Small firms using AI expect to hire more, not less. What the Fed's new data means for your office

What's new

On Oct. 8, two economists at the Federal Reserve Bank of New York, Will Aarons and Asani Sarkar, published an analysis of how small business owners who use AI see their next year. It draws on the Fed's Small Business Credit Survey, fielded from September to November 2025. The AI questions went to 5,248 employer firms, weighted to match the national population of businesses with at least one employee.

The headline: owners who use AI were much more likely to expect to add workers. The same week, a separate survey from the Small Business & Entrepreneurship Council found only 16% of small employers think AI eliminates more jobs than it creates.

That lands in the middle of a very different story from big tech. HubSpot said on Oct. 6 it would cut about 660 jobs, roughly 7% of its staff, as it reorganizes around AI. Both things can be true. They're about different kinds of companies.

The numbers

The Fed researchers use a "net" measure: the share of owners expecting an increase minus the share expecting a decrease.

Owners using AI expect more growth

Net share of small employer firms expecting an increase over the next 12 months (percent expecting more minus percent expecting less). Before adjusting for firm differences.

1. Employees AI usersNet expecting to add staff+33 pts
2. Employees Non-usersNet expecting to add staff+15 pts
3. Revenue AI usersNet expecting higher revenue+48 pts
4. Revenue Non-usersNet expecting higher revenue+21 pts

Source: Aarons and Sarkar, Liberty Street Economics, Federal Reserve Bank of New York, Oct. 8, 2026 (2025 Small Business Credit Survey). Expectations, not results.

The obvious objection is that growing firms try new tools, so of course AI users are optimistic. The researchers tested for that. After controlling for firm and owner traits, AI users were still 13 points higher on hiring expectations, and 10 points higher after also accounting for each firm's recent change in employment. They also checked whether naturally optimistic owners explain it and concluded they don't.

The Fed's full survey report, released in March, adds context. 46% of employer firms said they or their employees use AI, and another 15% planned to start within a year. Among users, 71% reported higher productivity, 31% higher sales, and the Liberty Street post says 77% saw no change in labor costs. Use is still shallow: the most common uses were writing or marketing (83%), and only 7% said AI was fully built into their business processes.

Small firms aren't using AI to shrink. They're using it to keep up.

What the other side has right

AI is showing up in layoffs. Challenger, Gray & Christmas, which tracks announced job cuts, says employers have cited AI in 120,136 announced cuts so far in 2026, about 21% of the total, making it the leading reason for the year. Technology companies account for 165,925 of this year's cuts. In September alone, though, AI fell to fifth, at 3,961 cuts.

Even HubSpot's case is mixed. Its CEO wrote that the company is shifting toward delivering AI-driven outcomes, but also that the cuts are not driven by AI-related efficiencies: "This is not simply a cost-cutting exercise," she wrote, as reported by Boston.com.

And the Fed numbers have limits the authors state themselves. They measure what owners expect, not what happened. Whether AI users actually hire more is, in the authors' words, a subject for future research. The survey was also fielded in late 2025, nearly a year ago. The SBE Council survey is smaller (516 owners of firms with 2 to 99 employees, fielded Sept. 11-19, 2026) and comes from a small-business advocacy group.

Our take for a 40-400 person company

Many of the biggest layoff headlines come from tech companies changing what they sell; by Challenger's count, tech accounts for about 29% of this year's announced cuts. A distributor or manufacturer in Dallas-Fort Worth is in a different spot. Your office isn't overstaffed with people writing code; it's usually short on time. Orders wait in an inbox, quotes go out a day late, and follow-ups get skipped.

That's where the Fed data fits. If AI takes the retyping off a desk, the same person handles more orders, answers customers faster, and gets to work nobody had time for, like chasing late shipments or cleaning up pricing. That's more capacity for a growing company, not fewer people. We wrote about that kind of work in the work your team never had time for.

It won't be painless everywhere. Roles change. The person who spent half the day keying POs now spends it on exceptions and customers. That's a better job for most people, but it needs to be planned and explained, not discovered.

What changes, and what doesn't

What to do this quarter

1. Pick one desk with a backlog, like order entry or AP, and time it: minutes per item, items per person.
2. Decide up front what freed-up time goes to. Name the work: faster quotes, follow-ups, a second shift's orders.
3. Tell the team before you start. Say plainly whether the goal is growth without hiring or a smaller team. Uncertainty does more damage than the tool.
4. Measure after 60 days against the same numbers. Before you hire for that desk again, read what to check before you hire.