Here’s a finding I didn’t expect to come from a central bank.
This month the Federal Reserve Bank of New York looked at 5,248 small American businesses, every one of them with at least one employee, and asked the owners what they expect to happen to their headcount over the next year.
The owners using AI expect it to go up. By more than the owners who aren’t.
The numbers
Among firms already using AI, the share expecting to employ more people outweighed the share expecting fewer by 33 percentage points. Among firms not using it, the margin was 15.
The obvious objection is that AI users are just a more optimistic crowd: younger owners, more profitable firms, bigger teams. The researchers checked. After accounting for owner and firm characteristics and location, the gap was still 13 points. After also accounting for whether the business had added staff in the past year, it was still 10.
Then the part I keep coming back to. The gap was wider among owners who described their finances as fair or poor (21 points) than among those doing well (10). If this were just confident people being confident, you’d expect the reverse.
And 77% of the businesses reported no change in their labour costs from using AI.
What it doesn’t say
I want to be careful with this, because it would be easy to oversell.
It’s US data, not New Zealand. It measures what owners expect, not what they’ve done; the authors say outright that whether the hiring actually happens is a question for future research. The fieldwork was a year ago. And only 31% of the businesses said AI had increased their sales, so this isn’t a story about AI working miracles either.
What it does say is narrower and, I think, more useful. In small businesses, AI so far looks like a way to get more done with the people you have. The owners using it are planning to keep those people, and add to them.
Now put it next to what your staff believe
Employment Hero’s New Zealand research found that 42% of workers believe heavy AI use shows they’re no longer needed.
So there’s a gap. On one side, owners (at least in this large US sample) using AI and planning to hire. On the other, staff reading their own AI use as evidence that they’re surplus.
Nobody closes that gap by accident.
The silence is doing the talking
Most owners I talk to have no plan to replace anyone with AI. A fair few would be surprised to learn their team suspects otherwise. They’ve never said so, because it never occurred to them that it needed saying.
It does. When nobody says what AI is for, people fill the gap with the headlines they’ve read, and the headlines are about job losses.
“AI won’t replace anyone here” is a promise nobody can fully make, and your team knows it. What you can say, and stand behind, is more specific:
- What we’re using it for. Name the actual jobs: the quote write-ups, the rosters, the first draft of the monthly report.
- What happens to the time it frees up. If the answer is “more of the work we never get to”, say that.
- What isn’t changing. The parts of each role that stay with people.
- Who decides. And how people will hear about it if any of this changes.
None of that needs a policy document. It needs ten minutes at a team meeting and the discipline to repeat it.
The honest version
I should declare an interest: I work on helping businesses bring AI in, so I’d naturally prefer the cheerful reading of this data. That’s why the caveats are up front.
A survey of what American owners expect is no guarantee to the person in your accounts team wondering about her job. But it’s good evidence that replacement isn’t the default plan for small businesses using AI. If it isn’t your plan either, the cheapest thing you can do this week is say so out loud.