The short version: American business applications hit 5.6 million in 2025, up 24 percent since ChatGPT launched, and it is not hype driving the number. AI is lowering what economists call the minimum efficient scale of a business, the revenue point at which hiring a specialist finally pays for itself. A solo founder can now cover accounting, marketing, customer service and compliance with AI tools instead of four separate hires. That is not a story about AI replacing workers. It is a story about AI making it possible to start the business in the first place, and about who is capturing the gains so far.
What does the new business formation data actually show?
New U.S. business applications reached 5.6 million in 2025, a 24 percent increase since ChatGPT’s public launch, according to research from Citadel Securities citing U.S. Census Bureau data. Citadel also tracked a smaller but telling shift: the median employee count at seed-stage startups dropped from five to four in the first quarter of 2023, right after ChatGPT’s release. Fewer people are needed to get a company off the ground, and more people are trying.
Citadel analyst Frank Flight put it plainly: those gains “may accrue most powerfully to the bedrock of the American economy: small businesses and entrepreneurs.” That is a notable claim from a firm that spends most of its time watching public markets, not Main Street.
Why does AI lower the bar for starting a business?
The mechanism has a name: minimum efficient scale, the revenue level at which it finally makes financial sense to hire a specialist instead of doing the job yourself or going without. Below that line, a founder who once needed separate people for bookkeeping, marketing, customer support and compliance either did all four badly themselves or didn’t start the business at all.
AI tools change that math without changing the founder’s skill set. A single person can draft marketing copy, answer customer questions, track expenses and check contracts against a checklist, all with the same laptop, none of it requiring a payroll. The business becomes viable at a smaller size than it used to be, which is exactly why more of them are getting started. We have covered pieces of this shift before, including which workflows are worth automating first and which AI tools actually earn their keep by function.
What does this actually look like inside one company?
Matthew Gallagher launched Medvi, a GLP-1 telehealth company, from his Los Angeles home in September 2024 with $20,000 and no employees. In its first year, Medvi generated $401 million in revenue, served 250,000 customers, and posted a 16.2 percent net profit margin, with 2026 revenue projected to reach $1.8 billion.
Gallagher’s toolkit was ordinary and publicly available: ChatGPT and Claude for code and copywriting, Midjourney and Runway for advertising creative, ElevenLabs for voice-based customer communication. He didn’t try to do everything himself. Regulated functions like clinical oversight went to specialist partners, CareValidate and OpenLoop Health, while he kept ownership of the customer relationship. His second hire was his brother. For comparison, telehealth competitor Hims and Hers reported $2.4 billion in revenue with 2,442 employees and a 5.5 percent net margin.
Medvi is an outlier, not a template. Few founders will land in a category moving as fast as GLP-1 telehealth, and $401 million in year one is not a realistic bar for most small businesses. What is transferable is the pattern: keep the parts of the business that require judgment and the customer relationship, hand the rest to AI and specialist partners, and stay small on purpose for as long as that works.
Is this actually helping every small business, or just the biggest ones?
Here is the honest caveat. PYMNTS Intelligence’s May 2026 Small Business Week report found a real divergence: SMBs generating over $1 million annually reported average revenue growth of 13.7 percent, while businesses earning under $150,000 grew just 0.6 percent. Digitally advanced small businesses, the ones already selling across multiple channels and accepting varied digital payment methods, were also more confident about future growth.
That is not proof that AI alone caused the gap. But it lines up with what we would expect if AI adoption is one of the things separating the two groups. The tools that lower the minimum efficient scale only help the businesses that actually pick them up. A smaller shop still doing everything on paper does not get the benefit just because the technology exists.
Does this mean AI is replacing jobs at small businesses?
The data says no, not primarily. Citadel’s analysis of S&P 500 earnings calls found that companies describing AI as complementary to hiring outnumbered those describing it as a substitute across nearly all job functions. The more common pattern is deferral: a business uses AI to cover a function until it grows large enough that a human’s judgment becomes the actual bottleneck, and only then does it hire.
That distinction matters for how owners should think about this. AI is not standing in for a receptionist, a bookkeeper or a marketer so a company can shed headcount. It is buying founders time before they need that first or fifth hire, and when they do hire, it is usually because the work has outgrown what software can judge on its own.
Real-World Example: Jevons Paradox
This trend follows the Jevons Paradox. This is an economic rule where making a tool more efficient makes it cheaper to use.
- Example: Imagine a team writing 10 reports a week. AI allows them to write 30 reports in the same time.
- The Outcome: Because the work is faster and cheaper, the business gets more value. The company then hires more people to handle this new demand, rather than firing the original team.
What should a small business owner take from this?
If you have been putting off starting something because you could not picture staffing it, the math has changed. The functions that used to require a team, at minimum, to get off the ground, accounting, first-draft marketing, routine customer replies, basic compliance checks, can now run through AI tools while you handle the parts that actually need your judgment. Our practical guide to AI for small business is a reasonable place to start if you are figuring out where to begin.
If you already run a small business, the divergence in the PYMNTS data is the more useful number to sit with. Growth is concentrating among businesses that adopted the tools, not just the ones with more revenue to begin with. Waiting is not neutral.
Frequently Asked Questions
What is “minimum efficient scale” and why does it matter for AI and small business?
It is the revenue level at which hiring a specialist, for accounting, marketing, or another function, finally pays for itself. AI tools let a founder cover those functions before reaching that revenue level, so a business can start smaller and still work.
How many new businesses started in the US in 2025?
5.6 million new business applications were filed in 2025, a 24 percent increase since ChatGPT’s launch, according to Citadel Securities research citing U.S. Census Bureau data.
Is AI actually replacing employees at small businesses?
The available data says mostly no. Citadel’s review of S&P 500 earnings calls found companies far more likely to describe AI as complementary to hiring than as a replacement for it. The more common effect is delaying a hire until the business is large enough that human judgment becomes the limiting factor.
What AI tools did Medvi use to reach $401 million in first-year revenue?
Founder Matthew Gallagher used ChatGPT and Claude for code and copywriting, Midjourney and Runway for advertising creative, and ElevenLabs for voice-based customer communication, while outsourcing regulated clinical functions to specialist partners.
If AI could run one function of your business starting tomorrow, which would you hand over first, and which one would you never let go of? Tell us in the comments.
