Is the 'one-person company' actually the ideal model?

On Alibaba.com — the world’s largest B2B marketplace — Chairman Kuo Zhang just announced that 30-40% of their customers are now “independent entrepreneurs.” No employees. Just one person managing things, while AI agents handle the rest: cross-platform product listings, social management, customer support, tax filings.

Suzhou has committed to training 10,000 OPC operators by 2028, with an investment of 700 million yuan. Chengdu offers 20,000 yuan subsidies to graduates who start a one-person company. Hangzhou spends 100 million yuan a year on its OPC community.

The entire Chinese administrative apparatus is pushing this model — not because AI is exciting, but because 1 in 6 young Chinese workers is unemployed, and 12.7 million students will graduate in 2026.

An entire generation of young Chinese is betting their future on AI, with government money behind them. And the story is spreading fast.

In early April, the New York Times published a viral profile. A Los Angeles entrepreneur named Matthew Gallagher had built a one-person startup called Medvi. The company sells GLP-1 (Ozempic, Wegovy) via AI, pulled in 401 million USD in 2025 revenue, and is projecting 1.8 billion USD in 2026.

Sam Altman — CEO of OpenAI — had long predicted an era of “billion-dollar one-person companies” powered by AI leverage. Medvi’s rise became a concrete proof point for that thesis.

The “one-person company” powered by AI is becoming a real wave. Not hype. Not the FOMO of a few KOLs.

Over the past few weeks, I’ve watched plenty of founders sit with the question: “am I doing this wrong?”.

Three things I think Altman gets right

Before I push back, there are three things I want to acknowledge.

First, someone working alone really does make decisions faster than any team. In the early years of GemPages, I worked entirely alone. Code, design, support, marketing. No meetings. No alignment to reach. No one waiting on anyone. Insanely fast.

Second, the cost math has genuinely changed. An AI tool stack for one person costs around 3,000–12,000 USD per year. An old-school team of 10 can chew through 1.2 million USD. That’s a 95-98% reduction. Real number.

Third, there will be billion-dollar one-person companies in the future. I have no reason to argue with Altman about technical feasibility.

My question lies elsewhere — can this business model actually survive? Especially in DTC ecom.

The label is doing a swap

Dig deeper into the “one-person company” stories, and they turn out to be very different cases.

Case 1 — Someone working solo. Creator, freelancer, using AI to scale personal output. Real. But small. Tens to hundreds of thousands USD in annual revenue. Wang Tianyi, 26, in Shanghai, makes 40,000 yuan a month producing AI ad videos — that’s Case 1. Nobody calls this a “billion-dollar company.”

Case 2 — A lean team of 2-10 + AI as leverage. This is what most viral pieces are actually labeling as “one person.” Medvi: 2 employees + brother + contractors + outsourced firms. Cursor: under 50 people, 500 million USD in annual revenue. Gumloop: currently 2 people, scaling toward 10 to chase 1 billion. According to Karen Dai — author of the book “One Person Company” in China — the OPC concept in China actually includes groups of 2-3. They call it “one person” because it sells better.

Case 3 — One real person + AI agents replacing every staffing function. Very rare. So far, no verified billion-dollar company belongs to this category.

Journalists call all three cases by the same name. They’re using the story to sell, not to describe accurately.

What’s actually working is Case 2 — a lean team + AI as leverage. That’s the trend worth learning. But “1 person making 1.8 billion” goes viral more than “5 people making 50 million,” so the narrative gets pushed into Case 3 — a category that doesn’t exist yet.

There’s something I want to underscore: when a story is too good to be true, it usually isn’t true. When the story runs faster than verification, DTC founders in FOMO mode are very easy to lead astray with wishful thinking.

If Medvi is the only flagship example we have of a “billion-dollar one-person company,” and you know that the company is mired in FDA trouble plus complaints from 100,000 plaintiffs, you might want to reconsider.

Is it worth betting on this model, or on another less viral one?

A one-person company for DTC?

In his own prediction, the Anthropic CEO listed the fields most likely to produce billion-dollar one-person companies: automated financial trading, developer tools, automated customer service.

DTC ecommerce isn’t on the list.

That’s because DTC has 4 factors that break the one-person model.

The first is supply chain. Cargo stuck at Long Beach, a supplier raising prices 12%, a SKU recall. These are edge decisions that require human judgment, not pattern matching. Tom Coshow at Gartner has been blunt: with LLM-based AI agents, you have to give them very simple decisions to get reliable answers. Automating the VP of Sales? Still far off.

The second is ad budget. A DTC brand doing 100,000 USD a month burns over half of it on ads. 1% conversion = profit. 0.5% conversion = bankruptcy. Do you dare hand an AI agent full authority over decisions at the edge of life and death? I’ve been building AI agents for CRO for 3 years and I don’t.

The third is complex customer service. AI handles tier 1 well. Complaints, returns, disputes, fraud, supplier issues — still require humans. And customer service is the actual defensive moat of a DTC brand. Tony Hsieh has a line I quote constantly:

“Customer service shouldn’t be a department. It should be the entire company.”

Handing customer service entirely to AI agents is handing over your biggest defensive moat.

The fourth is brand and community. 18,000 paying brands are with GemPages, most for 5-8 years — part of the story I told in 10 years building Shopify apps from Saigon. They didn’t stay because we were the best tool. They stayed because there are relationships, a community, people on our team they know by name. Customers loving a brand doesn’t scale through an API call.

One person + AI can build a high-revenue SaaS tool. Build a DTC brand that survives 10 years with the same formula? I’ve never heard of one.

What’s actually working?

I think the real trend here isn’t “one-person company” — it’s the lean team — 3-10 people + AI as leverage.

John Jackson, founder of Hitprobe, put it well: instead of a one-person unicorn, what’s more common will be teams of 3-5 scaling to the size that used to require 50+ people.

Why does “lean” beat “alone”?

Because solo + AI is essentially you talking to yourself in an echo chamber. AI doesn’t push back — it does what you tell it. Meanwhile, 5 people arguing over a decision is 5 chances to catch a blind spot.

Because you’ll have moments when you’re exhausted, sick, or lose motivation for two weeks. Alone, the company stops; in a team of 5, it keeps running.

Because customer trust is built across many people, not just the founder. After enough years, customers trust Gemians, not just me or any one name.

And because in the long run, decision quality matters more than speed. Solo is fast, but it also means a higher rate of being wrong.

There is one exception I have to acknowledge. A few exceptional individuals have honed processes + core tech over many years. Pieter Levels (Nomad List), Marc Lou (multi-product indie) — they succeed alone not because they have an AI doing it all. They succeed because they already had the processes before AI got good. AI is just leverage on what was already there.

This is a special case. Not a model to copy. If you don’t have 10 years of process honing, you’re in the majority — and the majority wins with lean teams.

Before you FOMO

It’s not that you don’t need AI.

It’s not that you need 50 more people.

You need exactly 3-5 people, using AI as 10x leverage — with processes good enough to keep momentum when the fall comes.

— Chris