AI Is About to Break the Real Estate Commission Model
AI Is About to Break the Real Estate Commission Model — Here's What It Means for Home Sellers and Agents A new industry report says consumers are overpaying for roughly half the cost of selling a home — and that the era of the automatic 6% commission is running out of road. For decades, the math on selling a home felt fixed. You hired an agent, you handed over somewhere between 5% and 6% of the sale price, and you didn't ask too many questions about what that money actually bought. That arrangement is now being pulled apart — not by a lawsuit, not by a discount brokerage, but by artificial intelligence and the consumers learning to use it. A new 32-page report from the consultancy Alloy Advisors, titled "The Home Sale Transaction, Reconsidered," makes a blunt argument: out of the 23 distinct tasks an agent performs during a home sale, real estate agents clearly outperform AI on just three of them. And while those three tasks are genuinely valuable, the report estimates they are worth thousands of dollars less than what most sellers currently pay. The authors — industry veterans Amit Kulkarni and Russ Cofano — conclude that about half of the total cost of selling a home is "overpriced," largely because software and AI have crushed the market value of work that used to justify a full commission. The headline prediction is the part that should make every agent sit up: "The commission standard is likely to break within 5 to 7 years" — meaning 5.44% stops being the default assumption. Notably, the report does not predict that agents disappear. It predicts that the flat, skill-blind percentage disappears. Those are very different futures, and understanding the difference is the whole game.
The $39,660 Question: What It Actually Costs to Sell a Home
Start with the number that anchors the entire report. On a typical $400,000 home sale, Alloy Advisors calculates total "hard" transaction costs of $39,660 — roughly 9.92% of the sale price — flowing out to third parties. The seller shoulders about three-quarters of that burden, around $30,200, while the buyer absorbs roughly $9,460 at closing beyond their down payment. Of that $39,660, real estate commissions account for about $23,000 — equal to 5.75% of the sale price and a striking 76% of all seller-paid friction. The remaining costs are spread across the usual suspects: transfer taxes, owner's title insurance, settlement and escrow fees, loan origination and underwriting charges, the appraisal, and the home inspection. The report also flags a quieter line item it calls a "platform tax" — the portal referral fees and MLS charges baked into an agent's commission that almost never show up as a standalone charge a consumer can see. Where the Money Goes on a $400,000 Sale Approximate Cost Real estate commissions (both sides) $23,000 (5.75%) Transfer taxes, title insurance, settlement fees Part of the remaining $16,660 Loan origination, underwriting, appraisal Part of the remaining $16,660 Home inspection and related buyer costs Part of the remaining $16,660 Total third-party transaction cost $39,660 (≈9.92%) Estimated "overpriced" portion $17,000 – $22,000 That last row is the provocation. The report argues that $17,000 to $22,000 of the total — the bulk of it tied up in agent commissions — is priced above what the work is now worth. As the authors put it, "The 3% listing commission was priced when agents controlled scarce information: access to listings, buyers, and valuations." Their verdict on that era: "AI has collapsed the market rate for most of that work to near-zero."
Why Commissions Didn't Fall After the NAR Settlement
If you followed the National Association of Realtors (NAR) commission lawsuit and its $418 million settlement, you might assume this compression already happened. The rule changes that took effect in August 2024 were supposed to unbundle buyer-agent commissions and let competition drive rates down. It largely didn't happen. The report cites data showing the national average commission actually rose to 5.44% in mid-2025, up from 5.32% the year before. A separate industry survey from spring 2025 found that 58.8% of agents said their buy-side commissions hadn't changed since the settlement, and nearly 12% said their commissions had gone up. So why has a percentage that everyone expected to fall proven so stubborn? - Seller-paid buyer commissions never really went away in practice. The business-practice changes shifted the paperwork, not the underlying habit of the seller covering both sides. - À la carte brokerage is effectively banned in many places. Thirteen states plus Washington, D.C., restrict the unbundled, pay-only-for-what-you-use models that would let price competition take hold. - Agents only get paid when a deal closes. The all-or-nothing, contingent-fee structure discourages the kind of itemized, hourly pricing that consumers can comparison-shop. The Alloy Advisors thesis is that regulation alone was never going to move the number — but informed consumers armed with AI are a different kind of pressure entirely. When a seller can ask an AI to evaluate every line item, model alternatives, and benchmark a quoted commission in real time, the negotiation stops being lopsided.
The Two Tiers Hiding Inside Every Commission
The most useful framework in the report is its split of the agent's job into two tiers. Tier 1 is the work AI has already commoditized. Tier 2 is the "human core" that still resists automation. The gap between what each is worth and what the commission charges is where the disruption lives. Tier What It Includes and What It's Worth Tier 1 — AI-compressed tasks Comparative market analyses, MLS data entry, listing descriptions, offer modeling, transaction coordination, and basic disclosure checks. Pre-AI market value: roughly $1,500 – $3,500 per listing. Today, the marginal cost for a competent AI user is close to zero, aside from $10 – $30 per deal for workflow software. Tier 2 — Human-value tasks Skilled negotiation, emotional coaching, on-site judgment, hyperlocal knowledge, and licensed fiduciary accountability. Estimated worth: roughly $2,000 – $6,500 per transaction — and critically, it does not scale with the price of the home. Here is the uncomfortable arithmetic. A 3% listing commission on a $400,000 home is $12,000. On a $1.5 million home, that same 3% is $45,000 — for the same Tier 2 work. The negotiation isn't four times harder. The emotional support isn't four times deeper. The local knowledge isn't four times sharper. As the report frames it: "On a $400,000 home, sellers pay $12,000 for that work. On a $1.5M home, they pay $45,000 for the same work." The commission model's indifference to skill and effort is, in the authors' words, the heart of the problem. There's a second wrinkle. Consumers can't reliably tell a top-decile agent from a median one before they sign — yet both typically charge the same percentage. That information gap is exactly the kind of thing AI is built to close, which is why the report sees premium pricing increasingly attaching to demonstrable performance rather than a standard rate card.
The Three Things AI Still Can't Do
This is the part agents need to internalize, because it's also the path forward. Of the 23 tasks studied, AI clearly beats humans on 10 — most of them low-skill clerical work like MLS entry, writing listing descriptions, and scheduling showings. Humans and AI both contribute meaningfully on six, including home valuations, staging recommendations, and offer review. Two tasks legally require a human, such as licensed fiduciary accountability. The winner was inconclusive on two more — repair priority guidance and disclosure document review. That leaves exactly three tasks with a clear, durable human advantage — and they happen to be among the most valuable in the entire transaction: - Negotiation execution. Cofano draws the analogy to buying a car: most people hate negotiating on their own behalf, they're not good at it, and the stress is real. A skilled agent acting as an active negotiator and buffer is worth real money — and the report's authors don't see that going away soon. - Hyperlocal tacit knowledge. The instinct for which street floods, which HOA is a headache, why one cul-de-sac commands a premium two blocks from an identical one — this is knowledge that lives in experience, not in a public dataset an AI can scrape. - Emotional support and crisis management. Selling a home is one of the most stressful financial events in most people's lives. Talking a seller off a ledge when an inspection goes sideways, or holding a deal together through a financing scare, is human work. But there's a sharp caveat the authors stressed in interviews: the quality of the agent decides whether the human or the AI wins. As Kulkarni put it, "You've got to really know your business." Cofano was blunter — an agent doing three deals a year loses to AI, while "highly-trained negotiators will win out over AI for some period of time, if not forever." In other words, AI doesn't replace agents. It replaces average agents and rewards excellent ones.
The Trust Gap That Keeps Skilled Agents in the Game
AI's reach into real estate is already wide. The report cites 2025 research showing about 82% of active or potential buyers and sellers have used AI for housing insights, and that consumers rate AI and agents nearly evenly on which made them "smarter" about the market — though agents still edge out AI on perceived accuracy. The decisive number, though, is about trust. A December 2025 survey found that 65% of Americans trust AI to compare prices on major purchases like homes — but only 14% trust AI to act on their behalf. That gap has barely moved in two years even as the technology has rocketed forward. People are happy to let AI inform a half-million-dollar decision; they are not yet willing to let it make one. That's the strategic opening. The report is clear that the winners are "not the pure-AI platforms" betting on a consumer trust they don't yet have, nor "the incumbents betting nothing will change." The opportunity is "in the middle — products and advisory relationships that use AI to make human judgment better and more transparent." The agent who uses AI to do the Tier 1 grunt work instantly and for free, then spends their energy on Tier 2, is the agent the data points toward.
Seven Ways to Sell a Home in an AI World
One of the report's most practical contributions is reframing how many real options a seller actually has. Most people believe they have two choices: hire a full-service agent, or sell it themselves. Alloy Advisors built a free public tool, The Home Seller's Reality Check, that models what a seller would net under seven distinct paths, factoring in carrying costs and live market conditions. Until consumers can see all seven, the authors argue, "the commission standard isn't going anywhere." Here's the menu: - Traditional full-service — the classic listing agent at a full commission, every task bundled into one percentage. - Exclusive preview — quietly marketing to a targeted pool of buyers before a public launch. - Private, off-MLS sale — keeping the listing out of the public MLS entirely, often for privacy or testing price. - Discount broker — full or near-full service at a reduced commission. - Flat-fee MLS — paying a one-time fee to get listed on the MLS while handling the rest yourself. - Full FSBO (For Sale By Owner) — no agent on the listing side, the seller runs the entire process. - iBuyer cash offer — selling fast to an instant-buyer platform for certainty, usually at a price discount. The point isn't that everyone should abandon full service — for plenty of sellers it's still the right call. The point is that an AI-assisted consumer can now model all seven on their own kitchen table, and an agent who can't articulate why their service beats the other six is going to struggle to defend a 3% rate.
What This Means If You're Selling in Colorado or Florida
Markets are local, and this is where the national averages meet the ground. In the Colorado Foothills — Evergreen, Conifer, Golden, Morrison, and the rest of Jefferson County — the report's "hyperlocal tacit knowledge" advantage is enormous. Pricing a mountain property correctly means understanding well and septic systems, wildfire-mitigation and insurance realities, defensible-space requirements, seasonal access, and micro-markets that can swing in value within a single zip code. No national AI model carries that the way a broker who has walked these properties for years does. The same is true across Southwest Florida — Cape Coral, Fort Myers, Naples, and Marco Island in Lee and Collier Counties — where flood zones, hurricane-hardening, seawall and dock conditions, HOA and CDD structures, and the rhythm of a heavily seasonal, relocation-driven buyer pool reward on-the-ground judgment. An AI can pull the comps. It can't tell you which canal lot is worth the premium or how to position a home to an out-of-state cash buyer relocating in January. This is also where the Colorado-to-Florida relocation lane becomes a genuine differentiator. A seller cashing out of an appreciated foothills property to buy in Southwest Florida — or vice versa — is running two transactions in two very different markets at once. That's a coordination, negotiation, and emotional-support challenge sitting squarely in Tier 2, exactly where human expertise still commands a premium. AI can handle the paperwork on both ends. Steering the whole move is human work.
What This Means If You're an Agent or Broker
The report is explicit that this is not a knock on the industry and not a scare tactic. Cofano framed it as puncturing a "fantasy" — the belief that AI somehow won't put downward pressure on compensation. Even if the analysis is only "marginally accurate," he argued, the implications are large enough that agents and brokers should already be "modifying their businesses to be sustainable." A few honest takeaways: - Stop selling Tier 1 as if it's valuable. Clients can now get the CMA, the listing copy, and the coordination for nearly free. Charging a premium for commodity work is the position most exposed to compression. - Build your brand on Tier 2. Make your negotiation track record, hyperlocal expertise, and client-care reputation the visible, provable center of your pitch — because that's what the data says holds its value. - Use AI aggressively yourself. The agent who runs Tier 1 through AI in minutes frees up time for the high-value human work and out-competes the agent still doing it by hand. - Get comfortable explaining the seven paths. A seller who feels you've shown them every option — rather than defaulting them into full service — is a seller who trusts your recommendation when you make one. - Volume matters. A few deals a year is no longer enough to out-skill the machine. The report's authors are clear that depth of experience is what separates the agents who win from the ones AI replaces. The reassuring part, repeated by both authors, is simple: "the great agent will win." As Kulkarni put it, great agents win because they put the consumer at the heart of the business and do what's genuinely right for the person paying them. For brokers, that means many models will need to pivot to put the consumer at the center and help agents serve a far better-informed client.
The Bottom Line for Sellers and Agents
It's worth being precise about what this report does and doesn't claim. The authors are candid that their service valuations are "constructed bottom-up estimates from freelance and professional market rates," not hard survey data, and that no controlled studies yet compare the final sale price of AI-assisted versus traditionally-assisted transactions. This is a well-reasoned argument, not a closed case. But the direction of travel is hard to argue with. Information that agents once monopolized is now in every consumer's pocket. The commoditized half of the job has collapsed in value, the human half hasn't, and the flat percentage that charges identically for both is the thing under pressure. The likely future isn't a world without agents — it's a world where what you pay is finally tied to the skill and the outcome you actually receive. For sellers, the move is to understand all your options, use AI to inform the decision, and hire human expertise where it genuinely moves the result — pricing, negotiation, and navigating a complex local market. For agents, the move is to become the kind of professional the data still rewards. Either way, the comfortable assumption that selling a home simply costs 6% is, by this report's clock, on a five-to-seven-year timer.
