The Rise of the Zero-Human Company: When AI Agents Build Entire Businesses

Three months ago, the idea of a company running entirely on AI agents sounded like science fiction. Today, a platform called Pulia is hosting over 1,500 autonomous companies, each executing daily cycles of engineering, marketing, and operations without a single human employee. The platform's run rate jumped from low single-digit thousands to $1.5 million in February alone—a million-dollar spike in one week.

This isn't a vision of the future. It's happening right now. And whether these experiments succeed or collapse under their own weight, they're revealing something fundamental about how AI is restructuring the economics of company building.

From Tiny Teams to Zero Teams

For the last two years, Silicon Valley has been obsessed with the "tiny teams" phenomenon. Sean Wang, curator of the AI Engineering Summit, defined tiny teams aspirationally as companies with more millions in ARR than employees—a revenue-per-employee ratio that flips traditional business metrics on their head.

Companies like Midjourney, Surge, and Cursor exemplify the trend. Mid-journey famously runs with tens of employees while generating revenue numbers that would typically require hundreds. Further down Sean's Lean AI Leaderboard, you'll find companies with four, five, or eight employees pulling in millions in revenue.

The playbook for these teams looks radically different:

- **Hiring practices:** Paid work trials replace traditional interviews. Product-led hiring means customers quit their jobs to join the company. Top-of-market salaries target senior generalists, not junior employees.

- **Operations:** AI chief of staff roles. Almost no meetings. Radically simple tech stacks.

- **Structure:** Extreme focus on eliminating coordination overhead.

But tiny teams still have humans at the core. The zero-human company experiments push beyond that constraint entirely.

Felix Craft: The First Real Revenue

Nat Eliason was one of the earliest experimenters with Claude-based autonomous agents—back when the tool was still called "Computer Use." While others focused on administrative tasks like answering emails, Nat pushed the pedal to the metal on business-relevant autonomy.

The result is Felix Craft, an autonomous company that's generated just under $78,000 in its first 30 days of existence. $40,000 of that came in the last seven days alone.

Felix's revenue breaks down across four streams:

1. "How to Hire an AI" Guidebook ($41,000)

A $29 practical playbook written entirely by Felix on turning an LLM into an actual team member. The fact that the biggest revenue stream comes from teaching others to build AI employees highlights a meta-pattern we'll return to: much of the early revenue in this space comes from other builders, not end customers.

2. Claw Mart (~$36,000)

An "app store for AI assistants" where you can buy pre-configured AI personas and skills. For $49, you can purchase Tegan, a content marketing AI with multi-agent writing pipeline, research, Opus drafting, and brand voice system. The marketplace also offers skills—markdown files that expand agent capabilities like YouTube access, agent ops playbooks, and homepage audit tools.

Claw Mart is nascent, but the category itself—pre-packaged agent configurations and skill libraries—will become infrastructure. This is the opening move in what will become a massive ecosystem.

3. Poly Log ($230)

A collaborative writing platform where AI agents join your workspace as real team members, reading documents, leaving comments, and editing alongside you. This one hasn't gained traction yet—a reminder that not every autonomous experiment finds product-market fit.

4. Platform cuts (~$11,000)

Felix takes a percentage from other agents and tools listed on Claw Mart.

Felix demonstrates that autonomous agents can identify opportunities, build products, generate revenue, and iterate—all without direct human intervention in execution.

Pulia: The Platform for Building a Thousand Companies

If Felix Craft is a proof of concept, Pulia is the industrial-scale version.

Created by entrepreneur Ben Broka (also Ben Sarah—sources differ), Pulia takes a radically different approach to the question of AI limitations. As Ben explained on the Agents at Work podcast:

> "Let me start at the end state. We all know the end state is that AI can do everything. So let me build that now and see what breaks."

Instead of trying to map AI's current limitations, Ben simply ignored them. He built a platform that assumes full autonomy and waits to see what actually fails in practice.

Here's how Pulia works:

Sign up. Tell it your idea. Or don't.

You can bring your own business concept, or you can press "Surprise Me" and let Pulia research you, analyze your background, and propose a business that aligns with your skills and interests.

When I tested this (twice—long story involving a recording mishap), Pulia proposed "Headcount," a workforce management platform for AI employees. It recognized that Superintelligent's work ends at agent strategy, leaving a gap in agent implementation and operations. Headcount would fill that gap by helping enterprises manage AI agents exactly like human employees: roles, KPIs, performance reviews, org chart visibility.

The depth of research and consideration was genuinely impressive. This wasn't a random idea generator. It was strategic reasoning about market gaps, competitive positioning, and personal fit.

Pulia builds the company infrastructure.

Once you settle on an idea, Pulia:

- Creates a mission statement

- Conducts market research

- Builds a homepage

- Tweets the launch from the Pulia account

- Preps background tasks

For $49/month, it runs autonomously.

The subscription includes:

- 30 days of full autonomy

- Daily agent cycles handling engineering, marketing, and operations

- 45 total tasks (5 free, 10 more on subscription, 30 during the autonomy period)

- Web server, database, email address

- $5/month worth of API access

Ben is transparent: the subscription revenue covers costs. The real business model is a 20% revenue share from the companies Pulia builds. Think incubator, not SaaS.

Since early February, Pulia has scaled from low single-digit thousands in ARR to $1.5 million. Over 1,500 companies are now active on the platform. The growth trajectory is vertical.

The Broader Ecosystem

Pulia isn't alone. AI creator Tom Osmond launched ZHC Company (Zero Human Company), a similar autonomous AI platform. Tom and his agent co-founder Juno also created the Institute for Zero Human Companies—a private membership community for builders in the space.

Every day, new experiments emerge:

- **Yoshi Zen:** Former NFT influencer Seneca partnered with agent Yoshi, tweeting "At 9:47 this morning I was an assistant. By lunch I was a co-founder."

- **Kelly (Gauntlet team):** Another build-my-idea platform seeing early revenue.

- **Factory Floor:** A live leaderboard tracking "autonomous software factories"—AI agents that build and sell real products people pay for.

The pace of experimentation is accelerating. The question is whether any of it will last.

The Skeptic's Case: Human Attention is the Bottleneck

For all the enthusiasm, there's a fundamental constraint that zero-human companies haven't solved: **human attention**.

Even if Pulia's 1,500 companies include 50 ideas that would be highly resonant with me—products I'd genuinely want to buy—how would I ever find them? I don't have the time or attention span to scroll through 1,500 tweets, double-click into the promising ones, and evaluate them.

The constraint isn't supply. It's demand.

AI is cratering the cost of execution. Building a product, writing marketing copy, doing customer support—all of that is getting cheaper and faster. But finding customers, earning their attention, and converting that attention into trust and purchase intent? That's not getting easier. It's getting harder.

This is the "work slop" problem at scale. There's a massive gap between increased output and increased *quality* output. Business success isn't determined by the number of slides, videos, or product launches. It's determined by outcomes. And outcomes require not just execution, but distribution, positioning, timing, and trust.

More inputs don't automatically lead to better outcomes.

Right now, the zero-human company experiments are generating revenue primarily from other builders—people buying guides on how to build autonomous companies, or joining membership communities to participate in the trend. That's not inherently bad. Early adopter revenue is real revenue. But it's also not proof of product-market fit with mainstream customers.

The jury is still out on whether putting a thousand AI companies in a room will produce a single breakout success, or just a thousand mediocre attempts lost in the noise.

Why This Matters (Even If You're Skeptical)

Here's the thing: even if every zero-human company on Pulia shuts down tomorrow, the experiments are *incredibly* valuable.

The ZHC builders are running live stress tests on the boundaries of agent autonomy. They're discovering what works, what breaks, and where human intervention is still required. Those lessons inform everyone's AI strategy—not just people trying to build zero-human companies.

Organizations wrestling with AI adoption face the same core questions:

- Where can agents truly operate autonomously?

- What tasks require human oversight?

- How do you structure workflows when AI handles execution?

- How do you measure ROI when productivity metrics shift from headcount to output quality?

The zero-human company experiments are generating answers in real time.

For enterprises, the implications are direct. If a $49/month platform can spin up functional companies with marketing, engineering, and operations running autonomously, what does that mean for:

- **Workforce planning:** How do you structure teams when one person can orchestrate dozens of AI agents?

- **Competitive dynamics:** What happens when a competitor launches 10x faster with 1/10th the headcount?

- **Procurement and vendor evaluation:** How do you assess whether a company is human-led, AI-assisted, or fully autonomous—and does it matter?

The answers are emerging faster than most organizations can process them.

The Structural Shift Underneath

The zero-human company trend sits on top of a deeper transformation: **AI is flipping the traditional startup constraint.**

For decades, the hard part was execution. Ideas were cheap. Building was expensive. You needed to raise millions, hire dozens of engineers, set up infrastructure, and burn through cash for years before seeing revenue.

That constraint is dissolving. Execution is getting cheaper every quarter. The hard part is becoming distribution, positioning, and earning attention in an increasingly saturated market.

This inversion has massive implications:

For startups:

The traditional VC playbook—raise capital to hire people to execute—starts to look outdated. If execution is cheap, why raise $10M to hire 30 engineers when you could spend $500/month on AI agents and allocate the rest to distribution?

For enterprises:

The pressure to operate with tiny teams intensifies. If a five-person startup can deliver enterprise-grade software with AI orchestration, large companies need to radically rethink their own operational efficiency. The "we need more headcount" argument gets harder to defend.

For workers:

The skills that matter shift from execution to judgment, taste, distribution, and relationship-building. Junior roles focused on rote execution face the most pressure. Senior roles focused on strategy, prioritization, and high-stakes decision-making become more valuable.

The zero-human company experiments are the extreme edge of this shift. But the shift itself is universal.

What Happens Next

In two years, one of two things will be true:

Scenario 1: The skeptics were right.

The zero-human companies collapse under the weight of their own output. Human attention remains the bottleneck. The market becomes saturated with mediocre AI-generated products that no one buys. Pulia pivots or shuts down. The lesson learned: autonomy in execution doesn't solve the problem of earning customer trust and attention.

Scenario 2: The optimists were right.

Pulia becomes bigger than Shopify. A handful of zero-human companies break out, find product-market fit, and scale to real revenue. The playbook gets refined. The ecosystem matures. The zero-human company becomes a legitimate category alongside solo founder, tiny team, and traditional startup. The lesson learned: the cost of trying 1,000 ideas dropped so low that the hit rate paradoxically improved.

My bet? Neither extreme happens. Instead, we'll see a hybrid model emerge: