The Cubicle Belt: The front line of A.I. displacement
The men who made the steel that built America found out they were obsolete on a Monday.
September 19, 1977, to be precise.
A morning announcement from Youngstown Sheet & Tube in Youngstown, Ohio: Five thousand jobs, terminated, effective immediately.
The locals called it Black Monday.
Within a decade, the Mahoning Valley would lose 50,000 steel and steel-related jobs.
The population of Youngstown would halve.
Today, nearly 40% of the people who remain live below the federal poverty line.
The pattern repeated up and down the Great Lakes.
Pittsburgh manufacturing employment peaked in the 1950s: 382,000 jobs.
Today: About 83,000. A 78% decline.
General Motors in Flint, Michigan, employed 80,000 people there at the peak.
Today? Fewer than 7,500.
Detroit became the only American city ever to grow past one million residents and then contract back below it.
The Packard alone employed up to 40,000 people.
It went from producing beautiful machines like this one I saw on the street this week:
To becoming one of the most photographed ruins in Detroit:
This American industrial decline took thirty years.
It happened slowly enough that the country had time to give it a name (the Rust Belt), make movies about it, study it, debate it, retrain its workers (badly), and then, eventually, mostly look away.
The defining image was that of a closed factory gate.
Today, the gate is closing again.
But it’s not in automotive factories and steel plants.
And it’s happening about four times as fast.
The canary in the Cubicle Belt
“I was literally digging my own grave.” — Geoff McGehee, 54, recently laid off from Sears Home Services, where he integrated AI into the customer-service operation that subsequently replaced him.
The Wall Street Journal published a piece this week headlined 'Phoenix Built an Empire of Cubicle Jobs. AI Is Coming to Tear It Down'.
The numbers in it deserve a closer look than they will probably get.
Phoenix's customer-service workforce peaked at 92,970 in 2021.
By 2025, it had fallen to 68,930.
That’s a 26% drop in four years.
It took the entire 1980s to deliver a hit of that scale to Youngstown steel.
Sixteen and a half million Americans work in office and administrative-support jobs.
That number is down from 18 million at the end of 2019.
The US Bureau of Labor Statistics projects another 4% decline over the next eight years — the steepest fall of any major employment category it tracks.
Phoenix is the canary. Or, more precisely, the cubicle is.
The story of how Phoenix became the back-office capital of America is not accidental.
American Express opened its first regional Phoenix office in the 1960s.
Long-distance phone lines and early computer networks meant payroll, claims processing and customer-service work no longer had to sit next to the executives who depended on it.
Phoenix had cheap land, low taxes, no unions, sunshine, and an airport.
The 1980s classified pages of the Arizona Republic were stuffed with ads for insurance-claims adjusters, customer-service supervisors, associate auditors.
By 2019, for the first time in its history, the Phoenix metropolitan area had more customer-service representatives than manufacturing workers.
The Cubicle Belt was the replacement economy for the Rust Belt. It absorbed the workers, the children of workers, and the American promise that you could move up in life without a college degree.
Now it, too, is going.
The quiet collapse of digital manual labour
The Rust Belt happened in public. Factories closed on specific dates. Workers stood outside locked gates with picket signs.
But the Cubicle Belt is going quietly. There is no Black Monday for call centers.
Companies aren't closing the buildings. They just aren’t hiring anymore.
Stanford's Canaries in the Coal Mine study, published last year, found that employment for 22-to-25-year-olds in the most AI-exposed occupations has fallen 13% since late 2022. For young software developers specifically, the drop is closer to 20%.
Wages, importantly, did not fall. There just aren’t any new jobs.
The new pattern is attrition. Not gates closing — doors just no longer open.
The Phoenix piece in the Journal describes this exactly: ‘Mass layoffs are rarely the cause. Instead, companies have taken advantage of the industry's high churn, cutting head count by not replacing workers who quit or were fired.’
You will not see this on the evening news. You will see it in the quarterly headcount line of a 10-K filing, three years from now, when someone notices the trend.
Manufacturing decline took 30 years.
David Autor, the MIT economist whose research named the original China Shock, has been explicit about the difference:
‘The greatest similarity is that this could happen quickly in certain areas, in certain activities, and it'll be extremely disruptive and scarring for the people who lose that work.’
Already, the medical-transcription profession has effectively disappeared.
Translation work has collapsed.
Tufts University's American AI Jobs Risk Index, published in March, projects 9.3 million US jobs vulnerable to displacement within two to five years, with a plausible upper range of 19.5 million (nearly the population of Australia).
Phoenix is listed by name in the Tufts data among the metro areas projected to lose between $10 billion and $70 billion in annual household income.
For context: the entire Youngstown economy in 1977 was a rounding error against this.
The retraining myth
When the Rust Belt collapsed, Washington's policy response was the Trade Adjustment Assistance program — federal retraining, income support, job placement. People have been studying this for decades. The evidence is unambiguous; it largely did not work.
The most rigorous evaluation (Mathematica, 2012) found that TAA participants had lower employment rates than the comparison group of laid-off workers for three years running.
Research by Autor and Gordon Hanson found that most trade-displaced workers ended up on Social Security disability rather than in new careers.
Jamie Dimon, the CEO of JPMorgan Chase, said the quiet part out loud earlier this year. On stage with Dario Amodei — the CEO of Anthropic, one of the leading AI labs and the company whose technology is automating the very work Dimon's bank employs hundreds of thousands of people to do — Dimon admitted that the government's promise to reskill manufacturing workers, decades ago, “didn't work”.
This is the same template now being readied for white-collar workers.
State retraining programs. Federal grants. AI literacy bootcamps.
The historical hit rate of ‘retrain the displaced worker’ is poor.
There is no reason to believe it improves when the displaced worker is 54, has 30 years of customer service experience, and is now competing with a 22-year-old AI literacy graduate for the same shrinking pool of work.
From coal mines to codebases, to where?
In December 2019, Joe Biden told a New Hampshire crowd that included coal miners: “Anybody who can go down 3,000 feet in a mine can sure as hell learn to program as well.”
It was widely mocked at the time, of course. But it captured the policy template perfectly: When an industry collapses, point its workers toward the rising one.
Today, the rising one of 2019 — entry-level coding — is the most AI-exposed occupation.
According to Anthropic, these are the most exposed:
The miners were told to learn to code.
So what do you tell the coder?
The Cubicle Belt probably will not collapse on a Monday.
It will fade out quarter by quarter in headcount disclosures, in regional unemployment data revised and then forgotten, in the slow disappearance of the entry-level rung that turned high-school graduates into call-centre supervisors into sales associates into something resembling a middle class.
Phoenix developers are already tearing down vacated call centers.
Where the cubicles used to be, they are pouring concrete for data centers — the physical infrastructure that runs the AI replacing the workers who used to sit there.
This week's quote:
"However much we may welcome the fruits of advancing technology… no one dare overlook or deny the fact that many individuals will suffer personal, mental, and physical hardships as the adjustments go forward."
— Joint Economic Committee of Congress,
Automation and Technological Change, 1955
Invest in knowledge,
Thom
The Benchmark
P.S. If you’ve read this far, thank you for your time! I put a fair bit of my own into writing The Benchmark each week, and some readers recently have been kind enough to share their appreciation. You're welcome to connect with me on LinkedIn where I post about the stories and insights in the newsletter.
Read more: 28 years later: Australia's capital gains gambit
Share: Forward this email to someone you know would appreciate it.
New here? Subscribe to The Benchmark.
Enjoyed this issue?
Subscribe to The Benchmark
Weekly insights on markets, investing, and portfolio strategy.