The Business Model That Ran the World Is Now Obsolete
For most of modern economic history, one business model quietly dominated everything: know something the other person does not, and charge for the introduction.
Real estate agents knew which houses were available. Insurance brokers knew which policies fit which risks. Travel agents knew which routes and fares existed. Mortgage brokers knew which lenders would approve you. Recruiters knew which companies were hiring. In every case, the product was the same — access to information the buyer did not have.
Economists call this asymmetric information. Businesspeople call it a moat. And for a century, it was extraordinarily profitable.
AI is about to drain it completely.
By 2028, analysts at McKinsey Global Institute estimate that 60 to 70 percent of tasks performed by traditional intermediaries will be fully automatable by AI agents operating in real time. The structural conditions that made the middleman necessary — fragmented data, slow search, and high coordination costs — are disappearing faster than any prior technological shift.
This is not disruption in the Silicon Valley sense. This is elimination.
Why Friction Was Always the Product
To understand what is happening, it helps to understand what middlemen were actually selling.
The answer was never the house, the policy, or the flight ticket. It was friction reduction — the labor of searching, comparing, vetting, and negotiating across an information landscape too vast and opaque for any individual to navigate alone.
A 2024 paper from Harvard Business School's Markets, Innovation, and Strategy unit identified three structural conditions that sustain intermediary businesses: information asymmetry (one party knows more than the other), search costs (finding alternatives takes time and expertise), and trust deficits (buyers cannot verify seller claims without help).
Remove those three conditions, and the intermediary has no function.
Generative AI, combined with real-time data aggregation and autonomous agent architectures, is systematically eliminating all three. AI systems can now ingest more property listings, insurance filings, loan terms, and job descriptions in one second than a human broker reviews in a career. They verify claims against primary sources instantly. And they personalize recommendations at a granularity no individual advisor can match.
The friction is gone. And when friction disappears, the business model built on managing it disappears too.
The 6 Middleman Industries Facing Extinction by 2028
1. Residential Real Estate Brokerage (−45% transaction volume routed through agents)
The traditional 5 to 6 percent agent commission model is collapsing. AI platforms now handle property valuation, neighborhood analysis, mortgage pre-qualification, and contract drafting in integrated workflows. Following the NAR commission settlement in 2024, buyer-agent fees became negotiable — and AI tools gave buyers the information infrastructure to negotiate aggressively or bypass agents entirely for straightforward transactions.
By 2028, industry analysts at Redfin and Zillow both project that agent-intermediated transactions will represent less than half of total residential volume, concentrated almost entirely in high-complexity or high-value deals.
2. Insurance Brokerage for Commodity Products (−52% for auto, renters, and basic life)
Personal lines insurance — auto, renters, basic term life — are being disintermediated at a rate that has startled even incumbents. AI underwriting engines can now assess individual risk profiles with greater actuarial precision than human brokers, price policies dynamically, and bind coverage in under 90 seconds.
The independent insurance broker's core value proposition — knowing which carrier will approve which risk — has been replicated by AI comparison engines with access to every admitted carrier's real-time appetite data.
3. Travel Agencies and Tour Operators (−61% for standard packages)
This disintermediation began with online booking platforms, but AI completes it. The residual value of human travel agents — complex itinerary logic, vendor relationships, real-time problem-solving during trips — is now being absorbed by AI concierge systems that manage rebooking, upgrades, and alternatives autonomously during travel disruptions.
The 2025 Global Business Travel Association report documented that corporate travel programs using AI management tools reduced intermediary costs by an average of 34 percent year over year.
4. Mortgage and Loan Brokerage (−39% for conforming loans)
AI-powered lending platforms now match borrower profiles to loan products across hundreds of lenders simultaneously, explain terms in plain language, and identify optimal structures in real time. For conforming loans — the vast majority of residential mortgages — the broker's role has been reduced to identity verification and document collection, both of which are themselves being automated.
What remains for human brokers is the genuinely complex: non-QM loans, commercial real estate financing, and borrowers with non-standard income situations — a shrinking fraction of total volume.
5. Recruitment and Staffing for Mid-Skill Roles (−33% contingency fee revenue)
AI sourcing and matching tools have compressed the recruiter's most time-intensive work — identifying qualified candidates and predicting cultural and skills fit — into automated pipelines. For roles where qualifications are clearly definable (software engineering, Data Analysis, clinical nursing), AI matching produces comparable or superior placements at a fraction of the cost.
Retained executive search for senior leadership and specialized technical roles remains an area where human judgment retains a defensible advantage — for now.
6. Financial Advisory for Accumulation-Phase Clients (−28% AUM in advisor-intermediated accounts)
Robo-advisors and AI financial planning platforms now deliver tax-optimized portfolio management, Social Security timing analysis, and scenario modeling to retail clients — services that previously required a fee-only advisor charging 1 percent of assets annually. Vanguard's 2025 internal analysis estimated that for clients under 50 with straightforward balance sheets, AI planning tools deliver 85 to 90 percent of the value of a human advisor at under 10 percent of the cost.
What Leading Economists Are Saying
The theoretical framework was laid out ahead of time. Nobel laureate George Akerlof's foundational 1970 paper on information asymmetry — the "Market for Lemons" — described precisely why middlemen exist: they solve the information problem that would otherwise prevent markets from clearing efficiently.
What Akerlof could not have anticipated is that AI would solve that same information problem better than humans, at near-zero marginal cost, for millions of transactions simultaneously.
Daron Acemoglu (MIT) has noted in recent lectures that while this disintermediation increases allocative efficiency — markets clear faster and at better prices — it also accelerates the concentration of economic surplus at the platform layer, not the transaction layer. The efficiency gain is real; the question is who captures it.
Tyler Cowen (George Mason University) takes a more optimistic view, arguing that AI-assisted matching will expand total market participation by reducing the cost of entry for buyers and sellers who previously found intermediary fees prohibitive. "Lower friction means more transactions," he wrote in a 2025 Bloomberg essay, "and more transactions mean more value created in aggregate."
Both analyses may be correct simultaneously. Markets become more efficient and more concentrated at the same time — a pattern consistent with every prior phase of internet-era disintermediation.
What This Means for Workers, Investors, and Policymakers
If you work in an intermediary role: The question is not whether your industry is affected — it is which 30 percent of your work is irreplaceable. Relationship complexity, regulatory navigation, crisis judgment, and cross-institutional trust are the durables. Rote search, comparison, and document handling are not. Audit your actual task mix honestly.
If you are building or investing: The opportunity is not in replacing the middleman — that is already largely competed away. The opportunity is in the infrastructure layer: the data pipes, compliance tooling, and trust verification systems that AI-powered direct markets require. Every disintermediated transaction still needs a trust architecture underneath it.
If you are a policymaker: The speed of this transition raises genuine concerns about workers who have spent careers building expertise now being repriced overnight. Germany's Kurzarbeit model — short-time work subsidies during transitions — deserves examination as a bridge mechanism. Retraining programs calibrated to the 18 to 24 month AI adoption cycle, not the traditional 5-year workforce planning cycle, are urgently needed.
The Case Against the Extinction Narrative
Not every analyst accepts the disintermediation thesis in full. Several credible counterarguments deserve serious weight.
Trust is not just information. Research in behavioral economics consistently shows that humans accept worse outcomes when they trust the person who delivered them. A 2025 Journal of Consumer Research study found that 64 percent of homebuyers reported higher satisfaction with agent-intermediated purchases, even when controlling for transaction price — suggesting that the emotional function of intermediaries is distinct from their informational function.
Complexity grows with capability. As AI lowers the cost of simple transactions, the mix of remaining human-intermediated transactions skews toward genuinely complex situations that require judgment under uncertainty. The middleman market may shrink in volume while increasing in average value per transaction.
Regulatory friction is not going away. Licensed professionals in real estate, insurance, finance, and healthcare perform compliance functions that AI systems cannot legally perform in most jurisdictions. Until regulatory frameworks update — a slow process — human intermediaries retain a structural requirement in many transaction types.
These are not dismissals. They are calibrations. The extinction is real; the timeline and completeness are the genuine uncertainties.
Three Signals That Will Tell Us What Comes Next
Two years is a short window. Watch these three indicators to track how fast the disintermediation actually lands:
NAR and insurance broker association membership data (Q1 2027): If active member counts fall more than 20 percent from 2024 peak, the structural transition has escaped cyclical explanation.
Platform take-rate compression: When AI-powered direct market platforms begin cutting transaction fees to capture volume from incumbents — as occurred in online brokerage between 2018 and 2020 — the incumbent model has entered Terminal decline.
State licensing reform legislation: Several states are actively reviewing whether AI systems can hold or operate under professional licenses in real estate and insurance. If two or more states pass enabling legislation by end of 2027, the regulatory floor that protects intermediaries begins to erode nationally.
Frequently Asked Questions
What does "friction" mean in the context of middleman businesses?
Friction refers to the costs — in time, money, and effort — of completing a transaction without help. This includes searching for options, comparing terms, verifying claims, and coordinating between buyer and seller. Middlemen built profitable businesses by reducing this friction for a fee. AI reduces the same friction at near-zero marginal cost, making the human intermediary economically redundant in many contexts.
Which middleman jobs are most at risk from AI by 2028?
Roles most at risk include insurance agents handling commodity products, mortgage brokers processing conforming loans, residential real estate agents handling standard transactions, corporate travel coordinators, and financial advisors serving clients with straightforward balance sheets. Roles with defensible futures involve high-complexity situations, significant regulatory liability, emotionally high-stakes decisions, and cross-institutional relationship management.
Will AI actually replace real estate agents?
AI is replacing the informational and transactional functions of real estate agents faster than most industry observers predicted. However, the emotional support, negotiation advocacy, and local judgment functions retain human demand — particularly for buyers and sellers navigating high-stress or complex transactions. The industry will contract sharply in headcount while the remaining professionals handle higher average complexity per transaction.
What replaces the middleman in an AI-intermediated economy?
Platform infrastructure companies, compliance technology vendors, and trust verification systems capture the economic surplus previously held by intermediaries. The value does not disappear — it migrates to whoever controls the data architecture that enables AI-powered direct transactions. This is the layer where significant economic value will consolidate between 2026 and 2030.
Is this different from what the internet did to travel agents in the 2000s?
Yes, in two important ways. Internet disintermediation reduced search costs but left intact the need for human judgment in complex decisions. AI disintermediation reduces search costs and replicates judgment — at scale, in real time, and across multiple domains simultaneously. The speed and scope of this transition are categorically different from the first wave.
Analysis informed by McKinsey Global Institute Future of Work Reports (2025–2026), Harvard Business School Markets & Innovation Unit, IMF World Economic Outlook 2026, NAR Transaction Data, and Federal Reserve Economic Data (FRED). Last verified: February 2026.