A nine-person business that adopted AI ended the year with roughly one more employee than a nearly identical business that didn't. Not one fewer.
That comes from Gusto, which compared actual payroll records—not survey answers—for 1,593 small businesses that adopted AI with 669 comparable businesses that hadn't, tracking both for a full year. The adopters grew headcount about 7% more. Among businesses with fewer than 10 employees, the gap widened to about 10%.
Then it stops. Above 10 employees, the researchers found no significant difference. That cutoff is the most useful number in the study, and it's the one nobody is quoting.
They Didn't Hire Assistants
Look at what the money bought. The added roles were therapists, dental hygienists, cooks, technicians, teachers, front-desk staff — people who serve customers directly. Health care and social-assistance practices under 10 people, the dental offices and mental health clinics of the world, grew headcount nearly 20% more than their non-adopting peers.
The pattern is legible once you picture the owner. In a nine-person business, the owner is also the bookkeeper, the marketer, the scheduler, and the person writing the intake email at 10 p.m. AI ate that second job. What it freed wasn't an employee's afternoon — it was the founder's capacity to sell, serve, and open another chair.
"This data pushes back on the idea that AI is coming for jobs on Main Street," says Nich Tremper, senior economist at Gusto. "The smallest businesses are the ones who are most used to doing more with fewer resources."
Why the Effect Dies at Ten
Here's the reading the coverage skipped. In a business of five, the constraint is the owner's hours. Remove the admin load and growth follows almost mechanically — there is nothing between the freed time and the customer.
In a business of thirty, the constraint isn't hours. It's coordination: handoffs, approvals, meetings, the internal traffic that grows faster than headcount. Time freed there doesn't reach a customer; it lands in someone's inbox. Which is precisely what ActivTrak's analysis of 443 million work hours found at larger firms: collaboration time up 34%, chat volume up 145%, focused work at a three-year low.
Same tool. Opposite result. The variable isn't the software — it's whether your organization has enough internal surface area to absorb the time before a customer sees it.
Independent evidence points the same direction. The JPMorganChase Institute, tracking real payments to AI services across 4.6 million Chase business banking clients, found small employer firms adopting at 27.6% versus 19% for large firms without employees—organizational capacity matters more than money, in a market where entry now costs $20 to $30 a month.
The Caveat That Matters
Gusto is upfront about the limits, and we should be too. The sample isn't random — it's Gusto customers who answered a survey. And the finding is correlational: businesses that adopt AI early may simply be the ones already going to grow. Ambition is hard to control for.
That doesn't void the result; it narrows the claim. The honest version is that AI adoption and hiring are moving together at the smallest end of the market, and that payroll data shows no measurable job destruction. Gusto's earlier survey work fits: 95% of regular AI users weren't cutting headcount, 34% were upskilling people, and only 5% reduced staff.
The Gap Between Buying and Using
The uncomfortable finding sits in that same research. More than 80% of small businesses have no formal AI strategy — and it shows in the returns.
Among businesses with a written strategy, 32% reported productivity gains above 20%. Among those without one, 14% did. Add training and the number climbs to 60%. Roughly 43% of small businesses offer any AI training at all.
So the tool is nearly free, adoption is broad, and the results are wildly uneven—separated mostly because some people sat down and decided what the thing was for.
What This Means
If you run a business under 10 people, this is close to permission. The data says the growth path runs through offloading the owner's back office, then reinvesting that capacity in the people your customers actually meet. Automate the intake email, not the hygienist.
If you're between 10 and 50, don't assume the same result transfers. Your bottleneck is probably coordination, and buying more tools into a congested organization usually creates more internal traffic, not more customers. Fix the handoffs first, then automate.
And whatever your size: write the strategy down. One page naming which three tasks AI owns, who is trained on it, and what you expect to gain moves you from the 14% cohort to the 32% one. That's a bigger effect than any tool choice you'll make this year.
The Main Street story of AI, at least in the payroll records, isn't replacement. It's leverage—and it lands hardest on businesses small enough to feel it immediately.
Discussion