract, the bill of quantities and the drawings. The professional can override any criterion, and the override is part of the Record. The engine proposes; the professional decides.

Images are taken through the app, with location, device and time embedded at capture; gallery uploads are refused. A Record cannot carry a professional's name unless that professional's own device signed it.

why The Registry

The single place a Record resolves, so every party relying on it gets the same answer, current and checkable. The standard is what we publish; the registry is what we own.

Resolution answers four questions: is this Record what was signed and sealed, is it current or superseded, under which standard version, and which parties are named as relying on it. Corrections supersede; nothing is rewritten.

Build order, settled: the registry comes first. The schema engine can be run by hand for a year; a registry that resolves cannot be faked, and it is the piece an acquirer eventually buys.

Tap a part to see what it holds.

The substrate never inspects, never finds, never lends and never decides. It defines, connects and resolves. Everything else is an engine built on top of it, by us or by others.

The precedents it borrows from

UCP 600a rulebook nobody owns, adopted voluntarily, defining when money moves against conforming documentsVerra VCSstandard body, accredited operators, a registry and reliance, across jurisdictionsISO 20022a message standard, not a rail

What it looks like on day one

Nothing to install. Something to read.

On the first day, the same professional the bank already appoints submits the same finding, in a structure instead of a PDF. Two screens carry the whole change.

Draw Condition ViewerFacility 0417 · 24-unit landed housingcredit officer

Released

50%

Evidenced

41%

Payment events

Agreement trancheRecord #A-0417-01 · issued 12 Mar

Relied on

Foundation trancheRecord #A-0417-02 · issued 14 Sep · awaiting your decision

Partial

Roof tranchenot yet requested

Pending

F-04

Cube results not evidenced on six units. Laboratory documents requested from the developer. Nothing to decide yet.

Quiet

Block D: no capture activity for 31 days.

Facility→Record→Evidence→Registry→Outcomefive screens from tranche request to decision

What stays the same

The professional already appointed, and the fee already charged
The release process and the credit judgement
The core system; nothing to install, nothing to procure
The panel of professionals the bank already accepts

What changes

A structured Record arrives instead of a PDF: condition, evidence, basis, finding, on one screen
Partial is a finding state, so a partial disbursement becomes a priced decision instead of a blind one
Released against evidenced, per facility and across the book: the number an auditor asks for and almost nobody can produce quickly
Exceptions arrive without being asked for
The next lender can read it too

The officer never opens a drawing, phones the appraiser or compares two photographs. The officer still decides, and the bank's own system acts on that decision. A Record states findings; it never authorises anything.

Protocol RunnerFacility 0417 · Foundation

Foundation complete

six criteria · derived from the contract and drawing S-02

F-01Excavation depth and footing dimensionsPass
F-02Reinforcement diameter and spacing per S-02Partial
F-03Concrete coverNot testable
F-04Cube results linked to the pourNot evidenced
F-05Setting-out and levelsPass
F-06Starter bars and anchorsPass

Completeness: 14 of 16 evidence items captured. Two missing before you leave site.

beforeAn inspection brief from the contract, not a template. You know what to look at before you drive.
on siteA completeness check counts evidence against the requirement. No return visit for the one measurement you missed.
afterThe report assembles itself. Your findings, in your scope, under your licence, already in the structure the bank reads.
if challengedA defence pack: the full trail, basis, overrides and versions. Better indemnity, because the file shows what you did.
over a careerA portable profile: credentials, Records issued, findings upheld. It moves when you move.

Everything on this side is free. Only the credential is paid, and it falls due after the first engagement it won has been paid. The licence is checked against the state register and never granted. The test for any feature here is whether it removes work: anything that adds a step is abandoned in week three, however principled.

The agentic layer

An AI can read anything. It cannot vouch for anything.

Banks are putting agents on PDFs, emails and vendor portals. Those inputs were written by the party being paid, carry no type, and name nobody who can carry liability. When an agent misreads a milestone and the money goes out, the bank carries the loss and answers the regulator. Not the vendor, not the model.

52%

of surveyed financial firms are piloting agentic AI or beyond

81%

expect agentic AI to be meaningfully achieved by 2030

79%

of regulators rate explainability as critical or important

38% vs 18%

of regulators, and of industry, who put primary responsibility on the institution

Source: Cambridge Centre for Alternative Finance, 2026 Global AI in Financial Services Report, executive summary (pp. 6–10)

Credit assistantFacility 0417 · reading Record #A-0417-02agent · no release authority

Ask it something.

What does the Record say?Foundation tranche, Record #A-0417-02, issued 14 Sep. Finding author: a named, licensed structural engineer, scope "structure", appointed by the bank from the register, fee fixed before the visit. Six criteria: four pass. F-02 partial: Blocks A and B evidenced, Block C spacing corrected on site and awaiting re-inspection. F-04 not evidenced: cube results not received.

Can we release?Your policy allows release against the defined portion. On this Record that is 16 of 24 units. Block C and the six units without cube results fall outside it. I can prepare the partial disbursement for your approval. I cannot release anything.

Anything in the book?This facility is 9 points released ahead of evidence, second-highest in the book. Two other facilities have had no capture activity for more than 30 days. Exception list attached.

The agent reads. The officer decides. Nothing here releases money.

PaymentConditionRecord
├── contract_id, contract_version
├── payment_event_id, payer_approval_ref
├── conditions[] → criteria[] → evidence[]
│     └── finding {status, basis, author_ref}
├── authors[] → licence, scope, declared_interest
├── protocol_version, reliance_framework_version
├── lifecycle_state, supersedes_ref
└── timestamps, signatures, integrity_metadata

A typed object with a named, licensed human behind every finding, resolvable in a registry. Auditable, integrity-stamped, version-pinned, append-only.

That is the input an agent can be allowed to act on. A PDF is not. Every agent a bank deploys is one more reason to adopt the standard now rather than later.

Liability and the perimeter

You are not relying on us.

The strongest objection is not technical: a credit officer who releases on a startup's record has outsourced due diligence to an unvetted vendor, and internal audit will say so the day a loan goes bad. The answer is that the bank never relies on VERA.

The professionalcarries the finding, under licence and indemnity, exactly as today. If the foundation cracks in year three, the bank pursues the engineer's credentials and cover, as it would now.

VERAwarrants the Record: that it is what was signed and sealed, that it resolves and is current. Never that the finding is true. VERA is not in the liability chain and never seeks to be.

The institutioncarries the decision. The reliance framework says how far reliance extends, to whom, under which version, and how a dispute runs.

The canonical sentence in every institutional conversation"You are relying on the same licensed professional you rely on today, in a format your credit officer can read, and with an evidence trail you can use if it ever goes to court."

Disputes are administered, never adjudicated: VERA runs the procedure, an independent panel of senior accredited professionals decides, and VERA charges the administrative fee. The DOCDEX structure, copied because it is what keeps a rule-maker out of the liability chain.

The first question a risk committee asks about any inspection regime is who chose the inspector and who pays them. The Record answers it on its face: who commissioned, who paid, on what fee basis; whether each finding was witnessed or received; and the engagement is commissioned by the party releasing money, never by the party being assessed.

£2.5m

Eiger Funding v Ridge and Partners [2026] EWHC 609 (TCC), 16 March 2026. A lender's monitoring surveyor had also been the developer's quantity surveyor on the same project, without the informed consent RICS guidance requires. The court found negligence, held that it was reasonable for the lender to rely on the report, and awarded £2.5m. The refinancing lender had relied on sign-offs nobody independent had made.

Source: 4 Pump Court, case note, 19 March 2026 · Mayer Brown, May 2026

Supervisory frameworks expect collateral valuation to sit apart from the credit decision, and the Indonesian appraisal profession faces a live accusation: an appraiser appointed and paid by the bank produces the answer the bank wants. Declared interest, fee-basis disclosure and a structured evidence trail are the answer to it. VERA is not asking those professionals to adopt a standard; it is offering them a way to show an independence they are currently accused of lacking.

Banks do not ask for a standard. They ask not to be that lender.

The position, as it stands pending counsel. Stated once and plainly, because a regulatory profile reads the whole sentence, not the subject of the verb.

Not a construction serviceQualified professionals test the conditions and record their findings. The licence is the regulated object; VERA checks it against the state register and never grants one. The entity registers as technology, not as construction consultancy.
Not in the money flowNo custody, no float, no settlement chain, no payment licence. The money moves between the parties as it does now, so there is nothing here for a payments compliance team to assess.
Not AI, under the Indonesian AI governance guidelinepending counselA deterministic protocol with a human finding in it. Calling it AI would route approvals through the slowest committee in the bank.
Not a trust-service providerSigned at capture on the professional's own device, resolved in a registry. Auditable, integrity-stamped, version-pinned, append-only. Never described as beyond correction, because a professional can override a criterion and the override is kept.
Data stays where the law puts itRecord store, registry and every identifying field in-country; de-identified technical payload processed from Portugal; party names stripped before anything leaves. Designed in, not discovered in a vendor-risk questionnaire.
No sandbox, no single-bank dealA financial-innovation sandbox frames VERA as a financial service. A deal with one bank alone tells every other lender it is their competitor's supplier. Banks mandate the standard, never the supplier.

The commercial model

Everything is free except the Record a disbursement relies on.

Visa's genius was never its take rate. It created a revenue line for its own members, and banks promoted cards for decades because cards paid them. VERA copies that half. Indonesian banks already negotiate bulk appraisal rates, charge the customer a packaged fee and book the difference: the intermediary earning a spread on verification exists today. VERA makes that line bigger, cleaner and defensible, and takes a small network fee from it.

The professionalTheir full fee, paid by whoever engages them. VERA takes nothing from it, ever. The platform that takes a slice of the job is the one answer a surety's risk committee rejects.

The bankKeeps the margin on a verification charge it already levies. One aggregated monthly invoice from VERA, in arrears, for Records relied on. This is the reason to adopt.

VERAA fixed, tiered network fee per Record relied on in a disbursement decision, banded by facility size. Fires on disbursement, never on registration: no Record, no release, no fee. Never a percentage of construction value.

The borrowerOne charge, or none they are not already paying, disclosed as it is today. Economically the developer bears it, as they bear the appraisal fee now. In subsidised housing the borrower cannot be charged at all.

Six lines, on the ICC and UCP 600 model

#	Line	Who pays	Starts
1	The standard, published free	nobody	Day one. The adoption engine, not a revenue line.
2	Implementation layer: sector annexes, assessor handbook, worked examples	operators, banks, developers, law firms	Before acceptance
3	Accreditation and training, examined and renewable; the fee falls due after the first paid engagement	professionals and their firms	Before acceptance
4	Dispute administration, filing and administrative fees	disputing parties	After the first disputes
5	Records metered: institutional reliance. The line that scales.	banks, sureties, escrow agents	After acceptance
6	Corpus licensing: base rates per condition type, underwriting data	insurers, funds, portals	Year 4 and later

Two of the six earn before any institution has accepted anything, and selling them builds the operator network the pilot needs. Services are capped at €300K a year and 40% of build capacity: a healthy consultancy with no standard is the failure mode this guards against.

Working estimates, not forecasts

Rp 75–150K

per verified event, three to four events per unit, against Rp 1.5–3.5M a unit already pays for an independent appraisal

€2.1M · €8.4M

revenue, base case, years three and five; services under a hard cap throughout

€3M – €18M

year-five spread between the slow and fast cases, and almost all of it is one variable: how many banks sign a framework, and what share of each book flows through

Blueprint estimates derived from the market frame, labelled as such. The distribution is step-shaped: one lender writing the standard into its underwriting guidelines moves a share of an entire book at once.

Nevercustody, escrow or floatrelease on a scorerating professionalsdispatching worka share of a professional's feea percentage of the draw"authorises you to release"

What else the Record opens for a bank, as hypotheses

A verified receivable is a different credit from a self-reported invoice. Supply-chain and sub-tier finance against Records, currently unavailable in Indonesia outside top-tier counterparties. Speculative, not a year-one line.
The closing package: when a construction loan refinances, the incoming lender's verified history of the asset assembles itself from Records already issued, at a moment when professional fees are already being paid.
Sureties and insurers reading the same Record the bank reads, pricing exposure they currently underwrite on paper.
None of these are modelled. They are the reason the Record's value grows once other people's money starts depending on it.

How it grows

Engines on the infrastructure.

The substrate stays neutral. Seven engines grow it, each feeding the others, and every Record relied on pays the registry a small fee that funds the next engine. Not every engine has to work. The first three start the loop; the registry does the rest.

we build itpartners build it on the standard

Engine 01 · supply

Turns a site visit into a signed Record.

who: engineers, quantity surveyors, appraisers, monitoring surveyors
why they use it: shorter day on site, no report to write, better indemnity, a credential that appreciates
produces: Records, the supply side
effect: density

We build this. If professionals do not produce Records, nothing else here has a subject.

Engine 02 · demand

Reads Records and routes decisions.

who: credit officers, technical advisers, risk teams
why they use it: released against evidenced per facility, exceptions without asking, a margin on verification they keep
produces: reliance events, each paying the registry
effect: demand

We build this. Alerts and recommendations only; it never holds or releases a payment.

Engine 03 · pressure from below

Evidenced work is paid against sooner.

who: subcontractors and tier-two suppliers
why they use it: a claim that can be checked is a claim that is paid in days, not weeks; cash is oxygen
produces: demand from the bottom of the chain
effect: subcontractors push contractors, contractors push professionals, professionals push the app

We build this. The party most helped by evidenced completion is the one furthest from the decision to adopt it, which is an argument for the free tier, not against the segment.

Engine 04 · capital depth

Underwrites against evidence instead of paper.

who: sureties, insurers, escrow agents
why they use it: a surety's classic loss is owner overpayment, not fraud; a disciplined obligee is a better risk
produces: capital confidence; the biggest balance sheets enter
effect: depth

Partners build this on the standard. Sureties carry the strongest strategic signal: they open the bond product.

Engine 05 · legitimacy

Aggregate benchmarks on the physical side of credit.

who: supervisors, auditors, government agencies
why they use it: today nobody can see the physical side of construction finance; the registry gives an aggregate view with no personal data in it
produces: legitimacy
effect: once a supervisor leans on the standard, banks stop hesitating

A relationship, not a product. Later, and never as a sandbox.

Engine 06 · volume

Verified construction receivables as an asset class.

who: pension funds, insurers, private credit
why they use it: pooled receivables backed by independent, dated findings against contractual conditions
produces: the deepest pool
effect: construction risk becomes tradable; the top of the funnel opens

Partners build this. A hypothesis for year three or later, named because it changes what the registry is eventually for.

Engine 07 · future-proofing

The only input an agent can be allowed to act on.

who: banks deploying agents on credit decisions
why they use it: an agent cannot carry liability; a Record carries a named, licensed author
produces: an accountable automated credit stack
effect: every agent decision is a reliance event; a five-year adoption curve compresses

An integration with customers we already have, built after human officers are using the monitor, not before.

Platforms carry Records

Procore, Autodesk, Sage, Doxa, Built. None of them attaches to a standard on its own initiative; each attaches when a bank asks it for the Record, or loses the customer. Distribution is a consequence of the first bank, not a partnership to pursue before it.

The buyer's unit

An off-plan buyer pays stages against milestones they cannot see. A Record before each instalment shows what was found, by whom and on what evidence, and the buyer's leverage runs through the bank's staged disbursement, never through pressure on institutions. Over the life of the asset the Records become a portable history every future owner, lender and insurer reads. A long-tail story; not phase one.

Every party wins somewhere

Each arrangement, argument and onboarding step is designed so that every involved party wins at some point in the process. That is the adoption strategy, not a slogan: a standard that only protects the party paying is a developer's tool, not infrastructure.

Phase 1 · months 0–12Supply loopApp and monitor, one institution. One real Record, one reliance event, one fee.

Phase 2Pressure from belowSubcontractors paid sooner become the advocates; more contractors and developers adopt.

Phase 3Capital loopSureties, insurers, then institutional capital reading the same Record.

Phase 4Agentic loopThe first bank whose agent reads only Records.

Phase 5LegitimacyThe standard becomes a reference, not a vendor.

The landscape

Nobody is doing it as a standard.

A great deal of money is digitising the paperwork. That is validation, not competition: the problem was never paper, it is that every professional, bank and platform records the same event in a different structure, so nothing travels. The asset is the Record; the work is standardisation, attribution and portability.

Show

Textura · Procore Paypay applications, lien waivers, invoices

Payment paperwork inside one vendor's system. Machine-readable within the walls; the claim is still the claimant's, and it resolves nowhere a second institution can reach.

Pillar · Siteline · Constrafor · Sinqsite updates reconciled to the accounts

The contractor's own paperwork, tidied and financed. Pillar raised a €12M seed in May 2026 on exactly this thesis.

Source: tech.eu, 12 May 2026

DoxaSingapore · OCBC, Visa

Developer-to-bank drawdown workflow since 2021; deep-tier financing on approved claims from September 2026. Roles, e-signatures and an audit trail inside the walls. The finding is the developer's consultant's, and it resolves inside one product. Evidence the market is real; a future reader of Records, not a rival.

Source: OCBC, Visa and Doxa press release, 2 Sept 2026

Built · Sitewire · RabbetUnited States · lender-side draw management

Built raised USD 125M at a USD 1.5bn valuation in 2021 and now offers a draw agent. It automates the decision side and leaves the evidence side untouched: complementary. Sitewire verifies progress for one lender inside one product.

Source: Built Technologies, Business Wire, 30 Sept 2021 · draw agent as reported

Automated release5D BIM and smart contracts

A decade of research binds the model to the payment and lets the data release funds. Every attempt answers the same gap with a sensor: drones, scanning, IoT, digital twins. A sensor can prove a pour reached temperature; it cannot carry liability, so no bank can rely on it and no court can weigh it. The literature calls this the oracle problem. Lack of standardisation is named as an adoption barrier in its own right.

Source: Sonmez, Ahmadisheykhsarmast and Güngör, Automation in Construction 139 (2022), and the surrounding literature

The handsappraisers, MK firms, quantity surveyors, engineers, inspection bodies

The people who make every finding and carry every liability. They are the network, not a cost to remove. In Indonesia two licensed professions already do this work nationally and are already accepted by banks. The state habitability gate for subsidised housing is complementary: a terminal check against a government checklist, not conditions against a contract during the build.

The insight underneath

The oracle problem is an accountability problem.

A construction progress payment is a transition from brute facts, what exists on site, to institutional facts, what the parties agree is owed. Every system above needs an attested, structured, liability-bearing input at that transition, and none of them can produce one. VERA sits at the transition and never on either side of it.

A platform never attaches to a standard on its own initiative. It attaches when a bank asks for the Record. The first yes is a bank's; the registry is the asset; the Record is the unit.

The same shape elsewhere

Construction is Profile 001.

The mark is not construction-locked, and neither is the schema. Five tests say whether a market has the shape; wherever all five hold, the same Record travels with a different annex.

test 1The claimant is the payee.

test 2A capital provider bears the loss.

test 3The condition is an observable state, not paperwork.

test 4A credentialed tester exists.

test 5Several institutions rely on one finding.

Construction

Stage payments against a building that does not exist yet; licensed engineers and surveyors; a lender, a surety and a buyer relying on one finding.

12345

Shipbuilding and heavy equipment

Stage payments gated by class surveys; refund guarantees as the surety. Five of five.

12345

Project finance

Independent Engineer drawdown sign-offs and completion tests, under FIDIC at scale.

12345

Results-based and grant finance

Impact bonds, programme-for-results, public programmes. Independent verification agents exist; every programme writes its own protocol.

12345

Trade-finance inspection

UCP 600 standardises how documents are examined; what an inspection report must contain is unspecified. One layer under UCP.

12345

Sustainability verification

Green, taxonomy and carbon findings by national inspection bodies, paid for once and reused free. In Europe, a criterion annex rather than a second country.

12345

Design consequence: the core schema is domain-neutral, a subject, an anchor, a condition and a finding, with construction in an annex. Verra, XBRL and ISO 20022 each shipped one domain first on a horizontal architecture. A multi-author Record is what makes it travel: film completion bonds and shipyards have the same milestone, several competences and a guarantor. First, one Record, one institution, one market.

Where

Indonesia first. Multi-geography by build requirement.

A market qualifies when capital is released in stages against construction progress and no standardised, machine-readable evidence connects the two. Indonesia qualifies three times over.

The milestone taxonomy is already national

Bank Indonesia's indent-property rule defined the tranches: up to 30% at the agreement, 50% at foundation, 90% at roof close. Every property banker recognises them whether or not the rule is currently enforced. The regulation named who verifies; it never defined what counts as proof. That gap is the product. Say standardise, never digitise, and never claim "already required".

The operator network already exists

Licensed appraisers registered with the financial regulator, and construction-management firms registered with the national construction board, already perform geotagged, app-based stage inspections and are already accepted by banks. The accredited supply exists on day one; the state tests their competence.

Three decisions, not two hundred

Mortgage credit sits with a handful of institutions, and penetration is low, so every lender has headroom and a reason to grow the book it has most reason to be careful with. A bank that wants to lend more against unbuilt property needs better evidence to do it safely.

USD 11–27bn

of annual commercial, non-subsidised construction value, on roughly 220,000 units a year at Rp 800M–2B each. An estimate, not a sourced figure, and the least-sourced input in the model; it scales everything. Subsidised housing is already gated by a state system, and informal self-build has no contract, no counterparty and no payer.

AustraliaLenders fund development against independent quantity-surveyor progress sign-offs; builders paid for 70% against 40% built are a live cause of insolvency. Same architecture, mature profession, no incumbent state system.

United KingdomThe Building Safety Act's golden thread was tested and found not to link to payment: higher-risk buildings only, safety only. The UK stays on the list for forensic provenance, which is the honest version of the claim.

UAEwarm doorOff-plan escrow release already conditioned on verified milestones; inspection throughput is the bottleneck. The behaviour is compulsory and the constraint is capacity.

The second geographyis chosen for proof, not revenue, and weighted by who will take the meeting as heavily as by market size.

NigeriaPresale-funded, stalled developments endemic, reported moves toward mandatory escrow. The largest African opportunity and the highest execution risk.

South AfricaA contractor late-payment epidemic and a sophisticated banking system. The cleanest fit anywhere for the receivables angle.

VietnamPresale-funded with a post-bond-crisis trust collapse. Real demand, less capital while it recovers.

PhilippinesOverseas workers remitting against units they cannot see. Retail buyers generate demand but do not pay; it needs the same lender adoption.

Botswanawarm doorRuns on FIDIC contracts, already standardised. Contractors reportedly decline to enforce payment rights for fear of being blacklisted; a machine-readable finding depersonalises that claim.

RwandaNot a revenue case. One public adopter is worth more to a category than a hundred private ones.

United StatesMechanics liens and lien-waiver bottlenecks, with title insurers as gatekeepers. Different legal machinery: a partnership market, not an entry market.

European UnionGreen-lending and taxonomy rules create a verification blindspot with real penalties. The second use of the schema, not the second country.

Everything beyond Indonesia is a research agenda, not a set of claims: no geography claim enters a bank meeting until it is sourced. The build consequence is already settled, though: the schema, evidence standard, Record format and registry are jurisdiction-agnostic, Indonesian specifics live in an annex, and FIDIC compatibility is a design target. Credit-side discovery is running from Lisbon now.

The one objective

An institution states in writing that it will accept a Record.

First Acceptance. Everything is sequenced against that sentence: it is the moment the company stops being a proposal and becomes infrastructure, and it reprices everything after it. Any activity that does not shorten the path to it is either cash to fund the path, or a distraction. There is no third category.

The acceptance ladder

Rung 1 · a deal decision, made by one person

Run alongside. Do not replace.

Find a project that already has a relying institution, a lender already disbursing or a surety already on risk, and supply the Record free, in parallel to whatever they use now. Zero switching cost, zero process risk, no procurement. Then ask the only question that matters: was it better than what you had?

Order institutions by decision speed, not strategic value. Developers decide fastest and signal weakest; sureties next, with the strongest signal; escrow agents; then banks, slowest and largest downstream. Approach banks only with Records already in existence.

The pilot shape: a regional development bank, one localised development of about fifty landed units. The clause sits in the credit agreement, not the construction contract. The bank appoints the operator from the register, the developer pays, and the fee is fixed before the first inspection and cannot move with the finding.

Rung 2 · the habit

Interim proof, not completion.

Fifty units to handover is nine to twelve months; First Acceptance cannot wait that long. After the first foundation tranche: did the officer act on the Record without asking for anything else? After the first superstructure tranche: days from inspection to disbursement, against the bank's own baseline.

Those two numbers win the second bank. Without a baseline they prove nothing, so the baseline is asked for before anything is signed.

Rung 3 · the committee matter

"We accept Records generated under VERA Standard 1.0."

This is First Acceptance in its usable form: protocol version, authorised operators, scope, named relying parties, permitted reliance and its limits, dispute procedure, retention, amendment rules. The twelve-month target is one institution here. Revenue without it means the thesis is unproven, however healthy the services line looks.

Standard 1.0 is not frozen until a capital-side co-author has shaped it. A standard rewritten after publication loses the credibility it was published to earn.

Rung 4 · where the flywheel starts

The flywheel starts when a financier requires it.

A bank requiring a VERA-conformant Record from any accredited operator is specifying an objective baseline in its collateral guidelines, the same shape as requiring a licensed insurer. Requiring VERA the company would tie a credit facility to a third-party service, which bank legal departments refuse and which would close the standard. Carrot over stick: better terms for compliant projects beats exclusion of the rest.

NHBC did not win on buyer preference. It won when lenders required it.

The first ninety days, with the kill criteria on the page

Registry first. Three things already depend on a Record resolving: the project badge, the developer's proof message to a buyer, the sample link a lender clicks. The schema engine can be run by hand; the registry cannot be faked.
Three Payment Condition Maps by hand, from three real Indonesian contracts, no platform. Kill: if three real contracts cannot be reduced to testable criteria by hand, the engine is not buildable and the thesis fails here, cheaply.
Three questions. A quantity surveyor: can you execute this without us telling you how to do your licensed job? A bank or surety risk officer: if this Record came from a professional you already accept, could you use it in your release process? A construction lawyer: does this perform a regulated function, or supply a protocol and a Record to the parties who hold that function? Kill: any one fails, and the protocol is fixed before 1.0, before any AI, dashboard or field operation.
The run-alongside offer, placed with one project that already has a relying institution. The shortest path to rung one.
Measure the relay, not revenue: advocacy rate, developer acceptance, protocol adoption, institutional reliance, and the share of projects where three or more stakeholder types touch the same Record. Below that, VERA is a collection of free tools. North Star: payment events made institutionally legible.

Where it stands, October 2026

Architecture complete. Standard, protocol, Record schema at version 0.1 draft with a reference validator, registry design, reliance framework outline, the free tools on every side mapped.
Discovery running. Credit-side interviews from Lisbon now; Indonesian professional and bank conversations next. Until the first institutional conversation is had in Jakarta, every fact here comes from a document rather than a counterparty.
No Record issued, no institution signed. That is the honest position, and it is what the next ninety days exist to change.
Open: the regulatory position is pending counsel; governance is proposed as a steering group with operators, a surety and eventually a lender, and is not yet decided.
Two vehicles, one disclosed boundary. Company A owns the standard and is the exitable asset. A separate, later, private vehicle buys what the standard's sensing finds, deliberately below the size institutions care about, disclosed in A's terms from day one. It does nothing before month nine.

Headwinds, stated plainly

Two constraints, in sequence. Busy professionals must produce findings in the schema at all before any bank conversation has a subject; then the growth constraint is bank framework count.
Main contractors may resist quietly. Front-loading an invoice is a working-capital strategy, and precise evidence removes that lever. They will object on cost and programme. The engagement is therefore commissioned by the party releasing money, never the party assessed.
Capital flight to risk-free paper. If lenders would rather hold central-bank securities than fund housing, the addressable book shrinks regardless of how good the evidence is. Not in our control.
Mid-market un-bankability cuts both ways. Developers declined because execution risk cannot be assessed are the ones verified progress makes lendable again. Testable: ask a credit officer what would have to be true to say yes to a borrower they currently decline.

Where this goes

One Record. One institution. One market. Then the rest.

The sentence everything is sequenced againstAn institution, a bank, a surety or a developer, states in writing that it will accept a Payment Condition Record produced under this protocol.

This is the state of the thinking in October 2026, at exploration stage, with the open questions left on the page rather than tidied away. The rest is a conversation.

Malik · Lisbon