Small Businesses Using AI Are Hiring More People, Not Fewer

Gusto just ran the payroll numbers on 2,262 small businesses. Companies using AI grew headcount 7% faster than the ones that only knew about it. For firms under 10 employees, the gap jumped to 10%.^[1]

Every doom take you've read this year says the opposite. "AI is coming for your jobs." "Klarna rehired humans." "Cost per agent is higher than the salary it replaced." All partially true. All missing the actual story for the $1M–$20M operator.

Here's what the data says — and what most of the noise gets wrong.

The three numbers that matter

Three separate reports landed in the last four weeks, and they all point the same direction.

Gusto (payroll platform, 2,262 customers): AI-adopting SMBs grew headcount 7% faster than comparable non-adopters. Under-10-employee firms grew 10% faster. The growth built quarter over quarter, from ~4.5% early to 6.8% by Q4.^[1]

General Assembly (526 HR leaders at US/UK companies with 50–999 employees): 67% plan to hire more workers by end of 2026. 30% are expanding headcount specifically to "scale faster, meet demand, and/or seize new opportunities." 87% already report ROI from AI. Only 21% adopted AI primarily to cut costs — 43% did it to save time, 36% to drive growth.^[2]

Axios covering both: "AI is leading to growth for entrepreneurs."^[3]

If you're running a business between $1M and $20M in revenue and you've been quietly worried you should be freezing hires until you figure out AI, the numbers are telling you the opposite. The operators who are shipping AI are the ones adding humans.

Why the doom takes get this wrong

The layoff headlines are real. Klarna. Salesforce. IBM. Duolingo. Big companies with big legacy headcount are cutting.^[2] That's true.

But those aren't your business. Those are companies where a 5% workforce reduction is 3,000 people. When they cut, they cut categories — support, sales dev, junior copy. And the numbers become a national headline.

Meanwhile, the $3M hospitality group down the street is adding a second location because their two-person ops team stopped spending 15 hours a week on invoice categorization. That doesn't make a headline. But it happens 10,000 times a quarter, and it's the majority of the actual economy.

The doom take confuses enterprise-scale cost-cutting with operator-scale capacity expansion. These are opposite motions. Same technology, opposite direction.

What the growth-hire actually looks like

The Gusto data has a detail most coverage missed. When small AI-using businesses hired more people, they didn't hire more admins.

They hired therapists at health care businesses and cooks at restaurants.^[1]

That's not a footnote. That's the whole point. AI didn't replace the front-of-house or the practitioner. It replaced the invoice categorization, the scheduling ping-pong, the reply-all customer-service loop. Which freed the owner to sell one more session, book one more table, take one more call — and eventually to add a human who does the revenue-generating work.

The pattern:

1. Owner or ops person is capped at 60 hours a week
2. AI takes 8–15 hours a week of administrative drag off their plate
3. That time gets redirected to sales, ops, or product — the revenue-side work
4. Revenue moves. Not doubled. But up enough that the next hire pencils out
5. New hire is a service delivery role, not an admin role

That's the mechanism. That's why the Gusto payroll data shows growth instead of contraction. It's not that AI grows the business. It's that AI frees the owner from the drag long enough for the business to grow itself.

What most operators are actually doing wrong

The under-10 team gets a 10% headcount lift. The 10+ team gets nothing.^[1]

That's the finding that should worry you if you're already at 15, 25, 40 people. The pattern that works for the two-person shop doesn't scale automatically. Something breaks between 10 and 100.

I have a theory. At 5 people, the owner picks up ChatGPT on Monday and uses it Tuesday for a real task. Feedback loop is 24 hours. At 40 people, the owner buys Copilot licenses, sends a memo, and 6 months later checks in and finds two people love it, three tolerate it, and 35 haven't logged in since onboarding.

This lines up with General Assembly's stat that 32% of SMBs say a lack of skills is their #1 challenge scaling AI.^[2] The tools aren't the problem. The workflow-mapping is the problem. The rollout is the problem.

Their most-effective fix, per HR leaders themselves: role-specific training. Not generic "here's how ChatGPT works" — training tailored to the marketing team's actual campaigns, the ops team's actual SOPs, the finance team's actual reconciliation flow. 35% of leaders picked it as the top-impact intervention. Only 18% said "let people experiment on their own" worked.^[2]

Which means: buying licenses without mapping workflows is the exact thing 82% of the smart-money HR leaders will tell you doesn't work. It's also the default rollout at most 20–50 person companies.

The actual play for Q4

If you're in that $1M–$20M range and you're planning Q4, the data suggests three things.

Don't freeze hiring because you're waiting to see what AI does. The operators winning right now are hiring service-side roles the AI can't do — cooks, therapists, sales, closers, delivery. The AI is doing the admin drag that lets those hires make sense.

Don't buy licenses without a workflow map. Pick 2–3 real workflows your team runs weekly. Map them. Then pick the tool. That's what the 87%-ROI cohort in General Assembly's survey is doing.^[2] The order matters.

Don't try to build the org chart around "AI + fewer humans." Build it around "same humans, less drag, more revenue-side output." The math on that gets you to a second location or a second product line. The math on "AI replaces the intake coordinator" gets you a broken customer experience and a Klarna-style walk-back in nine months.^[2]

McKinsey's latest number, for context: only 23% of enterprises are actually scaling an agentic AI system in a business function.^[4] Most of the rest are still in pilots. The gap between "AI is transforming everything" and "AI is deployed at your scale" is much wider than the podcast circuit lets on.

Which means: you're not behind. The window is still open. But it's open for the operators who map workflows before they buy licenses, and it's closing for the ones who treat AI like a headcount tool instead of a capacity tool.

What I'd build first

If a $5M business owner walked in tomorrow and said "where do I start," the answer for 8 out of 10 of them is the same: agent the intake or the ops backlog. Not the customer-facing work. Not the creative work. The stuff nobody wants to do and nobody's mad it went away — invoice categorization, appointment reminders, follow-up sequences, quote drafting, expense reconciliation, meeting notes to CRM.

That work eats 10–15 hours a week for at least one person in most sub-20-person businesses. Reclaim that first. Watch what the owner does with the time. Then plan the second build around what actually moved the number.

That's the pattern the Gusto data is showing. That's why the payroll went up instead of down.

If you want that map built for your specific business — the workflows worth automating, the ones to leave alone, and the order to build them in — that's what the audit call is for. 30 minutes, no pitch, you leave with the map either way. Book it here.

Sources 4 references

1. 1
   Small Businesses That Adopted AI Are Hiring Faster, New Gusto Research Finds

   PRNewswire / Gustoreport

   Gusto payroll data on 2,262 SMBs: AI adopters grew headcount 7% faster; under-10 firms grew 10% faster; hires were revenue-side roles.

   ↩
2. 2
   SMBs Attract Talent with AI Upskilling as Large Firms Cut Jobs and Benefits

   USA Today / General Assemblyreport

   526 HR leaders surveyed; 67% plan to hire more, 87% see AI ROI, 21% adopted AI to cut costs, role-specific training most effective.

   ↩
3. 3
   Small Businesses Using AI Are Hiring More, Not Less. New Study Finds a 10% Boost for the Smallest Firms

   International Business Timesnews

   Covers the Gusto findings; smallest firms (under 10) get a 10% headcount boost, and new hires are service-side roles (cooks, therapists), not admins.

   ↩
4. 4
   AI Agent Adoption Statistics 2026: Enterprise AI Usage

   GoGloby (citing McKinsey State of AI 2026)analysis

   Agentic AI scaling is at ~23% of organizations per McKinsey 2026 data; single-function scaling under 10%.

   ↩