$1.8 billion in projected 2026 revenue. Two employees. A net margin of 16.2 percent.
That is Medvi, a telehealth company founded by Matthew Gallagher with $20,000 in startup capital and a dozen AI tools. In its first year it generated $401 million. It is now on track to nearly quintuple that figure with the same team of two.
Its nearest competitor, the publicly traded Hims & Hers, runs essentially the same playbook: online healthcare and mail-order prescriptions. It employs 2,442 people and operates at a 5.5 percent margin. Same industry, same business model, roughly three orders of magnitude apart on headcount.
This is not a Silicon Valley curiosity. It is a structural shift in how a service business can be built, and it matters far beyond telehealth.
The Orchestration Layer That Replaces a Company
The first instinct when hearing numbers like these is skepticism. We will get there. But the architecture is worth understanding first, because it is replicable and largely unpatented.
Gallagher did not build a technology company. He built a thin operational core that coordinates commodity AI tools and outsourced specialists. ChatGPT, Claude, and Grok handle copywriting, customer communications, and internal code. Midjourney and Runway produce the brand assets and short-form video. ElevenLabs generates the voice. Everything that touches regulation, meaning pharmacy operations, compliance, and prescriptions, is outsourced to licensed specialists who appear on someone else's payroll.
That split is the real design choice. Gallagher did not try to automate the hard parts. He automated the scalable parts and rented expertise for the regulated ones. AI handles volume. Humans handle judgment. The orchestration layer in the middle is small enough to fit in two heads.
Anyone with a credit card can subscribe to the same tools. The moat is not the stack. The moat is knowing which tasks should be fully automated, which need a human checkpoint, and which should never be touched by a model at all. Most companies that stumble on AI point it at judgment-heavy decisions and then wonder why the output feels mediocre. Medvi points it at repetitive, high-volume execution and keeps licensed humans accountable for anything a regulator might ask about later.
Headcount Is the Hidden Margin Lever
The most revealing number in this story is not the revenue. It is the margin.
At 16.2 percent, Medvi is nearly three times more profitable per dollar of revenue than Hims & Hers at 5.5 percent. The reason is not exotic. In most service businesses, headcount is the largest cost line, and payroll compounds with benefits, management layers, real estate, and tooling. Taskade's analysis of one-person companies estimates that AI agents now absorb 80 to 85 percent of execution tasks at roughly 5 percent of traditional team cost. Shrink the biggest line item by an order of magnitude and margins expand even if the top line sits still.
This math is not telehealth-specific. It applies wherever a business processes high volumes of standardised requests, produces large quantities of content or documentation, coordinates specialists who do not need to be full-time, or runs rule-based workflows in scheduling, logistics, or claims. Food distribution fits. So does construction project management, energy dispatch, maritime bunkering, and most of mid-market insurance. The list is embarrassingly long.
The implication for established companies is not that they should fire everyone. It is that they should reexamine which roles are execution-heavy and which are judgment-heavy, because the cost of execution just dropped by roughly 95 percent and the competition will eventually notice.
China Has Turned This Into Industrial Policy
If this architecture is real, you would expect someone to build policy around it. That someone turns out to be China.
Suzhou, Shanghai, Wuhan, Shenzhen, and Hefei are running structured incentive packages for solo AI-augmented founders. Subsidies reach $1.4 million per registered one-person company. Qualifying founders receive free housing, free office space, and free compute credits for AI workloads. Suzhou alone plans 1,000 OPC registrations by 2028. Alibaba's B2B president, Kuo Zhang, has said the quiet part out loud: "The One-Person Unicorn is Coming." Thirty to forty percent of Alibaba's customers, he notes, already operate as solo entrepreneurs leaning on AI agents for daily operations.
When a superpower builds policy infrastructure around an idea, it stops being theoretical.
China is not subsidising solo founders because it is fashionable. It is doing so because the unit economics are demonstrably viable and the competitive implications for the next decade are enormous. The contrast with Europe is uncomfortable. Many Mittelstand companies, some of them global leaders in their niches, are still debating whether to run a first AI pilot. Meanwhile, a solo operator in Suzhou with free GPUs and state-backed capital is quietly building a competitor.
Where the Numbers Get Softer
Intellectual honesty demands the counter-case. Gary Marcus, the cognitive scientist and one of the more disciplined AI skeptics, has pushed hard on the Medvi narrative and his points deserve weight.
Medvi is private. The $1.8 billion projection is self-reported, and there are no SEC filings or independent audits in the public record. Telehealth has a short memory of companies that scaled fast and collapsed faster once regulators paid attention. The "two employees" framing also hides the actual labour: the pharmacists, compliance officers, and clinicians doing the regulated work exist, they are simply on another company's payroll. And the business operates one enforcement action away from a different economic model entirely.
The ethical questions deserve naming too. Reporting in The New York Times and PYMNTS has raised allegations around aggressive advertising practices and fabricated doctor profiles on the Medvi platform, and those are serious if they hold up. Even if Medvi itself turns out to have cut corners, the structural point survives: the architecture works, and someone will build the cleaner version of it.
These are fair objections. No one building a strategy on the Medvi story should treat the numbers as gospel. But a company with $900 million in revenue, a couple of founders, and a contractor network is still a radical departure from how service businesses were built ten years ago. The precise figure matters less than the shape of the thing.
The Floor Just Rose for Small Teams
The trend reaches well beyond startups. Across small businesses globally, 98 percent now use AI tools daily, 91 percent attribute recent growth directly to AI adoption, and 87 percent report measurable operational improvements. The capability gap between a Fortune 500 firm and a well-run ten-person team is closing fast, not because large companies are getting worse, but because the floor of what a small team can do has risen sharply.
For established mid-market companies, that cuts both ways. The next competitor may not be another firm with hundreds of employees. It may be three people with an AI stack, deep domain knowledge, and a willingness to accept margins the incumbent considers insulting. At the same time, incumbents already hold what solo operators spend years trying to buy: domain expertise, customer relationships, regulatory fluency, and brand trust. Layer AI-augmented operations on top of that foundation and the outcome is not a one-person company. It is a thirty-year firm that suddenly operates like one.
The practical move is to draw one honest line through the business and ask, role by role, which work is execution a model can now carry and which work is judgment a customer is actually paying a human for. Firms that get that line right will end up with the margin structure of a Medvi and the domain depth of a seasoned industry veteran. Firms that get it wrong will either automate the wrong things or automate nothing, and both paths lead to the same place.
Two Employees, Read as a Signal
The interesting question is not whether Medvi hits its projection. Private-company numbers rarely survive contact with an auditor intact, and Gary Marcus is probably right to discount them. The interesting question is what the world looks like when this architecture becomes ordinary rather than newsworthy.
In that world, the size of a company stops being a reliable proxy for its capacity. Two people with an orchestration layer and a network of licensed specialists can credibly take on a category that used to demand a few thousand employees. Industrial policy in Suzhou is betting on it. Alibaba's merchants are already living in it. The only question left is which incumbents notice early enough to reshape their own organisations, and which ones keep treating "two employees, $1.8 billion" as a headline they do not quite believe.
Sources: PYMNTS, "The One-Person Billion-Dollar Company Is Here" (April 2026); The New York Times, coverage of Medvi advertising and provider-profile allegations (2026); The Rundown AI, "AI Just Made the Billion-Dollar Solo Founder Real" (April 2026); Fortune, "Alibaba.com President on One-Person Unicorn" (March 2026); Rest of World, "China Mobilizes One-Person Company AI Startups" (April 2026); Taskade, "One-Person Companies: The Future of Work With AI" (April 2026); Inside Small Business, "How Small Teams Can Actually Win With AI in 2026" (April 2026); Gary Marcus, "The Back Story Behind the First $1.8B" (April 2026); Asrify, "AI Automation Playbook: Small Teams Outperforming Enterprises" (2026).