Over recent months, dire predictions have been swirling around artificial intelligence and its supposed power to eliminate vast swathes of employment. Tech CEOs warn of mass layoffs, dystopian futures where humans are redundant, and dramatic transformations in the workplace. But a new study from Yale’s Budget Lab, in collaboration with the Brookings Institution, offers a more measured counterargument: in the U.S., generative AI does not appear to have triggered widespread job loss since the debut of ChatGPT in late 2022 (in fact, its effect is not noticeably greater than earlier waves of technological change) (Financial Times).
What the Study Found
The research analysed official U.S. labour market data alongside measures of how heavily industries were exposed to AI use. The major takeaways:
- There was no clear correlation between high exposure to generative AI and disproportionate job losses across sectors.
- Graduate unemployment has increased, for example, degree holders aged 20–24 reached 9.3 per cent unemployment in August 2025, but that rise does not seem to be uniquely attributable to AI (Financial Times).
- In essence, the study argues that AI so far is reshaping job tasks and roles, rather than dismantling broad employment categories.
- The authors caution that the full labour-market impact of AI remains uncertain and evolving.
As Molly Kinder of Brookings put it: “Despite how quickly AI technology has progressed, the labour market over the past three years has been a story of continuity over change” (Financial Times).
This counters the narrative pushed by some industry leaders who forecast mass disruption. For example, Amazon’s CEO has warned publicly that AI will mean fewer “corporate” jobs (Financial Times). Others, such as the CEO of Anthropic, have suggested that entry-level roles in fields like law and consulting could vanish. Yet the empirical data does not clearly support these extreme projections (Financial Times).
Why the Alarm Bells Might Be Ringing
It might seem counterintuitive: AI is powerful, advancing rapidly, yet not (yet) slashing jobs wholesale. Here are a few possible explanations:
-
Lagged impact
Technological disruptions often take years, sometimes decades, to fully ripple through the economy. The internet, computers and earlier automation waves did not displace entire sectors overnight. -
Task substitution over job substitution
Rather than eliminating jobs outright, AI is more likely to automate specific tasks within roles, letting workers shift toward more complex, non-automatable tasks. This blend of substitution and augmentation is a key dynamic in how technology tends to evolve (see related academic research: Eloundou et al., arXiv). -
Hype, signalling and pressure
Executives face pressure to embrace AI to show innovation, cut costs or appear future-facing. Sometimes layoffs or restructuring are framed as “AI disruption” when other economic pressures may be at play. -
Industries, roles and demographics vary
Some sectors and job types (especially repetitive, well-defined tasks) are more vulnerable, while others are more resilient. Younger, less experienced workers may feel more pressure, potentially amplifying inequalities.
What Happens Next?
The Yale / Brookings study is not dismissing AI’s potential disruption, it is simply saying that the transformation is not yet manifesting in broad job losses. The authors plan to monitor the data monthly as AI adoption deepens (Financial Times).
That said, there are several areas we should watch closely:
-
Entry-level and junior roles
If firms see AI as a way to reduce costs, they may hire fewer junior staff, which could choke off talent pipelines and hinder long-term organisational renewal (Financial Times). -
Skill premium shifts
As AI handles more routine work, demand may increase for cognitive, creative and relational skills, potentially widening wage gaps. -
Asymmetric effects across sectors
Some fields (for example, law, accounting, content creation) have already seen early signs of disruption, while others may remain relatively sheltered. -
Policy, regulation and social safety nets
The social impact of AI will depend in part on how governments react: retraining programmes, income support or regulation of AI deployment.
A Balanced View: Not Utopian, Not Apocalyptic
It may be tempting to swing between extremes, either heralding AI as humanity’s saviour or doom-machine. But the evidence so far suggests a more moderate reality. AI is not (yet) a job killer in aggregate; rather, it is beginning to nudge the boundaries of how work is done.
For workers, adaptability and continuous learning will matter more than ever. For organisations, the challenge is how to integrate AI responsibly and sustainably, embracing its potential while mitigating disruption. And for policymakers, the imperative is clear: stay vigilant, monitor trends, and prepare to support transitions in an ever-evolving labour ecosystem.