Microsoft laid off nearly 5,000 people in early July while pouring billions into AI data centres. Amazon cut roughly 30,000 jobs between the end of 2025 and the start of this year. Oracle shed thousands. The July jobs report showed employers unexpectedly cut 23,000 positions. And nobody can agree on whether AI is actually to blame.
A Fortune investigation published August 8 lays out the problem: the data is not keeping up with the narrative. Companies attribute layoffs to AI to appear forward-thinking, or avoid mentioning AI for fear of public outcry — a phenomenon researchers call “AI washing.” The result is a labour market picture where CEOs blame AI one month and hail it as a job-creation engine the next, and economists are left trying to separate signal from spin.
💡 THE BOTTOM LINE: The AI-jobs debate has a measurement problem. Companies are using AI as a convenient layoff narrative, the data is too slow to confirm or deny it, and meanwhile 200 economists are warning that the displacement may be real and large-scale — just not visible in the numbers yet.
The Contradictory Data
The evidence cuts both ways, and neither direction is clean.
A study by financial services firm Ramp analysed more than 21,000 US firms and found that companies that invested heavily in AI grew their headcount. The top “high-intensity” AI spenders expanded overall staff by 10 per cent and boosted entry-level hiring by 12 per cent over two years — defying reports that college graduates face a barren job market. The bottom two-thirds of adopters saw no headcount growth at all.
A report by Google researchers found AI is mostly being used as a collaborative tool rather than a job-replacer. A California Policy Lab study found no statewide spike in unemployment insurance claims among AI-exposed roles like software developers and customer service reps since ChatGPT’s release in late 2022 — but did find elevated claims for college-educated workers in highly-exposed roles, and a significant increase in claims from high-exposed roles in the San Francisco area specifically.
Then there is the July jobs report: employers cut 23,000 jobs, an unexpected contraction. And a 2025 report co-authored by Stanford economist Erik Brynjolfsson found that workers aged 22–25 in AI-exposed roles suffered a 16 per cent relative employment drop compared to less-exposed peers.
The Ramp economist, Ara Kharazian, offered a reading that splits the difference: Big Tech “definitely overhired during the pandemic and are now making the decisions to correct that overhiring.” Some are “blaming it on AI. But what we’re seeing from firms that are using AI that didn’t have that overhiring problem is that they’re continuing to grow.”
Brynjolfsson, responding to the Ramp study, wrote on X that firms adopting AI “may grow by gaining market share from non-adopters, so employment can rise among adopters even as exposed occupations shrink economy-wide.” Both things can be true at once.
200 Economists Say Act Now
In July, nearly 200 economists and researchers signed a statement warning that AI could cause large-scale job displacement in the next decade. The signatories include Anthropic co-founder Jack Clark and Eric Schmidt, Google’s former chief executive.
“This could drive an unprecedented transformation of our economy, larger than the Industrial Revolution, but unfolding over a vastly shorter time frame,” the statement reads. It calls on policymakers to “act now” to better understand how AI is transforming the economy and to create legislation that will “steer A.I. in a direction that complements humans and benefits society.”
This aligns with other recent warnings. Verizon CEO Hans Vestberg warned of 25 million AI-driven job losses. A Stanford SIEPR analysis found the AI jobs “apocalypse” is not happening — yet, but that the exposure data is building. The MIT $2 trillion wage displacement study put coding at the front of the queue.
The tension between these warnings and the current data is the core of the problem. The displacement may be coming, but it is not yet visible in aggregate employment statistics — partly because it is concentrated in specific demographics (young, college-educated, tech-hub workers) that are too small to move the national numbers, and partly because companies are simultaneously hiring in new AI-adjacent roles.
The “AI Washing” Problem
Fortune identifies two variants of the phenomenon. In the first, companies attribute layoffs to AI to seem forward-thinking and strategically decisive. In the second, companies avoid mentioning AI for fear of public outcry, making the true scale of AI-driven displacement invisible.
Ben Zipperer of the Economic Policy Institute told Fortune that AI’s impact on jobs has so far been more limited than doomsday scenarios predicted. UCLA economist Till Von Wachter said pinning down the extent to which observed layoffs are really driven by AI has been “notoriously hard.”
Amazon’s CEO Andy Jassy has said AI would lead to a leaner workforce but also that AI could ultimately fuel job creation, and has framed Amazon’s layoffs as an attempt to flatten its organisational structure. An Amazon spokesperson said AI has not been the reason behind the majority of its layoffs and that AI adoption is not a factor in deciding layoffs — a claim that sits uneasily alongside the company cutting 30,000 jobs while investing heavily in AI infrastructure.
Workers on the ground report a different experience. An Amazon Employees for Climate Justice spokesperson told Fortune that employees feel “huge increased pressure” from executives to finish tasks faster using AI, and that AI tools have raised the demand for output. That pressure — more work expected per worker, enabled by AI — does not show up as a layoff in the data. It shows up as the same headcount producing more.
The NZ Angle
New Zealand’s exposure to AI-driven displacement follows the same pattern as the rest of the developed world — concentrated, specific, and hard to measure in aggregate. The jobs most exposed to AI in NZ are the same ones showing elevated unemployment claims in the US data: software developers, customer service representatives, and entry-level knowledge workers.
The difference is that New Zealand has fewer large companies making large public AI-driven layoff announcements. The displacement, if it comes, is more likely to look like the Amazon pattern — the same headcount, more output expected, pressure to use AI tools — than the Microsoft pattern of 5,000 names on a list. That makes it harder to see, not less real.
The policy question the 200 economists raise — whether governments should act now to steer AI toward complementing rather than replacing workers — is one that New Zealand has largely deferred. The NZ AI Governance Framework remains consultative rather than regulatory. The EU’s AI Act Phase 2, which began enforcement on 2 August, requires risk assessments and human oversight for AI used in employment decisions. NZ has no equivalent obligation.
❓ FAQ
What is “AI washing”? The practice of companies attributing layoffs to AI to appear strategically forward-thinking, or conversely, avoiding mentioning AI to dodge public criticism. Both distort the true picture of AI’s impact on employment.
Is AI actually causing mass layoffs? The evidence is mixed. Some studies show AI-heavy firms are hiring. Others show concentrated employment drops for young, college-educated workers in AI-exposed roles. The disagreement is partly because the data lags the narrative and partly because companies are using AI and hiring simultaneously.
What did the 200 economists warn about? That AI could cause large-scale job displacement in the next decade — potentially larger than the Industrial Revolution, over a much shorter timeframe. They called on policymakers to act now to steer AI toward complementing workers rather than replacing them.
What is the Ramp study finding? Companies that invest heavily in AI (“high-intensity adopters”) grew their headcount by 10 per cent and entry-level hiring by 12 per cent over two years. The bottom two-thirds of AI adopters saw no headcount growth. The lead economist noted that Big Tech firms overhired during the pandemic and are correcting, while blaming AI.
🔍 THE BOTTOM LINE
The AI jobs debate has become a Rorschach test. If you want to argue AI is creating jobs, the Ramp study gives you the numbers. If you want to argue it is destroying them, the Brynjolfsson data and the 200 economists give you that. The truth is probably both at once — AI is concentrating employment among adopters while hollowing out specific exposed roles, and the aggregate data is too blunt an instrument to see the difference. Meanwhile, companies are using AI as a convenient narrative for decisions they were making anyway. The real question is not whether displacement is happening now. It is whether policymakers will have a framework in place when it becomes visible enough that nobody can argue about it anymore.
