The U.S. residential brokerage industry collects roughly 5–6% of every home sale in commissions. In Citrini Research's 2028 scenario, agents acting on behalf of buyers and sellers compress that take to under 1%. That is not a discount. That is the disappearance of an entire occupational class.
The polite way to describe most of the modern services economy is "trust intermediation." The accurate way is friction. A real estate agent is paid 6% because the buyer cannot easily see the comps, schedule the tours, write the offer, and read the contract. A travel agent was paid 10% because flights, hotels, and rental cars lived in different reservation systems. A DoorDash courier and a 30% platform take exist because nobody wanted to call the restaurant, drive there, and pay in cash. Software ate part of that stack. Agents are eating the rest.
What disintermediation actually looks like
Strip the marketing language and a consumer agent in 2027 is a process that does four things: searches across multiple databases simultaneously, fills in forms it has filled in a thousand times before, negotiates within bounds you set, and signs on your behalf when the terms clear those bounds. None of these are technically novel. What is novel is that they are now cheap enough to run on every transaction in your life, every day, in parallel.
When that happens, the economics of intermediation collapse in a specific pattern. Citrini's projection puts real-estate commissions under 1%. The same logic compresses food delivery, which already runs on margins thin enough that DoorDash's contribution profit per order can swing on a few cents of routing efficiency. An agent that schedules pickups, batches three orders into one trip, and routes around traffic shaves the courier-side cost the platform's algorithms have already optimized for years — and gives the consumer the leverage to refuse the markup.
Travel intermediaries, insurance brokers, and small-business loan officers all run on the same machinery: information asymmetry plus paperwork tolerance. Both inputs are getting marked to zero.
The capital-flow problem hiding behind it
The displacement story everybody tells is about the agents, the brokers, the couriers. The harder story is what happens to the platforms whose entire stock-market valuation was built on collecting that friction.
DoorDash, Uber, Booking, and the long tail of SaaS marketplaces priced themselves as toll roads. Their margins depended on being the only feasible way to complete a transaction. Once consumer agents can route around the toll — or negotiate it down — the implied terminal margin in those discounted-cash-flow models is wrong by a factor that compounds every quarter. ServiceNow's projected ARR growth slowing from 23% to 14% in the same scenario tells the enterprise-software version of the story, and that is software whose buyers are corporations with procurement departments. Consumer-facing platforms have less defense, not more.
The signature is visible early. Mastercard revenue growth slowing to 3.4%. Card interchange under pressure as agents route consumers to lower-fee rails. White-collar job openings down 15% year over year. Each of these is a single sector reading the same news. Frictionless transactions are great for the consumer in any single case. They are catastrophic for the equity in any business whose terminal value was the friction.
This is the part the McKinsey deck about "AI productivity gains" does not address. The productivity is real. The capture is the question. If a consumer saves $30,000 in commission on a home sale, that money is not destroyed; it shows up in the buyer's bank account or the seller's. But the $30,000 used to flow through a brokerage that paid two agents, a broker, and a chain of vendors. Those wages are gone. The capital that financed the brokerage chain is impaired. The compounding effect on local economies — which is how the Yale and Sonnenfeld work on entry-level employment is already documenting — is what turns a productivity story into a demand story.
The honest question
The question worth asking is not whether agents will collapse intermediation margins. The capability is here, the cost is falling, and the consumer incentive is overwhelming. The question is which structures pick up the slack when an entire layer of the labor market — the layer that converted moderate education into stable middle-income work through the friction of human transactions — is compressed into a sliver of its former size.
Citrini's piece treats this as the prelude to a credit cycle, not the climax. If they are even directionally right, the firms whose business model is the friction are not going to be re-rated cyclically. They are going to be re-rated structurally. What replaces them is the open question of the decade.