Has the Great Winnowing commenced?
Is AI-driven unemployment (and underemployment) being underestimated?
I had dinner with the CEO of a tech company last week. He said, “We just cut almost 20% of our workers, and chances are we’re going to be doing it again in the next year or two. There are a lot of efficiencies we are getting with AI. At the same time, my daughter is in college and she’s looking for a job. She’s trying to avoid tech because she sees what I’m doing. I’m not sure what her classmates are going to do for jobs in a few years.”
Andrew Yang, Andrew Yang Newsletter, 4/8/25
New claims for unemployment insurance have risen. Companies that once complained about dire labor shortages are more easily filling jobs, and hiring has slowed sharply. Wage gains have slowed… The new dynamism the U.S. economy gained in the wake of the pandemic was wholly unexpected. The danger now is that it could go away.
Justin Lahart, WSJ, 4/8/25
Goodbye, $165,000 Tech Jobs. Student Coders Seek Work at Chipotle.
NYT headline, 10/8/25
For the last eight months, I’ve been attempting to warn people, Paul Revere-style, of the AI-driven, mass-automation tsunami that’s about to hammer the labour market, or at least the white-collar worker part of it.
Has anybody paid attention?
Well, let’s just say my impact on the public debate has thus far been minimal.
So when I shared the Andrew Yang quote above in a Substack note late last week, I never expected anything to come of it. I’ve posted many notes in recent months. While some got a handful of likes or reshares, none went viral.
The Yang one did. At the time of writing, it’s notched up over 140,000 impressions, which must be at least 100,000 more impressions than anything else I’ve posted on Substack over the last three years.
The post has also been enthusiastically liked, restacked, emailed and commented upon. I’ve gained 27 new subscribers, mostly Americans. I’m now hoping, Steve Irwin-like, to become a prophet with honour in my own country after first cracking the US market.
If I were more craven, I’d currently be developing a ‘How to 20X your Substack readership overnight!’ online course and attempting to make out like a bandit.
Unfortunately, I don’t think it’s within my capabilities to reliably 20X my audience, let alone anybody else’s.
Nonetheless, I have been thinking about why so many people were so taken with my note.
Playing armchair psychologist, I suspect it boils down to it being a gentle warning from a figure most people regard as amiable but competent.
Yang is a jolly, self-deprecating Taiwanese-American who comes across as warm and non-threatening. But he’s also a successful tech entrepreneur turned presidential candidate, lending him some gravitas. Finally, his message came in a form that was easily digestible to a mass audience.
Yang didn’t scream, ‘WAKE UP, SHEEPLE! AI IS SETTING IN MOTION REVOLUTIONARY ECONOMIC, POLITICAL, SOCIAL CHANGE AND, ALMOST CERTAINLY, THE END OF THE WORLD!’
(That’s certainly been my approach for the last eight months. But given recent developments, I’m starting to question its efficacy.)
In contrast, Yang merely noted that a tech company bigwig of his acquaintance was using AI to (relatively) slowly but steadily whittle down his workforce. I imagine Substack note browsers found this oddly reassuring because it accords with their own experiences and also suggests there will be a somewhat gentle transition.
Most white-collar employees probably haven’t seen their own employer make swingeing cuts yet. But they may well have seen colleagues – a customer service representative here, a technical writer there – be automated out of a job in recent months.
However, if Yang had simply noted a business-owner acquaintance was automating away some jobs, I don’t imagine the note would have struck a nerve.
The businessman expressing concern for the next generation – including, presumably, his daughter – was what sealed the deal by subtly introducing some uncomfortable but unavoidable questions.
What exactly are the young people meant to study to maximise their chances of securing employment?
Regardless of what credentials they attain, what prospects will they have by 2030 if business owners continue to automate away a fifth of their workforce every year?
What if even the well-connected, expensively educated children of CEOs are to be rendered jobless by the Great Automation?
Which brings me to my main point – I fear the Great Labour Market winnowing is already upon us.
How rubbery are the unemployment figures?
If you have the unmitigated gall to suggest the growing availability of impressive artificial intelligence is likely to reduce the demand for the significantly more expensive and troublesome human variety, you’ll attract the ire of AI sceptics.
These sceptics will point out that unemployment rates in technologically advanced nations have been at historic lows in recent years, thus proving that AI short-term impacts are being overhyped, and suggesting AI is unlikely to be a wrecking ball over the longer term.
Hmmmm.
Let’s start with what unemployment rates measure.
Metrics differ from nation to nation, but individuals usually only count as unemployed if they are (a) actively looking for work and (b) haven’t worked at all for at least a week. Crucially, those who’ve given up looking for a job might be deemed as ‘out of the labour force’, but they aren’t counted as unemployed.
While those metrics were reasonable enough in 1975, I’m not sure how useful they are for 2025.
Half a century ago, you either had a job or you didn’t. You were either working five days a week at the coal mine, or car factory, or butcher’s shop, or you weren’t. There was very little gig economy to speak of.
Things are more fluid nowadays. About half of all young people now go on to tertiary education and don’t start looking for a proper job until their early to mid-twenties. If they struggle to find an entry-level position – as many reportedly now are – they’ll often live with their parents and may not show up in the unemployment statistics.
At the other end of the age scale, I imagine there are a lot of older workers reluctantly heading into early retirement. Or perhaps doing some ‘consulting’, which generates enough money to prevent them from being counted as unemployed but not enough to allow them to maintain their previous lifestyle.
To be fair, NEETs (Not in Employment, Education and Training) were around long before ChatGPT arrived in late 2022. But I’d be surprised if AI isn’t now further swelling their ranks. A NEET is a young (16-24) person who doesn’t do anything. Even the industrious Japanese and Germans have a NEET issue (i.e. one in 20 youngsters is wasting away), and the Brits are currently freaking out about 13% of their young ’uns being NEETs.
Also, the gig economy now allows lots of people to live in a limbo land between being employed and unemployed. Back in May, my fellow Substacker Shawn K wrote about getting laid off from his prestigious, well-remunerated, software engineer job and applying, unsuccessfully, for 750-plus coding roles while supporting himself by working for DoorDash. (His post, which everybody should read, is titled ‘The Great Displacement Is Already Well Underway’.) I presume the Shawn Ks of the world aren’t showing up in the unemployment data either.
Indeed, I suspect many white-collar workers who are no longer in the labour force will be unable or unwilling to register as unemployed for quite a time.
Working the refs
But let’s put the above quibbles to one side and assume that the unemployment data is an accurate reflection of labour market realities.
While it’s true labour markets have been tight since Covidtide, they are now clearly becoming looser. Granted, unemployment rates remain relatively low, but the direction of travel is worrying.
What’s even more disturbing is how both politicians and employers are now acting.
Let’s assume that current upticks in unemployment, especially youth unemployment, were no more than statistical noise and nobody expected the upward trend to continue.
In such a scenario, you wouldn’t expect political leaders to be strong-arming government statisticians. After all, if they didn’t expect rising unemployment to be an ongoing and potentially serious issue, they wouldn’t be incentivised to cook the books in such a heavy-handed manner.
Likewise, if employers were worried about ongoing labour shortages, it seems unlikely they would be gleefully issuing RTO mandates.*
Signs of things to come
Le Monde reported back in April that Chinese economists were coming under pressure to produce, shall we say, optimistic forecasts. Subsequently, the Wall Street Journal reported
Land sales measures, foreign investment data and unemployment indicators have gone dark in recent years… China’s National Bureau of Statistics stopped publishing some numbers related to unemployment in urban areas in recent years.
Back in mid-2023, when the official youth unemployment rate topped 20%, Chinese authorities said they needed to review their calculations. Half a year later, when new data were released, the youth unemployment rate had miraculously fallen to below 15%.
Strangely, despite this encouraging turnaround, there still seems to be plenty of cynicism or outright despair among China’s young people, with a growing number choosing to “lie flat” or “let it rot” rather than strive.**
Trump recently pulled a Xi, calling for unemployment data that wasn’t “rigged” after July employment figures suggested a cooling labour market. The President immediately had the Bureau of Labor Statistics Commissioner dismissed and posted, “No one can be that wrong? We need accurate Jobs Numbers…She will be replaced with someone much more competent and qualified.”
Let’s conclude by observing the difference between how employers were behaving a couple of years ago and how they are acting now.
Dear reader, I give you this vibe-shifting email from AT&T CEO John Stankey
We run a dynamic, customer-facing business, tackling large-scale, challenging initiatives. If the requirements dictated by this dynamic do not align to your personal desires, you have every right to find a career opportunity that is suitable to your aspirations and needs… if a self-directed, virtual, or hybrid work schedule is essential for you to manage your career aspirations and life challenges, you will have a difficult time aligning your priorities with those of the company and the culture we aim to establish.
Given my day job, I speak fluent corporatese, so let me translate: The party is very much over. If you’re not going to be in the office grinding at least five days a week, fuck off and be a hippy somewhere else.
Like I keep saying to readerships of varying sizes, winter is coming.
*After the dark years of widespread remote work, it appears long-frustrated corporate peacocks are salivating at the prospect of getting back to five-day-a-week, on-site posturing and office politicking.
**Chinese young people are exercised about not being able to get into the housing market, or land the kind of job their qualifications led them to expect, or generally live the (post-1979) ‘Chinese Dream’ of comfortable middle-class prosperity. It’s such a relief that we don’t have these problems in the West’s well-functioning capitalist democracies.



Most corporate jobs are fake anyway. Good riddance!
You’re shrinking my topic coverage because you’re doing such good job covering AI! All that’s left for me to explore is the adaptation angle. 😂