Within a month of launching in February 2025, it had generated nearly $1.5 million in revenue from subscriptions. By June, Wix had acquired it for $80 million.

Last year, Dana Snyder, founder of a nonprofit consultancy called Positive Equation, used AI coding tools to build a software platform that works as an on-demand consultant for nonprofits. The platform guides organizations through building a monthly giving program step-by-step—generating fundraising strategies, donor communication plans, and program names tailored to each organization. 

With no technical background, Snyder built the platform using Replit’s AI coding tools over six months, with the aim of targeting the roughly 93% of U.S. nonprofits too small to afford a human consultant. Snyder said the platform allowed her to reach a far larger market than she imagined would be possible as a solo founder and offer her services at more affordable rates. Today she manages most of her clients through the platform and is still the company’s only full-time employee. 

Snyder and Shlomo are part of a growing wave of AI-enabled solopreneurs—founders building and scaling companies without the headcount they once needed. Data published in May last year by the U.S. Census Bureau counted 29.8 million non-employer companies generating around $1.7 trillion in revenue, roughly 6.8% of total GDP. It found that applications for new businesses are running at over 440,000 a month, more than 90% faster than pre-pandemic rates—though those figures date back to 2022. More recent estimates suggest the number of U.S. solopreneurs now likely exceeds 41 million.

The barriers to starting and running a business alone have been declining for years thanks to advances in cloud computing, e-commerce infrastructure, and freelance platforms. But the new crop of AI tools is now compressing the time, cost, and expertise required to build something significant, and in doing so, beginning to redraw the economics of entrepreneurship.

Chasing the solo unicorn

Tech companies have long championed the idea that AI would usher in a new era of solo entrepreneurs. Sam Altman, the CEO of OpenAI, said in 2024 that a tech CEO group chat he was in even had a running bet on when the first one-person billion-dollar company would appear. 

High-performing companies with limited headcount are not exactly new in Silicon Valley. Instagram had about 13 employees when Facebook agreed to acquire it for roughly $1 billion in cash and stock in 2012. WhatsApp had roughly 55 when Facebook struck its 2014 deal with the company, valued at up to $19 billion including restricted stock units. Mojang, maker of Minecraft, had around 40 employees when Microsoft bought it for $2.5 billion in the same year. 

The one-person version of this, some experts say, is just the logical endpoint of a trend that was already well underway.

“The average employment by companies less than a year old has been going down pretty steadily for 20 years. A one-year-old company was seven, eight, nine people 15 or 20 years ago. Even just two or three years ago it was down to three or four,” J.P. Eggers, a professor of entrepreneurship at NYU’s Stern School of Business, told Fortune.

Technology has played a role in that shift, but so has the rise of the fractional workforce. As it became easier for startups to hire outside contractors to handle specific functions—marketing, legal, design—the need for full-time staffers in those roles diminished. Now a lot of those gaps can be plugged by AI tools entirely.

Founders say they are using AI agents or coding tools to automate the workflows that would once have required dedicated hires, replacing both the labor of individuals and some of the expertise those roles carried.

Shlomo, for example, said he spent the first months of Base44 tracking exactly where his time went, then building automations to reclaim it. He created AI agents to sift through user feedback tickets and surface product ideas, to crawl his platform and flag UX issues, and to build and run quality-assurance tests—work that would typically be split across a product manager, a QA engineer, and a developer. He also built an application that monitored the code he was shipping and automatically turned it into marketing content: feature update posts, charts drawn from revenue data, published daily. 

“It took a while to fine-tune to generate content that sounds like me,” he said. “But once it worked, it was incredible.”

He also built a customer support bot, but he shut that down after two weeks. Going through support tickets himself turned out to be something he actually needed to do to stay close to what was happening in the product, he said.

Snyder is also using AI to automate work that would traditionally have required dedicated hires or outside contractors. One service she offers has nonprofits answer a series of questions about their mission and donor base. An AI agent loaded with Snyder’s own methodology then generates three curated name ideas for their giving program, each with an explanation of why it might resonate with the public—a task that would previously have taken hours of consultant time. She also uses AI to handle outreach before conferences, supplying a list of speaker LinkedIn URLs and a connection message but letting the agent handle the sending.

Snyder says automating these tasks frees up capacity for the work that actually requires a human. “If we can use AI for the manual, repeatable tasks,” she said, “we then have more brainpower to spend on ideating—which is the only thing that, as humans, we should really be spending our time on.”

The limits of one

How far the model can scale depends largely on the type of business. For consumer software products with limited supply chains and minimal regulatory exposure, some say a solo founder or very small team is genuinely viable. Industries with complex compliance requirements, physical supply chains, or enterprise sales relationships are harder, however, and tend to require the need for human oversight at too many points in the chain. 

AI is also considerably better at some tasks than others; coding is one area where the tools have advanced rapidly, which goes some way toward explaining why vibe coding founders like Shlomo have been among the earliest to demonstrate what the model can do at scale.

Even within software, a lack of domain expertise can become a liability. Last year, in an experiment at NYU Stern co-run with Microsoft, Eggers had his MBA students use AI agents to try to build startups from scratch. The AI was good at executing discrete tasks and accelerating brainstorming, he said, but it couldn’t substitute for the judgment that comes from having specialists in the room. “You’re kind of taking it on faith that what the AI is producing is pretty good,” he said. “No one really has the deep, specialized knowledge you need in lots of different areas.”

The economics of the model are also more complicated than they first appear. Monthly AI bills at lean startups can run into the hundreds of thousands of dollars, especially if the company is running on always-on agents. These costs can quickly become comparable to the headcount salaries they replace, Eggers said. However, compute costs scale more elastically than staff and don’t come with equity expectations, meaning founders who build this way tend to own considerably more of what they build.

But that concentration of ownership has broader implications. If more companies are being created with fewer people, it does not necessarily mean more will be successful. Experts say the market can only sustain so many winners, and as AI takes on work once distributed across larger teams, the wealth generated by successful startups could flow to an increasingly small number of people.

The cost of going alone

 It’s also important to take into account the day-to-day grind of running something solo.

In the early months of Base44, Shlomo had no one to watch the platform overnight, so he set an alarm every two to three hours to check that his servers were still running. It was only because of those alarms that, when the platform went down under a traffic spike one night, he managed to catch it within 10 minutes rather than six hours. 

And it was part of what eventually led him to sell. Shlomo said he recognized that building something truly global required expertise he didn’t have—specifically the consumer marketing capabilities that Wix had spent years developing. “I’m a product person,” he said. “But eventually, in order to actually scale this and make this a company that people might someday remember, I need help.”

Snyder is already building her next set of AI agents—a podcast producer, a keynote creator, roles she would never have hired for but that have always eaten into her time. For her, the tools are simply making possible what she could never have done alone. 

Shlomo and Snyder both say their businesses wouldn’t have been possible without the AI tools available today. As those systems become more capable—and start handling more of the product, operations, and even the business development work—the boundaries limiting how much one person can technically run are likely to expand again. What’s less clear is how big these businesses can get before a single person can no longer keep up.