AI and small business hiring: the NY Fed data

Q: Does AI reduce jobs at small businesses?

So far the evidence points the other way. In a Federal Reserve Bank of New York analysis published on October 8, 2026, based on 5,248 U.S. employer firms in the 2025 Small Business Credit Survey, firms using AI reported a net expectation of higher employment over the next twelve months of 33 percentage points, against 15 for non-users, and 77 percent of respondents said AI had not changed their labor costs. These are expectations, not realized outcomes, but they suggest small firms use AI to add capacity rather than to cut staff.

Q: What did the New York Fed find about AI and small businesses?

The New York Fed's Liberty Street Economics post (October 8, 2026) found that 46 percent of small employer firms were using AI tools and another 15 percent planned to adopt within a year. Among users, 51 percent had integrated AI partially or fully into their business processes. AI users expected stronger employment and revenue growth than non-users, and the gap held at about 10 points for employment and 14 points for revenue after controlling for firm characteristics.

Q: Where should a small business start with AI?

Start with one high-volume, rule-heavy process where people spend hours on repetitive work, such as order intake, invoice matching, scheduling or inventory updates. Measure how many hours it takes today, automate that one process with a human approving anything financial or customer-facing, and redeploy the hours saved to work that grows revenue. Expand to a second process only after the first one has proven its value.

Q: How do I keep my data safe when adopting AI?

Data privacy is the most common reason small businesses hold back: 38 percent of non-adopters in Homebase's 2026 Main Street AI Gap Report named it. The practical answer is architecture, not avoidance. Run the agent inside your own environment where possible, give it read-only access by default, keep a person approving every write to your systems, and keep a full log of what it did.

AI capacity is the extra output a business gets when software takes over the repetitive part of a job, so the same people can handle more customers, more orders and more growth. For two years the loudest question about AI has been how many jobs it will remove. On October 8, 2026 the Federal Reserve Bank of New York published an answer from the companies that matter most to this blog: small employers. Firms that already use AI expect to hire more over the next year than firms that do not, and they expect faster revenue growth too. Most of them report that AI has not changed their labor costs at all. For a traditional or small business, that reframes the decision. The question is no longer whether AI will replace your team. It is how much more your current team could handle if the routine work stopped landing on their desks.

Key takeaways. In a New York Fed analysis of 5,248 U.S. small employer firms, 46 percent were using AI tools and another 15 percent planned to within a year. AI users reported a net expectation of higher employment of 33 percentage points, against 15 for non-users, and 77 percent of respondents said AI had not changed their labor costs. A separate Homebase survey found adoption rising from 64 to 74 percent in a year, with the biggest gap between the smallest and largest firms. The practical lesson: small firms are using AI as capacity, not as a layoff tool. Start with one repetitive process, measure the hours it frees, and point those hours at growth.

What the New York Fed actually found

Here are the numbers, read carefully. The Liberty Street Economics post by Will Aarons and Asani Sarkar draws on the 2025 Small Business Credit Survey, fielded between September and November 2025, covering 5,248 firms with at least one employee. Of those, 46 percent said their business or employees were using AI tools. Among users, 63 percent called AI somewhat or very important to production, and 51 percent had integrated it partially or fully into their business processes. The headline is the hiring outlook. AI users reported a net expectation of higher employment over the next twelve months of 33 percentage points, compared with 15 for non-users. Importantly, the gap stayed at about 10 points after the authors controlled for past employment changes and other firm characteristics. Revenue followed the same pattern: a net 48 points for users against 21 for non-users, and a 14-point advantage after controls.

Why this is capacity, not replacement

The authors are careful to say these are expectations, not proven outcomes, and that is the right way to read them. Still, the shape of the data is telling. If small firms were using AI mainly to cut payroll, you would expect labor costs to fall and hiring plans to shrink. Instead, 77 percent of respondents reported no change in labor costs from AI, while users expected to grow headcount faster. In other words, the typical small business is using AI to absorb work it could not staff, not to remove people it already has. That matches what we see in the field. A 30-person distributor does not fire the person who keys in orders. It lets that person stop retyping purchase orders and start chasing the quotes that were going unanswered. This is the same pattern behind the McKinsey earnings gap: value comes from redesigning a process, not from removing a role.

The gap inside small business is widening

However, not every small business is moving at the same speed. Homebase's 2026 Main Street AI Gap Report, a survey of 750 small business decision-makers, found overall adoption rising from 64 percent in 2025 to 74 percent in 2026. Yet adoption ranged from 54 percent among businesses under $500K in revenue to 81 percent among those above $5M, a 27-point gap. The most common uses were practical back-office work: data analysis at 47 percent, payroll and timekeeping at 45, inventory management at 44 and administrative paperwork at 43. Meanwhile, the top reasons non-adopters held back were data privacy and security concerns (38 percent) and fear of mistakes (31 percent). Both are solvable with architecture rather than avoidance, which we cover in our guide to deploying AI inside your own environment.

What this means for a traditional business

For a company without an R&D department, the takeaway is simple. The firms pulling ahead are not buying AI to shrink. They are using it to say yes to more work with the team they already trust. Consequently, the right first project is not "where can we cut a position" but "where are skilled people spending hours on work a machine could prepare for them". Order intake, invoice matching, scheduling and inventory updates are the usual answers, and they are exactly the processes our back-office automation guide walks through. Moreover, framing the project as capacity changes how your team receives it. People adopt a tool that removes the part of their job they dislike far faster than one they suspect is measuring them for replacement. That is often the difference between a pilot that stalls and an automation that sticks.

What to do this quarter

List the work you are turning down. Write down the orders, quotes, follow-ups or customers your team cannot get to today. That list is your capacity problem, and the business case.
Pick one repetitive process feeding that list. Choose something high-volume and rule-heavy, and measure how many hours it takes per week right now. Our small-business playbook shows how to choose.
Automate it with a human in the loop. Read-only access by default, a person approving anything financial or customer-facing, and a full log of what the agent did.
Redeploy the hours on purpose. Decide in advance where the freed time goes, such as faster quotes or more customer calls, so the savings turn into revenue rather than slack.
Measure, then expand. Compare hours and output against your baseline after 60 to 90 days. Our payback benchmarks help you judge whether it worked before you add a second process.

Frequently asked questions

Does AI reduce jobs at small businesses?

So far the evidence points the other way. In the New York Fed's October 2026 analysis, small firms using AI expected net employment growth of 33 points against 15 for non-users, and 77 percent said AI had not changed their labor costs. These are expectations, not outcomes, but they point to AI as added capacity.

What did the New York Fed find about AI and small businesses?

46 percent of small employer firms used AI tools and another 15 percent planned to within a year. AI users expected stronger hiring and revenue growth than non-users, and the gap held after controlling for firm characteristics: about 10 points on employment and 14 on revenue.

Where should a small business start with AI?

With one high-volume, rule-heavy process such as order intake, invoice matching or inventory updates. Measure the hours it takes today, automate it with a human approving financial or customer-facing steps, and redeploy the hours saved to growth work.

How do I keep my data safe when adopting AI?

Treat it as an architecture question. Run the agent inside your own environment where possible, read-only by default, with a person approving every write and a full log of what it did.

Small businesses using AI expect to hire more, not less: what it means for yours