Europe's Last Toll Booth

Preface

On a Tuesday morning in 2026, in a notary office in Cascais, a property transaction will close. Five chairs at the desk. The notary. The lawyer. The buyer. The owner. The fifth, where the agent used to sit, will be empty. The transaction will close in forty minutes. The parties will pay around three thousand euros for the public infrastructure pieces required to consummate the transfer of property title. Six months earlier, the same transaction would have carried twenty-five to forty thousand euros in agent commission for coordination work that, by then, was already done by software.

This is not a forecast.

This is what is now possible. The infrastructure that compels this outcome is already operational. The position the empty chair represents is the last residual gatekeeper in the largest offline transaction most Europeans ever conduct in their lifetime, and its persistence is no longer a function of value created. It is a function of structural inertia.

Property mediation is the last category of European mass-market commerce yet to migrate to digital infrastructure. It is also the largest.

European real estate brokerage commissions drain between one hundred and seventy-five and two hundred and seventy-five billion euros annually from buyers and owners across the Union, by estimates that vary depending on methodology.

The first eight to ten years of my career, I assumed the position was indispensable. The years that followed taught me it was the bottleneck.

In Portugal and Spain alone, those commissions exceed seven billion euros a year, extracted across roughly seven hundred and fifty thousand annual transactions. The position that collects them no longer authenticates anything that public infrastructure does not authenticate faster, cheaper, and with greater traceability.

This memo is not a pitch. It is a thesis about a category that the next eighteen months will define. We are building property rails — one specific layer of the infrastructure this category requires. The architecture is described at the level of the system, not the company.

The argument that follows traces the same arc twice. First, in two financial markets where it has already played out. Then, in the one where it is now playing out. The pattern is consistent, the conditions are met, and the chair is empty.

Iberia is the proof. The pattern repeats wherever the conditions converge. The category is being defined. Right now.

I. The Pattern — Disintermediation as Recurring Cycle

This is not the first time a position has outlived its function. Twice in the last fifteen years, the same arc has played out in financial infrastructure, against incumbents who looked unassailable and ended up irrelevant.

The first case is Stripe. Until 2010, no merchant could accept a card payment online without an intermediary. Card networks required merchants to be onboarded through credentialled merchant account providers who ran underwriting, set rates, and absorbed the risk of chargebacks from fraud. The merchant rep existed because the alternative did not yet: PCI DSS was nascent, fraud scoring was rule-based and brittle, identity verification at scale was not yet a standard product. Onboarding required a person to spend hours reviewing each business and assuming risk on the network's behalf.

The underlying conditions then changed. PCI DSS matured into a deterministic compliance framework. Machine-learning fraud scoring outperformed any human underwriter. KYC went digital. The technical justification for the merchant rep evaporated by the late 2000s. The position itself did not. It persisted on inertia for three or four more years, charging fees for work the infrastructure beneath it now performed automatically.

Stripe replaced the position in 2010 by collapsing merchant account onboarding into a single API endpoint. By 2024 the company processed one point four trillion dollars in payment volume in a single year, equivalent to one point three percent of global GDP, on infrastructure that does not require a merchant account intermediary at any point. Stripe did not argue with merchant reps. It built the rails on which they were no longer needed.

The second case is Wise. Until 2011, no individual could move money across currencies without paying a brokerage spread to a bank or an FX broker. Banks did not have direct retail-grade cross-border matching infrastructure. International wires moved through correspondent banking networks that were slow, opaque, and priced on layers of intermediation. The FX broker existed because the gap between what the customer paid and the underlying interbank rate could not be closed any other way.

The underlying conditions then changed. Peer-to-peer matching algorithms made it possible to pair a customer in London sending pounds to Berlin with a customer in Berlin sending euros to London, settling each in their local currency at the mid-market rate. SEPA made domestic-rate transfers across the Eurozone instantaneous and cheap. Open Banking under PSD2 opened the underlying account infrastructure. The technical justification for the FX broker evaporated. The position persisted for years on inertia, charging spreads of two to four percent on retail volume the infrastructure beneath it could now move at near-zero marginal cost.

Wise replaced the position by routing customers through P2P matching directly. In fiscal year 2024 alone, the company moved one hundred and eighteen point five billion pounds across borders, saving customers one point eight billion pounds in fees that would have flowed to incumbent brokers and banks. Wise did not argue with FX brokers. It made them economically untenable.

The two cases share a structure that is not coincidental. In each, a position that had been legitimate for decades lost its technical justification incrementally, persisted on inertia for years after, and was supplanted by infrastructure that performed the same function natively. The successor companies did not win by being cheaper or faster than the incumbents. They won because they were performing a different operation. They were not intermediaries who had improved the intermediation. They were rails on which intermediation was no longer required.

Each time, the same arc. A legitimate position outlives the conditions that legitimised it. It persists by inertia. It is supplanted not by argument but by architecture.

I have not watched Stripe or Wise from the inside. I have watched the equivalent positions in real estate — the agent, the notary's clerk, the bank's loan officer — outlive their function in real time. The pattern is not theoretical for me. It is biographical.

The third case follows the same arc. The architecture is operational. The position is the last to know.

Real estate is the third case. The next section traces how it begins.

II. The Origin and Obsolescence of Real Estate Mediation

The story of real estate mediation begins in the gap between strangers. Two parties, both transacting at scale for the only time in their lives, with no shared frame, no shared registry, no shared anything. The position emerged because the gap was real.

In nineteenth-century European cities, formal estate agencies arose to bridge five recognisable problems at once. First, owners and buyers had no efficient way to find each other. Second, neither could verify the other's identity or capacity to contract. Third, neither had reliable access to the legal status of the property changing hands. Fourth, the documentation required to consummate a property transfer was idiosyncratic, locally drafted, and difficult to coordinate without physical presence. Fifth, there was no infrastructure for the secure transfer of large sums between strangers across distance. The agent solved all five at once, in person, and the position institutionalised because it worked.

The twentieth century scaled the model: Multiple Listing Service systems formalised inventory pooling, franchised networks like RE/MAX and Century 21 standardised brand and process, and by the time the internet arrived, agency had become the architecture of European real estate's last value chain — a default, not a service.

What followed was elimination, one technical layer at a time, of every original justification.

Information asymmetry was the first to evaporate. Idealista, Rightmove, and Fotocasa, all founded around the year two thousand, made the cost of bringing buyers and owners into the same view zero. The function the agent performed in nineteenth-century Lisboa, walking the inventory between two people, is now a search filter.

Identity verification was the second. The eIDAS Regulation, binding across twenty-seven member states since 2016 and extended by eIDAS 2.0 in 2024, made cross-jurisdictional digital identity legally equivalent to in-person presentation. Portugal's Chave Móvel Digital and Spain's DNIe perform in thirty seconds the verification that, before 2016, required an agent to stand in a notary's office and vouch.

Property authentication was the third. The Caderneta Predial digital in Portugal, the Nota Simple in Spain, the digitalisation of Land Registries across the EU. All of these turned the legal status of a property into an API call against an official source. The agent did not authenticate property in 1995 because they were trustworthy. They authenticated because there was no other source. There now is.

Document coordination was the fourth. Qualified electronic signatures under eIDAS QES carry the same juridical weight as a notarised wet signature in all twenty-seven EU jurisdictions. The Multicert and Adobe Sign infrastructures process them in seconds. The position of “person who walks the contract back and forth” no longer requires a person.

Payment infrastructure was the fifth. SEPA Instant, the SIBS rails that process over a billion transactions per year in Iberia alone, and segregated escrow APIs all make the secure movement of three hundred thousand euros between strangers a standard product, not a courtesy extended by a trusted intermediary.

Legal coordination was the sixth. Portugal's Lei 4/2020, Spain's Real Decreto-ley 6/2023, and the EU Digital Tools Directive 2019/1151 made remote notarisation legal. The notary still notarises. The agent does not.

Six original justifications. Six specific pieces of public infrastructure that supplanted each. The position that once unified them now coordinates none.

The arithmetic of what remains is unsentimental. Decompose a typical Cascais residential transaction conservatively, at a property value of five hundred thousand euros:

In plain language: the mark-up against the public infrastructure that already delivers the same outcomes starts at six hundred and fifty-three per cent at five per cent commission and reaches one thousand one hundred and five per cent at eight per cent commission.

The infrastructure already delivers the outcomes. It is operational, regulated, auditable, and priced. The commission persists not because it pays for any of these. It persists because the position has not yet been formally retired.

The mathematics of capture extends beyond this single transaction. Stripe captures around three percent of payment value. Wise captures less than one percent of transfer value. Real estate agents capture five to eight percent of property value. The infrastructure that delivers the same outcome can operate at zero point seven percent at scale, the same convergence point Stripe and Wise reached in their mature markets. The take-rate available to the position that captures the deterministic operation in any major asset class converges to under one percent over time. This has happened in payments. It has happened in remittance. It has happened in trading commissions and in mortgage origination. It has not yet happened in real estate mediation. It will. The category is the gap between what is captured today and what the infrastructure now permits.

The economics describe only one half of the structural problem. The other half is conflict of interest, encoded into the architecture of the position itself.

The agent represents the owner and serves the buyer. The agent is paid by one side and shows properties to the other. The agent decides which proposals are forwarded and which are filtered, which counterparties are introduced and which are withheld, which information is disclosed and which is held in reserve. In Portugal and in Spain, dual representation is not the exception. It is the default.

The same is true in the United States. In Spain and the United States, both sides of the transaction — owner and buyer — pay commission to the same agent, symmetrically, each side responsible for delivering three to five percent of the transacted property value. From each side. For the same single deal. In these cases of double “taxation”, the math is straightforward.

This is not a claim about individual agents. It is a structural fact about the architecture they inhabit. The position rewards opacity. It punishes transparency. It survives only because no party can verify what the position chooses to disclose. The agent who lunches with the seller on Tuesday cannot be the agent who serves the buyer on Wednesday. But this is the model. The structure conceals the contradiction from both. The only way the contradiction stays concealed is if no one builds the architecture that exposes it.

What, then, does the agent in 2026 actually do? The industry's answer is “expertise, negotiation, market knowledge”. Expertise is the residual product of the asymmetry the position itself perpetuates. Where information is public and cross-referenced, expertise becomes a search interface. Negotiation requires interests aligned with both parties simultaneously, which is impossible by structure when the same agent represents both. Market knowledge — comparable transactions, time-on-market, price-per-square-metre, absorption rates — is public data, filterable and free.

Each of these refutes structurally. The position survives by collecting commissions on functions the network has already absorbed.

When, like me, one sits on both sides of transactions — the buyer's and the owner's — for over two decades, one quickly begins to map exactly which functions the position adds and which it merely imposes itself upon. By the time the first decade rolls into the second in this sector, the map fits on a single page. In two or three lines.

There is an objection that the human relationship is itself the value. That a buyer wants a person to call when something goes wrong. This is not refuted. It is reframed. The trust the buyer seeks is in the outcome of the transaction. When the architecture verifies identity, property, payment, and contract against official sources in real time, the trust is delivered by the architecture. The person at the desk in Cascais on the Tuesday morning is the lawyer, who serves both parties under fiduciary obligation. The position the empty chair represents is not the trust. It was, until 2025, the substitute for trust where verification did not exist. Trust was the function. Verification absorbed the function. The position survived the function.

The commission of twenty-five to forty thousand euros on a five hundred thousand euro transaction does not retribute value created. It retributes the persistence of a position that delivered value in 1925, less in 1995, and almost none in 2025. Translated into hours of real labour, the rate is over a thousand euros an hour to forward emails and coordinate visits. More than a senior surgeon charges. For a function the infrastructure has already absorbed.

Throughout my career in real estate development, I have personally signed escrow instructions for approximately one point six billion euros in property transactions across twenty years. I have paid agents millions in commissions. I have logged the hours of agent work those commissions paid for.

I decided to build property rails at a notary's desk — outside Portugal, where I had assumed the functional structure might be different, only to find it identical — on a specific date, after signing escrow instructions for one transaction in which the commission cost dozens of times more than the lawyer, the notary, and the verification combined. The decision happened there.

The math is one I can do in my head, and the math is the reason this memo exists.

If the function is gone, why has the structure survived? Because no single missing condition was binding alone. Five had to converge before substitution became inevitable. The next section names them.

III. The Five Conditions — Why Now

The architecture described in the previous section is not new in concept. Each component has existed for a decade or more. What changed in early 2024 is that five conditions, each necessary and none sufficient alone, converged for the first time. Until that convergence, the substitution was theoretically possible but practically blocked. After it, the substitution is not just possible. It is structurally compelled.

The first condition is AI document processing at production accuracy. Document understanding models such as LayoutLMv3, Donut, and the multimodal generations of GPT have reached over ninety-nine percent accuracy on key-value extraction across structured document types since 2023. Production-grade systems from Hyperscience, Mindee, Klippa, and Rossum have been deployed in regulated workflows across banking, insurance, and government. Before 2023, AI in document workflows was assistive — a tool that flagged anomalies for human review. Since 2024, it is deterministic for structured documents. The certidão, the caderneta, the nota simple — all of these are structured. None requires human interpretation when the architecture has access to the source.

The second condition is eIDAS binding across twenty-seven EU member states. Regulation 910/2014 has been binding since 2016, but qualified electronic signature adoption did not reach mainstream maturity until eIDAS 2.0 (Regulation 2024/1183) and the rollout of the EU Digital Identity Wallet in 2024. Cross-jurisdictional digital identity is now legally equivalent to in-person presentation in every member state. Verification cross-referenced against Chave Móvel Digital, DNIe, or any other binding national source carries the same legal weight as a notary's wet stamp.

The third condition is SIBS-grade payment infrastructure operational in Iberia. SIBS processes over one billion transactions per year. Segregated escrow custody via API has been available to qualified B2B partners since 2022. SEPA Instant, mandated by Regulation 2024/886 from March 2024, settles five hundred thousand euros between two banked counterparties in seconds, not days. Real-time settlement of property-grade transaction values is no longer a privilege of institutional banking. It is commodity infrastructure.

The fourth condition is the digital-native buyer majority. First-time buyers under forty now constitute the structural majority of new residential transactions in Iberia. This is the first time the generational tipping point has been reached, and the implications for digital adoption are not gradual. A buyer or an owner who has never deposited a paper cheque, never queued at a teller, and signs nothing on paper has no sentimental attachment to the analogue procedures of real estate mediation. They expect verification. They expect transparency. They expect freedom. They expect traceability. They will not pay twenty-five thousand euros for the privilege of being told what their lawyer can confirm directly. The digital-native will not sign a six-month exclusivity clause with an agent that strips them of freedom and decision-making power over their own asset.

The fifth condition is the EU DSA reshaping the regulatory baseline. Regulation 2022/2065 has been binding since February 2024 for online intermediary services and very large online platforms. The DSA does not directly govern offline brokerage. It governs digital services. The structural consequence is asymmetric: as property transactions migrate to digital infrastructure, brokerage practices operating within that infrastructure inherit by default the transparency obligations the DSA imposes — algorithm transparency under Article 27, recommender system disclosure, uniform terms enforcement. The incumbent dual-representation model, opaque by design, is structurally exposed to regulatory challenge under DSA principles whenever it operates digitally. It is not yet banned. It is in a position that is defensible only as long as it remains offline.

Each of these conditions, taken in isolation, is insufficient. AI without eIDAS is information without legal weight. eIDAS without payment rails is signature without escrow. All of this without a digital-native majority is a solution without a market. And without the DSA shifting the regulatory baseline, the status quo remains the default. Five conditions had to converge. They have.

Each of the five conditions has arrived in my professional lifetime. The first to land was eIDAS in 2016. The next time I signed a cross-border contract digitally, I understood the position had just been retired. The position took another nine years to notice.

The substitution will repeat in every economy where the five conditions converge. Iberia is first because the conditions converged here first, not because the conditions are unique here.

If the conditions are met, why have no existing players built the architecture? The next section diagnoses each category structurally.

IV. What Replacement Looks Like — The Architecture

What replaces the position is not a marketplace. Not a digital agency. Not a better listing portal. None of these substitute the function. They sit alongside the agent, supplement them, or extend their reach. The replacement is something different in kind. It is infrastructure — property rails — a coordinated layer that performs the integration the position once performed, but through architecture rather than through people.

Property rails are not a product that competes with the agent. They are the technical fabric coordinating identity, authentication, escrow, legal coordination, and payment as a single deterministic operation. The category did not have a name. It does now.

The architecture has five mandatory layers. None is optional. None can be omitted without compromising the integrity of the whole.

The first is identity verification. A biometric capture cross-referenced in real time against eIDAS-binding sources — Chave Móvel Digital in Portugal, DNIe in Spain — produces a cryptographic statement that the official source confirms identity. Verification, not assumption.

The second is property authentication. Direct API integration with the Caderneta Predial digital, the Nota Simple, and the Land Registry retrieves title, encumbrances, and ownership chain from the official source itself. The legal status of the asset is the legal record, not a representation of it.

The third is transaction security. Segregated escrow custody on SIBS or SEPA Instant rails writes each lifecycle event to an audit trail no party can rewrite, and funds move only when verification preconditions resolve. The conditions are mechanical, the trail permanent.

The fourth is legal coordination. A lawyer assigned algorithmically rather than chosen by an interested party, with fiduciary obligations binding to both buyer and owner and a fee fixed in advance. The assignment optimises for transaction integrity, nothing else.

The fifth is payment rails. SIBS-grade settlement infrastructure, SEPA Instant, segregated client funds. The secure transfer of three hundred thousand or five hundred thousand euros becomes a deterministic operation, not a negotiation between trusted intermediaries.

These five layers operate under three structural principles, each of which is non-negotiable.

The first is prove, don't verify. The architecture does not certify identity, property, or capacity. It causes the official source to certify them, in real time, with the result written immutably to the transaction record. The platform is a coordinator of authoritative answers. It is not the source of those answers, and it does not interpret them.

The second is all points or zero. Verification is not graded. There is no “ninety percent verified” state. Every required cross-reference passes, or the transaction does not advance. This is the inverse of how traditional mediation operates, where partial information, friendly omission, and selective disclosure are not just possible but routine. Architecture makes partial impossible.

The third is transparency as architecture, not promise. Every cross-reference, every verification, every state transition is auditable for the lifetime of the transaction record. Transparency is not a service the platform chooses to offer. It is the structure of the platform itself. The buyer, the owner, the lawyer, and the regulator can each see the same record. They cannot see different records.

Twenty years inside this architecture's absence is twenty years of watching the same loss vector close every transaction in the same way. Information held back. Counterparty vetted in private. Outcomes shaped by the position's preferences. The architecture above is not utopian. It is what I needed and never had.

An algorithm has no friends.

It does not lunch with the seller and dinner with the buyer. It does not perform errands for one party at the expense of another. It does not receive incentives to filter information. It does not hide proposals. It does not block options. It does not show properties to one buyer and withhold them from another on personal grounds. It does not change its behaviour in the presence of A or B.

An algorithm pulls from official sources, cross-references identity with property with funds with contracts, and either passes or fails. No human override. No vested interest. No asymmetric loyalty.

The traditional position rewards opacity. The infrastructure makes opacity impossible. These are not architectures that compete. They are architectures that contradict.

The agent who is generous with the owner on Tuesday's coffee and frugal with the buyer on Wednesday's email is performing the function the architecture forbids. The function is structurally extractive. Generosity to one side is theft from the other. The agent's daily friendliness is the structural unfriendliness of the position.

The architecture does not eliminate humans from the transaction. It coordinates them. The lawyer remains, but is now assigned algorithmically rather than chosen by an interested party. The notary remains, but acts on documents that are already verified. The ambassador who conducts the property visit remains, but operates under a certified protocol. Humans are present where their presence is structurally required. They are absent where their presence was the substitute for verification that did not yet exist.

This distinction matters. The thesis is not that automation replaces every person in the transaction. The thesis is that automation absorbs the functions that did not need a person to begin with, and reassigns the functions that do need a person to people whose interests are not in conflict with either side.

Stripe built the rails on which payments now flow without merchant accounts. Wise built the rails on which currency moves without FX brokers. Property rails do the same work in the third domain. The position that once unified information, identity, authentication, documentation, payment, and legal coordination is replaced not by a faster agent but by the architecture that performs each function natively.

Every relevant cross-reference must resolve before a property reaches a buyer. In the traditional architecture, that number is zero. The agent's word is the verification, and the agent's word is unverifiable.

This is the category. It is not a vertical SaaS. It is not a marketplace. It is the orchestration layer that performs the integration the position once performed. Property rails do not compete with agents. They render the position unnecessary by absorbing the function.

Property rails describe what the replacement looks like. The next question is structural. Why has no incumbent built this, and what category of player is positioned to do so?

V. Existing Players — Surgical Diagnosis

The graveyard is more instructive than the survivors.

Capital deployed and destroyed across the last decade, across three categories of approach, runs into the tens of billions. Not one substituted the agent.

Failure and success both confirm the same conclusion. Nobody substituted the agent — including the winners.

Capital-intensive iBuyers: Casavo (over eight hundred million euros raised in equity and debt across multiple rounds), Opendoor (down eighty-four percent from its peak, reporting one point three billion dollars in net losses in 2025), Zillow Offers (shut down in 2022 with an eight hundred and eighty-one million dollar write-down), Redfin Now (shut down 2022), and Knock (multiple pivots) — each tested the proposition that pure capital could replace the agent. Each ran into the same ceiling. The model does not survive without distribution, and distribution at scale requires either humans or infrastructure. They chose humans back. The infrastructure was not yet built.

Even Casavo, the European reference case, concluded that the architecture had to absorb the human distribution layer. Its CEO, Giorgio Tinacci, declared in November 2025: “We believe the European real estate intermediation market will move toward consolidation, catalysed by a shift from franchising to direct networks and the impact of AI on agent productivity.” Casavo did not return to working with agents in submission. It returned to them because the infrastructure to make distribution autonomous had not yet been built.

Hybrid listings-and-tools plays at scale: QuintoAndar in Brazil, peak valuation five point one billion dollars, still operational and still scaling — chose a different path. They built portals plus CRM for agents. They became a better channel for the position they could have replaced. Success in PropTech, in this category, has meant becoming the rails on which the agent runs faster. The extraction continued. The platform subsidised it.

Agent-empowerment plays in public markets: Compass IPO'd in 2021 at seven billion dollars and currently trades at less than forty percent of its open price, having lost more than sixty percent of its market capitalisation. Compass merged with Anywhere, the parent of Century 21, in 2025. Even consolidation did not reverse the structural extraction. It just made the consolidation cheaper.

Listing portals: Idealista, Rightmove, Fotocasa, Zillow — are the most successful PropTech businesses ever built. Their paying customer is the agent. Disintermediating the payer destroys the P&L. The conflict is mathematical, not strategic.

I have been close enough to the strategy of two of the categories above to watch the conversation converge, every time, on the same point: distribution. Whoever has it controls the transaction. The companies above tried to buy it, partner with it, or replicate it. None concluded that it had to be replaced.

Three categories. Capital deployed and destroyed in the tens of billions. Every one of them, in its own way, confirmed the same conclusion. The architecture had not yet been built. The agent did not survive because the agent was strong. The agent survived because no one had built the substitute.

There is structurally room for one operator of property rails per regulatory perimeter. The first operational rails compound regulatory licensing, B2B integration lock-in, and verification API state into a position that subsequent entrants cannot replicate at equivalent depth. The first operational rails capture the perimeter.

If incumbents cannot build this, and the evidence shows they cannot, the question becomes how the architecture reaches the market. Three concrete paths follow.

VI. Distribution — How the Thesis Reaches Market

Every two-sided architecture faces the same problem at launch: which side recruits first. Get demand without supply, and demand churns. Get supply without demand, and supply loses faith. Sector evidence consistently shows marketplaces that invert the natural sequence struggle to reach liquidity. NFX and a16z have documented cold-start sequencing as a primary failure mode in two-sided platforms, and the failure is rarely about marketing spend. It is about which side of the architecture has structural reasons to participate first, before liquidity exists.

Property rails resolve cold start through three distinct distribution wedges, each grounded in evidence already on the ground.

The first is B2B-first inventory. Property developers, investment funds, and banks holding non-performing real estate hold significant supply that flows by default through agents. A developer with twenty-five units pays roughly three hundred and seventy-five thousand euros a year in agent commissions, assuming standard five percent rates and average prices. A larger project with two hundred units approaches three to five million euros. API integration converts that commission cost into a direct distribution channel in forty-eight hours. The decision is not a sales motion. It is a balance-sheet trade. Stop forfeiting one hundred and seventy-five thousand euros per unit in commission, in exchange for an integration that takes two days and produces a transparent transaction record. The architectural analogue is the early Stripe pattern. The flow that brings new participants onto the network is not a marketing campaign. It is a financial trigger that one party can act on unilaterally because the math is already on their side.

I have built developer balance sheets for years. The agency commission line is consistently the largest discretionary expense. It has remained discretionary because the alternative was not credible. Now it is.

The second is lawyers as an organic supply-side network. Hundreds of lawyers have registered through this architecture without a single euro of customer acquisition cost. The reason is structural. A lawyer coordinating a property closing is the only neutral professional in the transaction whose interests align with completion rather than with capture. Switching cost emerges naturally after the first transaction, because workflow integration, KYC stack, and document templating compound. A lawyer who has closed two transactions through this architecture has effectively zero incentive to return to the friction of unstructured paper-based closings. The supply side is already organic. The bottleneck was never demand.

The third is Iberia as a beachhead for the broader EU. The architecture has been designed to absorb new jurisdictional requirements without reconstruction. eIDAS uniformises identity binding and signatures across all twenty-seven member states. Notary law and fiscal frameworks remain national, but they sit on the same technical substrate. Each new geography is integration plus localisation, not a rebuild. The cost of expanding from Iberia into a third or fourth EU market is materially lower than the cost of building the first market, because the regulatory and technical foundations are already in place. The replication economics of the architecture compound with each new jurisdiction.

The sequence is supply first, demand second. Months one through six prioritise B2B integration partners and lawyer onboarding. Month seven onward focuses on demand-side acquisition once liquidity is secured. The precedent is consistent. Uber recruited drivers in target cities before opening rider acquisition. Airbnb built supply quality through hand-curation before scaling demand. Stripe built API completeness before announcing a single paid merchant. Marketplaces that invert this sequence rarely reach liquidity, because the side that arrives first without a counterparty churns, and the cost of regaining trust is rarely paid back by acquisition.

Three wedges describe how the architecture reaches market. The next section addresses the risks honestly.

VII. Risks — Honest Diagnostic

The thesis has risks. None destroys it. Each is structural, and so is its mitigation.

The first risk is PropTech category timing. The sector is not currently fashionable in venture capital. Investment in property technology peaked in 2021 and has cooled materially since. The mitigation is structural rather than tactical.

Property rails are not PropTech in the vertical sense. They are payment and verification infrastructure that happens to apply to real estate. The re-categorisation is part of the work this memo does. A category that had been read as adjacent to a cooling sector reads differently when properly framed as adjacent to fintech infrastructure.

The second risk is adverse regulatory shift. Incumbent agency networks could attempt to capture regulation in their favour, particularly at the national level in Portugal or Spain. The mitigation is in the data. The EU DSA already binds in the opposite direction. Agency networks are fragmented across roughly forty-seven thousand entities in Iberia alone, with no unified lobbying voice. The European Commission has a consistent track record of blocking incumbent capture in adjacent verticals. The Uber and Airbnb rulings established that information society services cannot be retroactively re-classified as transport or hospitality businesses through national lobbying.

The most concrete signal that the regulatory direction has already shifted is across the Atlantic. In March 2024, the United States National Association of Realtors agreed to a four hundred and eighteen million dollar settlement that materially restructured the buyer-broker commission model. For the first time in a century, the structural extraction was challenged not by technology but by the courts. Europe's regulatory direction under the DSA points the same way.

The third risk is incumbent replication. Established agency networks could attempt to build the architecture in-house. The mitigation is time and depth. Eighteen months of regulatory work cannot be shortcut. The cost of replicating the technology stack is in the range of five to ten million euros, before any of the proprietary verification API integrations. None of those integrations is publicly available. The B2B integration lock-in deepens with each institutional partner. By the time a credible incumbent attempts replication, the architecture has compounded for two cycles.

The fourth risk is AI document accuracy ceiling. AI does not yet achieve one hundred percent accuracy across every document type. The mitigation is the trust layer, not a technology fix. The lawyer assigned to each transaction coordinates the residual cases that the architecture cannot resolve autonomously. The human layer covers what the machine does not, and is structurally aligned with both parties because it is assigned algorithmically rather than chosen by an interested party.

What scenario, if true, destroys the thesis? Three candidates are material. The first: digital-native majority among first-time buyers regresses. Improbable. Eurostat confirms the demographic transition as structural rather than cyclical. The second: the European Commission revokes the DSA or suspends eIDAS. Improbable. Both are Regulations rather than Directives, revocation is a multi-year process, and the political direction of the Commission is reinforced by eIDAS 2.0. The third, and the only genuinely existential risk: institutional B2B integration does not convert into transaction volume. The mitigation is the same as the test. The ROI for the developer or fund is a balance-sheet trade, not a behavioural change. The validation point is the first three institutional contracts.

Stripe was told incumbent banks would crush it. Wise was told regulated FX was untouchable. Each time, the inversion was the same: the structure built, not the argument made, ended the gatekeeper's term.

I have argued this thesis to myself for years before writing it down. The first argument concluded only that the math did not work. I lacked the architecture. The architecture had to arrive before the argument could be operational.

This memo is not a claim that real estate agents are bad people. They inherited a position that made sense for a century and stopped making sense over the last decade, without anyone deciding it should stop.

This memo is not a manifesto against intermediation. Some intermediation is structurally necessary. Lawyers remain. Notaries remain. Both for reasons the architecture cannot replace — because the State has not yet dematerialised.

This memo is not a forecast of the death of mediation agencies. Many will continue to operate, particularly in segments where the buyer or owner actively chooses to delegate. The claim is structural. The position no longer governs the transaction. The infrastructure does.

Hugo Ferreira
CEO, Turnki