Jevons' Revenge: How a 160-Year-Old Economic Law Is About to Hand the Next Decade of American Housing to Realtors

By Tim Harris · April 28, 2026

🎧 Stream us on Apple Podcasts and Spotify!

On a Tuesday morning in San Francisco last October, Marc Benioff stepped onto a stage at Salesforce's Dreamforce conference and told a room full of executives that artificial intelligence is now doing roughly half of the work at his company. A few weeks later, Ford CEO Jim Farley said AI will "replace literally half of all white-collar workers." JPMorgan told its managers to stop hiring. Goldman Sachs CEO David Solomon said his bank is rebuilding its org chart "front to back." Amazon cut 30,000 corporate jobs. Salesforce laid off 4,000 customer support reps in a single September week.

If you read those headlines and concluded the American economy is heading for a cliff — and that housing is going down with it — you are about to make the most expensive misread of your career.

Because while the cable-news doom merchants were busy filming thumbnails of crumbling skyscrapers, something else was happening, quietly, in the data:

The National Association of Realtors is now forecasting a 14% jump in existing home sales for 2026. The number of practicing radiologists — the profession Geoffrey Hinton famously declared dead in 2016 — has hit an all-time high, with average radiologist income up 48% since 2015 to roughly $520,000. And 88% of homebuyers and 91% of home sellers in the most recent NAR Profile used a real estate agent in their last transaction.

Welcome to the Jevons Paradox. It is the single most important economic concept in American real estate right now, and almost no one in the industry is talking about it.

In January 2025, when the Chinese AI lab DeepSeek released a model that did roughly what OpenAI was doing for a fraction of the price, tech stocks cratered. Microsoft CEO Satya Nadella's response on social media was three words long: "Jevons paradox strikes again." He went on: as AI gets cheaper and more accessible, "we will see its use skyrocket, turning it into a commodity we just can't get enough of."

Nadella was citing William Stanley Jevons, a 26-year-old English economist who in 1865 published The Coal Question and noticed something nobody expected. James Watt's steam engine had just made coal-burning dramatically more efficient. Common sense said England would burn less coal. The opposite happened. British coal consumption tripled by 1900. Why? Because when the cost of using a resource collapses, total demand for that resource doesn't politely decline — it explodes.

Cheaper coal meant more factories. More factories meant more trains. More trains meant more cities. More cities meant more coal. The efficiency gain didn't shrink the market. It built an empire.

Now run the same equation on cognitive work — the writing, analyzing, summarizing, scheduling, drafting, researching, and decision-supporting that defines white-collar America. The marginal cost of all of it is collapsing toward zero. And the doom narrative says the same thing the 1860s critics said about coal: this efficiency will destroy demand for the resource (in this case, human workers).

Jevons would tell them they are wrong. The historical pattern is unambiguous. Y Combinator CEO Garry Tan put it this way in a recent keynote: AI will not eliminate human labor; it will redefine it. Box CEO Aaron Levy said it more bluntly: "The reason AI isn't going to wipe out jobs in the way that some predict is that we consistently make the mistake of thinking that when we make something more efficient, you need commensurately less supply. It turns out that in a significant number of fields, better productivity levels actually means more demand for that service."

But — and this is the part that matters — the paradox does not save everyone. It saves a very specific kind of worker.

Let's be honest about the carnage, because if we don't name it, we sound like cheerleaders.

According to Challenger, Gray & Christmas, U.S. employers announced over 1.2 million job cuts in 2025 — the most since 2020. Of those, only about 4.5% were officially attributed to AI, but modeling-based estimates from PwC and McKinsey put the actual AI-displaced or AI-foregone figure at 200,000 to 300,000 jobs. The Bureau of Labor Statistics reported in early 2025 the lowest rate of professional-services job openings since 2013 — a 20% year-over-year drop. About 40% of white-collar job seekers in 2024 didn't land a single interview. Computer science graduates are facing a 6.1% unemployment rate — nearly double the rate for philosophy majors.

This is what Anthropic CEO Dario Amodei has been warning about for two years. He has publicly predicted AI could wipe out half of all entry-level white-collar roles by 2030. He may turn out to be early. He will not turn out to be wrong.

Here is the pattern in the actual layoff filings. The jobs being eliminated, compressed, or simply not refilled share one trait. They live entirely inside a screen.

Look at that list. Every job has the same DNA: a desk, a monitor, a keyboard, and zero requirement to physically be anywhere or look anyone in the eye. These are pure cognition jobs in fluorescent-lit cubicles. They live exactly where AI lives. And AI does them faster, cheaper, and at 3 a.m. on Christmas Eve without complaining about overtime.

The doom narrative gets this part right. The screen jobs are in trouble. The 22-year-old finance major who thought a remote analyst seat at a Fortune 500 was a safe bet just discovered it isn't. That is a real problem for a real generation of workers, and pretending otherwise is dishonest.

But here's where the story turns.

Erik Brynjolfsson, the Stanford economist who literally wrote the book on AI and labor, said it cleanly in a recent interview: physical jobs and trust-based jobs "are so far shielded from AI disruption." That category has a clear floor and a clear ceiling. The floor is anything that requires a human body to be physically present in a real-world location. The ceiling is anything that requires a human being to be trusted by another human being.

Plumbers. Electricians. HVAC technicians. Roofers. Framers. General contractors. Home inspectors. Appraisers walking the property. Locksmiths. Landscapers. Painters. Pool guys. Mortgage loan officers who actually meet their clients.

And — this is the entire ballgame — real estate agents.

These are not jobs threatened by AI. These are jobs about to be supercharged by it. And the proof is already in.

In 2016, Geoffrey Hinton — the man who later won the Nobel Prize for inventing the foundations of modern AI — stood at a conference and declared that medical schools should "stop training radiologists now." Within ten years, he said, AI would read scans better than humans and the profession would be finished.

Ten years later, here is what actually happened, drawn directly from the Bureau of Labor Statistics, the American College of Radiology, and a recent CNN Business deep-dive on the field:

Hinton himself has now publicly walked it back. He told The New York Times last year that he now believes AI will simply "make radiologists a whole lot more efficient." Even Nvidia CEO Jensen Huang, who probably has a stronger commercial incentive than anyone alive to claim AI replaces human workers, has gone on record saying AI is boosting demand for radiologists, not killing it.

What happened? When AI compressed the cost and time of reading a single scan, doctors didn't order the same number of scans more cheaply. They ordered dramatically more scans, on more patients, for more conditions, earlier in the diagnostic process. The pie didn't shrink. It exploded. The slice that AI automated grew slower than the new pie that AI made possible. That, in a single sentence, is the Jevons Paradox in 21st-century clothing.

This is not a fringe theory. The Y Combinator CEO is talking about it. The Microsoft CEO is talking about it. Erik Brynjolfsson is talking about it. The CNN Business piece on radiologists, published in February 2026, called it "a rare win-win scenario where economic efficiency and human employment rise together."

Now apply that exact mechanism to the residential real estate transaction. It is the same shape, exactly.

Five things. Pay attention to all five, because the agents who internalize them in the next twelve months are going to look like prophets in 2028.

In November 2025, NAR Chief Economist Lawrence Yun delivered the 2026 housing forecast in Houston. His top-line number: existing-home sales are projected to rise approximately 14% in 2026, with home prices up another 4%. "Next year is really the year that we will see a measurable increase in sales," Yun said. "Home prices nationwide are in no danger of declining."

Most of the analyst class is attributing this to easing mortgage rates and pent-up demand. They're partly right and missing the bigger story. The friction cost of a real estate transaction — the search, the disclosure paperwork, the comp-pull, the marketing copy, the buyer-matching, the showing logistics, the document prep — is collapsing right now because of AI. When friction collapses in a market with deep latent demand, the Jevons Paradox guarantees one outcome. Volume detonates. A 14% jump in 2026 is not a peak. It is a starting line.

The 2025 NAR Profile of Home Buyers and Sellers reported that 88% of buyers and 91% of sellers used a real estate agent in their most recent transaction. That number has barely moved in twenty years, through Zillow's launch, Redfin's launch, the iBuyer wave, the commission lawsuits, and the entire portal era. Why? Because the actual job of a real estate agent has very little to do with information transfer and almost everything to do with what AI cannot do: walking a buyer through a house and reading their face, knowing the seller is bluffing, calling the listing agent and pulling the truth out of them in 90 seconds, talking a first-time buyer off the ledge at 11 p.m. on a Saturday, and physically getting deals across the finish line through the seventeen things that go wrong between contract and close.

AI handles the data layer. The agent handles the trust layer. As AI strips out the tasks that never made the agent valuable in the first place, the agent's value goes up, not down.

If U.S. home sales rise 14% in 2026 — and the productivity revolution underneath that number is permanent, not cyclical — then the cascade through the trades is mathematically guaranteed. More transactions means more inspections, more pre-listing repairs, more renovation projects, more remodels, more new builds, more move-in punch-lists. The plumber, the electrician, the HVAC tech, the roofer, the GC — these workers are about to set their own prices for the rest of the decade. Parents who steered their kids out of trade school and into "safe" desk careers are about to look catastrophically wrong. The 24-year-old master electrician working in any growing metro is, on a risk-adjusted basis, in a stronger labor position than a 24-year-old data analyst at a Fortune 500.

This is the part most coverage is missing entirely. AI doesn't just make selling a home cheaper. It makes building one cheaper, faster, and more permittable. AI-assisted architectural design is collapsing iteration cycles. Generative permitting tools are starting to compress what was a 6-month entitlement nightmare into weeks. Construction-management AI is helping small and mid-size builders run schedules and material orders the way only national homebuilders used to. Off-site and modular construction — which depends on cheap, fast design and engineering — is finally getting an economic tailwind.

If the Jevons Paradox holds on the supply side the way it has on the demand side, we get more homes, built faster, at lower marginal cost, by the same builders. The chronic undersupply that has driven the affordability crisis since 2008 finally cracks. Not because of a government program. Because of a 160-year-old economic law applied to drywall and permits.

This is the part nobody likes to say out loud, so we'll say it. The part-time agent who closes four houses a year, doesn't return calls until Tuesday, and runs their business out of generic templates? AI is going to eat them. Not because AI will replace agents. Because AI-equipped agents will replace them. The agent who shows up, builds relationships, knows their inventory cold, runs an actual business, and uses AI as a force multiplier on every administrative and content task is about to do the volume of a 15-person team — solo, with one assistant.

The same shift is hitting brokerage models. When a producing agent doesn't need 47 admin staff and a giant corporate back office to do the volume of a 47-staff team, the economics of the traditional brokerage stop working. This is exactly why agent-centric, cloud-based, revenue-sharing models — eXp Realty being the obvious example, with Libertas as the most aggressive co-sponsoring brand inside it — are going to be the structural winners of the next decade. The fixed-cost brokerage was built for a world where information was scarce. AI just made information free. The agent is the product. Everything else is overhead.

Now, the strongest objection. It's the one we hear in every coaching call: "What about Zillow? What about Opendoor? What about the commission lawsuits? What about the day an AI agent does the whole transaction?"

Take them one at a time.

Zillow and the portals. Zillow has had two decades to disintermediate the agent. Result: 88% of buyers still use one. The portal owns the lead-gen layer; it has not touched the relationship layer, and it cannot, because the relationship layer is not an information problem.

Opendoor and the iBuyers. Opendoor's stock is down roughly 95% from its peak. The iBuyer model fundamentally misread what consumers want. Most sellers do not want a robotic, slightly-below-market cash offer. They want maximum price plus a human who handles the chaos. The iBuyer experiment is over and the market answered.

Commission lawsuits. The 2024 settlement changed how compensation is disclosed and negotiated. It did not eliminate the agent. It eliminated the bad agent — the one who couldn't articulate their value. Top producers report that buyer agency conversations have actually clarified their pitch. The settlement was a filter, not a death sentence.

The full-stack AI agent. The single most-cited fear. Here is the reality: a residential real estate transaction is a federally regulated, state-licensed, fiduciary-duty event involving the largest financial decision most American families ever make, in physical buildings in physical neighborhoods, with humans on both sides who require trust to sign. Even radiology — which is digital pixels read by a doctor — has been unable to regulatorily, legally, or culturally hand the work over to AI in the past ten years. The idea that residential real estate, which is radically more physical, more local, more emotional, and more legally entangled, will fall to a chatbot in the next decade is not a serious argument. It is a thumbnail.

Strategy without action is fan fiction. Here is what serious agents do this week.

Stop doing your own admin. Every minute you spend writing your own listing description, scheduling your own showings, drafting your own follow-up emails, building your own CMAs from scratch, or formatting your own marketing materials is a minute the Jevons Paradox is begging you to give to AI so you can spend it face-to-face with humans who pay you.

Build the AI stack now, not in 2027. Pick your tools — there are dozens — and integrate them into your daily workflow this quarter. Listing copy, transaction summaries, market reports, social content, lead follow-up sequences, contract review prep. The agents who are 18 months ahead on this in early 2026 will be uncatchable by 2028.

Double down on the parts of the business AI cannot touch. Door-knocking. Open houses. Past-client calls. Coffee meetings. Hand-written notes. Showings. Negotiations. Closings. These are the activities that increase in value as everything else commoditizes. This is the entire game.

Reposition your brokerage economics. If you are paying a giant fixed-cost overhead for services that AI now handles for the cost of a streaming subscription, you are subsidizing a structure built for a world that is ending. The agents winning the next decade are operating leaner, on platforms that pay them back the savings. This is the structural argument behind eXp's growth and the case Libertas has been making to thousands of agents about co-sponsoring inside that ecosystem.

Sell the trades into your sphere. Every client you ever closed needs a plumber, an electrician, an HVAC tech, a handyman, a roofer. Become the connector. The agent who is the go-to referral hub for a metro's best tradespeople is the agent whose database compounds in value every year for the next twenty.

The American housing market is not heading into a crash. It is heading into the most explosive volume decade since the early 2000s — and this time, without the subprime time bomb under it. NAR's 14% forecast for 2026 is the front edge of a Jevons-paradox wave that the people behind keyboards in cubicles cannot see, because they are watching their own jobs disappear and assuming everyone else's must be next.

They are wrong. The desk jobs are getting cleared out. The handshake jobs — the plumbers, the electricians, the roofers, the contractors, and the realtors — are about to be the most valuable workers in the American economy.

AI is not your competition. AI is the most powerful tool ever placed in the hands of a salesperson. The plumber doesn't compete with the wrench. The surgeon doesn't compete with the scalpel. The radiologist did not compete with the algorithm — and is now earning $520,000 a year because of it.

The screen jobs are going away. The handshake jobs are about to print money.

Pick the right side of the paradox. The data is already in.

If you’re a producing agent and your brokerage still treats you like an employee…
it’s time for a broker upgrade.

Two paths:

📲 Text: Tim — [512-758-0206]

Want intel first? 🌐 https://whylibertas.com/harris (watch the 3 videos + $40K+ benefits)

Partner with Tim & Julie Harris at eXp Realty. Let’s go.

— Tim Harris
Host, Power House Talk

💬 Want your question featured on an upcoming podcast? Reply to this email or DM us on Instagram— we might break it down next.

📩 Know an agent who should be in this room?
Forward them Power House Talk.