Labor economist Kathryn Anne Edwards is pushing back on the Silicon Valley narrative that AI will produce a permanent class of unemployable Americans, arguing the framing itself is classist and the underlying data does not yet support catastrophe. Speaking on Platformer's miniseries on AI and jobs, Edwards pointed to US Bureau of Labor Statistics data for May showing employment in 18 AI-exposed occupations declined just 0.2% while employment rose 0.8% across the rest of the economy. She called the figure illustrative but not definitive. Edwards writes a column for Bloomberg Opinion and co-hosts the podcast Optimist Economy.
Her target is the rhetoric coming from figures including OpenAI chief executive Sam Altman, who has invoked the prospect of an 'idle class' of permanently unemployed workers displaced by AI. Edwards argues that projections in the range of 7 million to 7.5 million displaced workers should not change the policy response — and that fixating on the precise number is a way of avoiding the harder question of what to do about it.
The pattern Edwards highlights is one tech has run for at least 15 years: when a new technology arrives, firms prefer to lay off existing staff and hire younger workers fluent in the new stack rather than retrain employees, particularly those over 50. That makes attributing current layoffs at Amazon, Salesforce, and elsewhere specifically to AI difficult, because the industry already had high turnover and a documented preference for younger labor.
“I am very firmly of the belief that the US worker is an incredible being, and they don't deserve to be written off by their former employer as never being able to work again.”— Kathryn Anne Edwards, Labor economist and Bloomberg Opinion columnist
Key facts
- 01US Bureau of Labor Statistics data for May showed employment in 18 AI-exposed occupations fell 0.2% while employment rose 0.8% everywhere else.
- 02Edwards rejects Sam Altman's framing of a permanently unemployed 'idle class' as classist toward American workers.
- 03Projections of AI job loss range from 7 million to 7.5 million people, but Edwards argues the policy response should be identical at either figure.
- 04Edwards points to tech's long-standing practice — going back 15 years — of laying off workers over 50 rather than retraining them.
- 05She prescribes overhauling unemployment insurance, healthcare, relocation subsidies, and the estate tax to prepare for AI-driven displacement.
Edwards is skeptical of the 'AI-washing' of recent corporate layoff announcements. Companies citing AI as the reason for headcount cuts are often also working through pandemic-era over-hiring, macro uncertainty, and shareholder pressure that rewards the AI narrative in the stock price. All of those causes can be true simultaneously, she said, and the company itself may not know which dominates.
The skepticism extends to productivity claims. Edwards invoked the Solow paradox — economist Robert Solow's 1980s observation that 'you can see the computer age everywhere but in the productivity statistics' — to question what measurable economic output Claude, ChatGPT, and the broader crop of subscription chatbots have actually produced. Anthropic's Boris Cherny, the creator of Claude Code, told the same series that the end of software engineering as currently practiced is here, but predicted a new 'builder' profession would generate more jobs than it destroys.
Other tech leaders interviewed in the series struck similarly measured tones. Box chief executive Aaron Levie argued the 'last mile' of human labor will resist automation. Google's James Manyika emphasized that automating a task is far easier than automating a job. Edwards' position aligns with that 'AI as normal technology' frame in the near term, while diverging sharply on what policymakers should do regardless.
“It's like teenage sex: everyone's talking about it, no one's doing it.”— Kathryn Anne Edwards, Labor economist and Bloomberg Opinion columnist
On the question of whether AI is already destroying jobs, Edwards does not deny it. She says she genuinely believes AI is causing job loss now — but warns that 'at the same time' is not the same as cause and effect, and that economists looking at occupation-level data face confounding variables including interest rates, investor sentiment, and sector-specific hiring cycles. The 0.2% decline in AI-exposed occupations could reflect tighter capital conditions in venture-funded sectors as easily as it reflects automation.
Her prescription is structural rather than predictive. Edwards argues the United States is unprepared for mass unemployment from any cause, AI included, and points to tools she says are well understood: overhauling unemployment insurance, decoupling healthcare from employment, subsidizing relocation for displaced workers, and raising the estate tax. The barrier is political will, not economic knowledge. Many millions of Americans are already out of work and unsupported, she noted, a population that exceeds most projections of AI-driven job loss.
Edwards describes herself as a cynical optimist on the policy question. The bar for improvement is low because so little has been tried, which she frames as upside rather than despair. The frustration in her position is that the AI debate has consumed political oxygen that could be spent fixing problems that predate the technology by decades.
The counterweight to Edwards' view is straightforward: if frontier models continue improving at current rates and agentic systems like Claude Code keep absorbing software engineering tasks, the 0.2% decline in AI-exposed occupations could steepen quickly. Cherny's own forecast — that software engineering as currently practiced is ending — implies a sharper labor adjustment than May's data captures. Edwards' framework assumes the policy response is the same whether displacement is 7 million or 7.5 million, but it is less clear the framework holds if the figure compounds into something larger over a shorter horizon.
For the AI industry, Edwards' critique is awkward because it does not contest the technology's capability — it contests the labor narrative being built around it. The 'idle class' framing serves a specific function: it positions AI labs as the inevitable beneficiaries of a workforce transition they describe as already decided. An economist arguing that American workers are adaptable, that the data does not yet support catastrophe, and that the real problem is a safety net Congress has refused to fix removes the inevitability from the pitch. That matters for how AI companies talk to regulators, to investors selling on total-addressable-market expansion, and to the workers whose jobs are the implicit collateral in the agentic-AI roadmap.
Working on something we should cover, or seeing a story we missed? Send leads, documents, or feedback to hello@aichatdaily.com. For sensitive tips, see our secure tips page for Signal and PGP options.
Spotted an error? Email hello@aichatdaily.com with the URL and the issue, or read our full corrections policy.
