The loudest AI-and-jobs claims of 2026 have run in opposite directions: predictions of a white-collar apocalypse on one side, blue-chip companies blaming AI for restructuring that was always coming on the other. A National Bureau of Economic Research working paper revised this month cuts through both with something rarer than a forecast — a very large sample of what companies actually say happened to their own workforces.
🔍 THE BOTTOM LINE: Nearly 9,000 executives across seven countries report that AI has had essentially no net effect on firm employment over the past three years, and most expect little over the next three. But the average is hiding the real story: more than 40 percent of firms expect meaningful job gains or losses that offset each other only in aggregate. AI looks less like a wrecking ball and more like a reshuffle — new roles appearing in the same firms where old ones go.
What the survey found
The paper — “Firm Data on AI”, co-authored by a team led by the Bank of England’s Ivan Yotzov and Stanford’s Nicholas Bloom — surveys CFOs, CEOs and senior executives at firms in the US, UK, Germany, Australia, Canada, Mexico and South Africa. It was first issued in February and revised in October 2026 with fresh summer data. As Forbes’ Joe McKendrick summarised it, the topline numbers are strikingly calm:
- 78 percent of firms now use AI, with US firms expecting to spend about $2,000 per employee on it in 2026 — roughly $300 billion in aggregate, and heavily concentrated: 2 percent of firms account for over half of all spending.
- 87 percent of executives report no effect on their firm’s employment over the last three years; 7 percent saw growth, 6 percent saw cuts.
- Looking forward, 57 percent still expect no impact, while 28 percent expect AI to reduce headcount and 16 percent expect it to add jobs. The expected average decline is about 0.4 percent over three years — modest, and measured only at existing firms.
- More than 40 percent of firms anticipate positive or negative employment effects that roughly cancel out only across the whole sample.
- Productivity effects are small but growing: about 0.2 percent a year so far, with executives expecting closer to 0.6 percent per annum over the next three years.
As the Yahoo Finance syndication of the Forbes piece adds, the co-authors interpret the offsetting moves as evidence that “AI could reshape the distribution of employment across firms, generating major labor market churn, while leaving overall employment levels broadly unchanged.” New-firm hiring, they note, isn’t even counted in the projection — the 0.4 percent average covers incumbents only.
The caveats worth holding onto
Executive surveys measure belief, not payroll. stated intent to hire or cut is not the same as observed hires, and self-reported “no effect” can describe a firm that quietly froze hiring just as easily as one that genuinely held steady. The sample also covers seven economically large countries — New Zealand is not among them — so local readers are reading an analogue, not home data.
There’s also a timing argument, made forcefully by Harvard economist Doug Elmendorf in Harvard Business School’s Working Knowledge recap of related research: what we’ve seen so far “has very little predictive power” for what AI does to labour markets over the next five to fifteen years. HBS’s Joseph Fuller, in the same piece, puts a number on why the disruption has been slow — only about a third of corporate AI experiments succeed, and most employees use the tools “like glorified Google”. Neither reason is exactly reassuring. One says the quiet is temporary; the other says companies are bad at the thing that’s supposed to be displacing people.
Where New Zealand fits in
Firm-level surveys can’t be read straight onto a 5-million-person economy, but the shape of this finding matches what New Zealand’s own narrow data has been showing all spring. Local accountants who use AI at work overwhelmingly don’t expect job cuts — our coverage of the NZ accounting survey last week found 94 percent expecting no reductions — while AI mentions in NZ job ads have been climbing much faster than hands-on adoption and PwC’s barometer found NZ AI hiring more than doubled in a year. Meanwhile the US monthly job-cut tracker keeps ranking AI well down its causes list and falling.
The useful career reading from the NBER churn finding isn’t “AI won’t take my job”. It’s that within-firm movement is the default expectation: task mix shifts, some roles shrink while neighbours grow, and the person best placed through a reshuffle is usually the one already holding the new tasks. That’s a different — and more actionable — picture than the one the layoff headlines suggest.
FAQ
Has AI actually eliminated jobs so far, per the data? In this survey of nearly 9,000 firms, 87 percent of executives report no employment effect from AI over the last three years, against 6 percent reporting cuts and 7 percent growth. Net effect across firms: essentially zero.
What is “job churn”? The NBER authors use it to describe simultaneous hiring and shedding at different firms that cancels out in aggregate. Over 40 percent of firms in the survey expect gains or losses large enough to matter individually — they just don’t sum to much nationally.
Will AI cut jobs in the next three years? The average expectation in the survey is a modest 0.4 percent decline at incumbent firms, with 28 percent of executives expecting cuts and 16 percent expecting growth. The authors stress new firms’ hiring isn’t included and could offset some or all of that.
What does this mean for New Zealand workers? NZ wasn’t surveyed, but local signals — accountants not expecting cuts, AI hiring up strongly per PwC, AI mentions in job ads rising — rhyme with the survey’s finding that AI so far reshuffles work more than it subtracts it.
— CJ Murden, editor of Singularity.Kiwi. Former digital technologies teacher, author of AI-focused books. Writing with a New Zealand focus.
