Portals, CRMs, and Traditional Agents: Which Part of Real Estate Survives the Autopilot?

In March 2026, Sequoia Capital published a piece called ‘Services : The New Software.’ It argues that the next trillion-dollar company won’t look like a software company. It will look like a services firm — but run entirely on AI infrastructure. The thesis is built on one uncomfortable ratio: for every dollar spent on software, six dollars are spent on services. Every major tech wave so far has played in the 1x market. The next wave plays in the 6x market.

Real estate brokerage sits at the top of Sequoia’s opportunity map.

Real estate is the world’s largest asset class at $393 trillion — bigger than global equities and bonds combined. And the industry selling it still runs on 1995 infrastructure.

The brokerage fee layer — the commissions paid to agents — amounts to nearly $200 billion per year in Europe and North America alone. Virtually none of it has been automated. The industry that handles the single largest financial transaction in most people’s lives has not fundamentally changed in three decades.

That is about to change. But not slowly, and not incrementally. This is a structural collapse of the incumbent model — and it is already underway. The question is not whether it happens. The question is who captures the category when it does — and why it took this long.

In most European markets, the average property owner transacts once or twice in a lifetime. But even in more active markets like the US or Australia, where owners move more frequently, the structural problem persists: no two transactions are comparable. Different market, different cycle, different agent, different life stage. Past experience does not translate into negotiating leverage. The agent has pattern-matched hundreds of deals in a specific market. The seller is always, effectively, doing it for the first time.

The AI-Framework: Intelligence vs. Judgement

Sequoia makes a distinction that sounds academic but has enormous practical consequences: the difference between intelligence work and judgement work.

The pattern is already visible in other industries. In software development, writing code, generating tests, and debugging syntax errors are intelligence work — GitHub Copilot does all of it today. What remains for the human engineer: deciding what to build, system architecture, the judgement call on whether to refactor or rebuild from scratch. In accounting, bookkeeping, tax calculations, and standard reconciliations are intelligence work — automated. What remains: advising on an acquisition structure, navigating a regulatory grey area, the conversation with a founder about what they actually need versus what they asked for.

Real estate is the same pattern — at larger scale and with higher emotional stakes. And it is approximately five years behind where accounting and software development are today.

Intelligence work is rule-based. Repetitive, learnable from data, automatable. Scoring a lead, generating a property valuation, writing a listing description, processing a follow-up sequence. The rules can be complex — but they are rules. AI has already crossed the threshold where it can do most of this autonomously.

Judgement work is different. It requires experience, instinct, and contextual awareness that cannot be extracted from a dataset. Deciding which offer to accept when three are on the table. Knowing when a buyer’s hesitation is cold feet versus a real problem. Building trust with someone in 45 minutes who is about to make the biggest financial decision of their life.

“The higher the intelligence ratio in any field, the sooner AI autopilots will win.” — Sequoia Capital, March 2026

In real estate, the intelligence ratio is very high. Most industry participants just haven’t mapped it honestly.

What This Means for Each Player

Traditional Brokers and Independent Agents

The traditional agent’s value proposition rested on information asymmetry. They knew the market, the comparables, the pricing intelligence. That asymmetry is gone.

The average traditional agent spends roughly 70–80% of their working hours on intelligence work — tasks AI either already handles or will handle within 24 months. This does not mean agents disappear. It means the value proposition shifts entirely.

The agent who survives is the one who stops doing intelligence work and starts doing exclusively judgement work: negotiation, trust, emotional navigation, strategic advisory. Equipped with AI infrastructure, that agent can handle 50 transactions per year instead of the current industry average of under 5. That is a 10x productivity multiplier.

The agent who doesn’t make this shift faces a different outcome. Commoditised. Competing on price in a market moving toward outcome-based pricing.

Agent CRMs and PropTech Tool Vendors

The dominant model in agent-facing technology has been the copilot. Better CRM. AI-powered follow-up. Auto-generated listings. Make the agent 30% more efficient.

This is Sequoia’s copilot definition precisely: a copilot sells the tool. The professional remains the customer, the tool makes them more productive, and they take responsibility for the output.

The structural ceiling: copilots capture the tool budget. The tool budget is 1x. The services budget — the commissions — is 6x.

“Selling the work means cutting their own customers out of doing it.” — Sequoia Capital

A CRM vendor whose customers are agents cannot simultaneously build the autopilot that replaces those agents. The business model and the product vision are in direct conflict. By 2030, agent CRMs face a stark choice: remain a productivity tool for a shrinking pool of high-performing agents, or fundamentally restructure as transaction infrastructure. Few will manage the transition.

Discussing the autopilot thesis at PPW Bangkok — portals sell the tool, platform brokerages sell the transaction. The distinction, at scale, is not incremental. It is categorical.

Real Estate Marketplaces and Portals

Portals have dominated the discovery layer of real estate for 20 years. Rightmove. Scout24. REA Group. The model: aggregate inventory, monetise with agent subscriptions and lead generation, raise prices every year.

The structural challenge portals face is not AI. It is that AI is eating their most defensible asset.

The portal’s moat was search — the ability to aggregate inventory in a way individual agents could not. That moat narrows as AI-native search makes the traditional portal UX redundant. When a buyer describes their needs in natural language and receives a curated shortlist, the reason to browse 1’200 listings disappears.

There is a deeper structural limitation that no amount of AI investment can fix from the portal side: portals have listings data, but they have no transaction data. They know what was listed, at what asking price, and for how long. They do not know the actual sale price, who bought it, how much was financed, why deals fell through, or what negotiation patterns led to a close. Their AI can optimise the discovery layer. It cannot build a model of what a successful transaction actually looks like — because that data only exists inside the transaction itself. A platform brokerage that owns the full process has both. Every closed deal adds a data point that a portal can never access. That asymmetry compounds with every transaction.

Portals are responding with copilot moves. Zillow launched a ChatGPT plugin. Zoopla integrated AI search. But these don’t change the fundamental revenue model. The portal still earns on visibility and leads, not on transactions.

Some have already moved further. Scout24 Group’s Immobilienverkauf24 subsidiary goes beyond lead generation entirely: through the ONE system, it provides agents with software, seller leads, buyer leads, valuations and remote transaction support — in exchange for up to 50% of commission revenue. The agent handles viewings and client contact. Scout owns the intelligence layer and takes a cut of the transaction. Germany’s largest portal is no longer just selling the tool. It is selling the work. Germany’s IVD real estate association compared it directly to Uber and Airbnb — platforms that captured the value chain while the original service providers became execution contractors.

Scout ONE is not a strategic pivot. It is a portal that watched Compass happen in the US, saw 80% of Germany’s independent agents structurally defenceless, and made the rational move. The direction is right. What it reveals is that Europe’s dominant portal has no answer to an AI-native platform brokerage — so it is building one from the agents up, before someone else does it from the platform down.

The same movement is visible in Switzerland. RealAdvisor began by selling visibility and leads to agents, then added a CRM to deepen platform dependency — classic copilot economics, building lock-in through the tool layer. They are now attempting a revenue share model, capturing a portion of the transaction rather than the listing fee. Whether this constitutes a genuine autopilot pivot or a revenue diversification move remains to be seen. But the direction confirms the thesis: every player with a discovery product is looking at the transaction layer.

In Asia, FazWaz and PropertyScout abandoned the pure portal model years ago and earn directly on commissions. The reason is structural: in most Asian markets, consumer purchasing power is too low to sustain the listing fee and agent subscription economics that Rightmove or Scout24 rely on in Europe. Without a thick subscription revenue base, portals had to go deeper into the transaction to capture value. The result — whether by necessity or strategy — is the autopilot model Sequoia describes. Europe’s portals have not faced this pressure yet. That does not mean they won’t.

If portals move into the transaction layer, they face the same structural limitation as CRM vendors — their customers are agents, and building an autopilot means marginalising those customers. The portal that solves this contradiction will become one of the most valuable companies in the industry. Most won’t.

Platform Brokerages

This is where the Sequoia thesis becomes most interesting — and most demanding.

A platform brokerage is not a traditional broker, not a CRM vendor, not a portal. It is the category Sequoia’s autopilot thesis describes: a company that sells the outcome — the closed transaction — directly to the seller or buyer, using AI to handle the intelligence layer and human agents to handle the judgement layer.

This is not theoretical. properti has closed over 11’000 transactions in Switzerland on exactly this model since 2019 — earning on the transaction, retaining the customer data, and compounding the AI infrastructure with every deal. The Swiss market is the proof of concept. DACH and Europe is the scale opportunity.

Every deal that closes on the platform makes the next deal smarter — better valuations, better lead scoring, better matching. This is the compounding data moat Sequoia identifies as the long-term defensibility of autopilot businesses.

The model requires brutal clarity on one strategic question: is the platform selling the work, or selling the tool? A platform that earns revenue share on every transaction — and retains the customer data from every deal — is in autopilot economics. The difference at scale is not incremental. It is categorical.

The fragmentation problem in traditional brokerage makes the platform advantage even larger than it first appears. The average independent agent today operates across six to ten separate tools: a CRM for listings and contacts, a valuation tool for pricing, a document platform for contracts, a communication tool for follow-up. Each solves one slice. None compound. Every transaction starts from scratch because the data never flows between layers.

This fragmentation is not accidental. It is the natural result of a market where copilot vendors each compete for a single budget line. Every tool optimises for its own metric — portal clicks, CRM logins, valuation requests. Nobody optimises for the closed deal. The seller pays the price in friction, fallthrough risk, and a process that feels exactly as disjointed as it did in 1995.

A platform brokerage with an integrated stack does not just replace one of these tools. It replaces the entire assembly. Data create leads. Valuation informs lead scoring. Lead scoring informs outreach timing. Outreach timing informs negotiation position. Every layer feeds the same data model and every transaction makes the next one smarter. The compounding effect is not additive. It is multiplicative.

The advantage extends beyond the transaction itself. The average independent agent or traditional brokerage also lacks the operational infrastructure that makes a modern service business function: centralised lead generation, systematic training and development, centralised compliance and legal support, professional debtor management, an one tier-community, shared back-office and finance infrastructure, and a brand that compounds across every deal rather than starting from zero with each new client. A platform brokerage provides all of this as shared infrastructure — owned centrally, available to every agent on the platform.

The result: agents on a well-built platform can reach 50+ transactions per year while the market average sits below 5. That is not just AI. It is the elimination of every operational distraction that prevents a skilled negotiator from spending their time negotiating.

The Outsourcing Wedge

Sequoia describes the autopilot playbook as starting where outsourcing already exists:

“If a task is already outsourced, it tells you three things: the company has accepted that this work can be done externally, there’s an existing budget line that can be substituted cleanly, and the buyer is already purchasing an outcome.”

Real estate brokerage is entirely outsourced. Every seller who lists a property hires an external agent. The budget exists — typically 2–4% of the transaction value. The substitution is frictionless: an old agent contract is replaced by a platform contract. No internal restructuring required.

This is what makes real estate brokerage one of the largest remaining autopilot opportunities on the planet. It is not a market that needs to be educated about outsourcing. It has always outsourced. It just hasn’t found a platform worth switching to. When it does, the value transfer will not be incremental. It will be total.

The Convergence: Why This Matters More Than It Appears

“Today’s judgement will become tomorrow’s intelligence. As AI systems accumulate proprietary data about what good judgement looks like in their domain, the frontier will shift.” — Sequoia Capital

The platform that closes 50’000 transactions in a market does not just have revenue. It has a model of what good judgement looks like in that market — one no individual agent and no portal without transaction data can replicate.

This is why timing matters more than execution speed in this category. A platform that started building its transaction data asset in 2019 is not a few years ahead of a 2026 entrant. It is potentially a decade ahead in model quality. The compounding nature of data advantages is not linear.

The Future of the Agent

The future of the agent is not extinction. It is bifurcation.

The number of agents will fall significantly. In Switzerland alone, over 5’000 agents close fewer than 5 transactions per year. Most of that activity disappears — absorbed by platforms, AI systems, or simply by a market that stops tolerating mediocre service. The agents who remain will handle more transactions, at higher value, with better tools. Expectations rise. Tolerance for underperformance collapses.

What remains will look like two distinct models — and both have genuine legitimacy.

Fixed-price neo brokerages — charging a flat fee instead of a percentage commission — solve a real problem: commission opacity. A seller who understands what they need, wants to control the process, and does not want to pay a percentage on a CHF 2M property has a rational case for a flat fee. The value proposition is transparency and cost reduction. It is a legitimate market position for a specific segment of sellers.

Platform brokerages solve a different problem: outcome quality at scale. A seller who wants the highest achievable price, the fastest close, and the least friction does not primarily want a cheaper agent. They want a better one — backed by AI infrastructure, transaction data, and a team that has executed this process hundreds of times in their specific market. The value proposition is not cost. It is result.

The critical difference is incentive alignment. A fixed-price agent earns the same whether the property sells for CHF 900’000 or CHF 1’200’000. A platform brokerage that earns on the transaction outcome is structurally aligned with the seller’s interest. That distinction matters more than any fee comparison.

Both models will coexist. They are not competing for the same customer — and the market is large enough for both. What disappears is the middle: the average agent charging full commission, delivering average results, with no data advantage and no platform infrastructure to justify either.

Where This Ends

The real estate industry in 2030 will look structurally different from today — not because all agents disappear, but because the distribution of value shifts.

Traditional agents without platform support face commodity pricing and shrinking share. CRM and PropTech tool vendors consolidate — survivors specialise or get absorbed into platform infrastructure. Portals face the most uncertain future: those that move decisively into the transaction layer have a path, those that keep optimising the listing interface are optimising a product whose primary use case is being eroded by AI search. Platform brokerages with genuine transaction data, AI infrastructure, and autopilot economics compound into dominant positions.

The Sequoia thesis is not about technology. It is about which business models capture the 6x market and which stay stuck in the 1x market.

The window is open. But it has a hard close.

The first platform brokerage to close 20’000 transactions in the DACH market will own a proprietary data asset — on valuations, buyer behaviour, negotiation patterns, timing — that no competitor can buy, replicate, or shortcut. That threshold is approximately 18–24 months away from being claimed. After that, the category leader is set and the compounding gap becomes insurmountable.

Every portal, CRM vendor, and traditional brokerage reading this article is deciding right now — whether they know it or not — which side of that gap they will be on.