The Curse of AI and Automation
Disappearing Displaced Workers
Where Do Displaced Workers Actually Go?
I’ve been digging into something that’s been nagging at me for a while. The last couple of times I streamed on Twitch I discussed AI and robotics and it got me thinking about how it’s all affecting workers. And what the hell happens to the displaced workers? I mean they have to go somewhere and they’re not all being re-employed 1:1. We hear about this displacement, by AI presently, but what I don’t hear, what nobody seems to be actually asking, is, where do these people go? Like, what actually happens to them?
So I went and researched it. And what I found is both pretty clear and, honestly, a bit unsettling. The results weren’t necessarily surprising (they kinda were, though), but the pattern has been playing out for over a century and we’re still not talking about it honestly. That was the most surprising part.
Historically Telephone Operators
I want to start here because this is the most granular data we’ve got on large-scale worker displacement, and it’s nearly a century old. Which, when you think about it, says something on its own.
Between 1920 and 1940, AT&T automated telephone switching. 350,000 operators—mostly young women—had their occupation systematically eliminated, city by city. And researchers actually tracked what happened to those workers using census data. So we have real numbers on where they ended up.
A decade after displacement:
13% became secretaries
11% waitresses
9% stenographers
8% receptionists
5% nurses
3% beauticians
20% moved into other sales and service work
Overall, they were 28% more likely to be doing clerical work and 43% more likely to be in service occupations than comparable women who hadn’t been displaced.
These workers took a wage hit of roughly 30% below what they would have earned if they’d stayed as operators (which remember no longer existed).
The telephone companies actually tried to help; something you’d never see in this day and age. They went to local businesses and said, look, we’ve got a bunch of displaced workers, can you find them something? And businesses did—but the jobs that materialised were waitressing, reception, clerical work. Not equivalent, nor promotions.
That’s what the historical record shows when a major occupation gets systematically eliminated en mass. Workers move, sure, but they move sideways and down, rarely up.
What’s Happening Now
I want to be clear about something before going further. I’m not relying on other people’s projections for this. I’m working from documented job losses and extrapolating from the historical pattern. So when I cite numbers, they’re actual data—not “X million jobs will be at risk by 2030” type stuff. Those headlines are agenda-driven and meant to be click-baity. I don’t get money for writing what I write, I don’t give a shit about its click-baitiness. I’ll also freely make up words. So there.
Below is what’s actually documented across five sectors over the last 15-20 years.
Manufacturing
1.7 million jobs lost over the past 20 years. Of those, somewhere between 420,000 and 756,000 are directly attributable to industrial robotics in the 1990s and 2000s. These factories were able to function with far fewer people. Robot density in global manufacturing has more than doubled in seven years, sitting at 162 units per 10,000 workers on average globally. Some countries are way further ahead—South Korea is at 1,012 per 10,000, Japan at 1,562.
The narrative says displaced workers “move into technician roles maintaining the robots.” The math doesn’t doesn’t even math. You’d have to be delusional to think there’s a 1:1 ratio of replacements, in reality 50 workers get replaced with maybe one or two technicians.
Warehousing and logistics
Highly automated warehouses now operate with 25% fewer workers than manual facilities. Amazon has deployed over a million robots globally. By the end of 2026, around 4.7 million commercial warehouse robots will be installed across more than 50,000 facilities worldwide. The turnover rate in warehousing was already over 40% annually before automation started accelerating. Workers were already leaving the sector in droves and automation is hurrying it all along.
Where are those workers going? No idea, because the data doesn’t exist, not that I could find, anyway.
Retail
180,000 documented job losses in the recent period, mostly driven by self-checkout expansion. Interestingly, the displacement is actually slowing down a bit because self-checkout hit proper operational problems. Theft spiked biggly, 15% of self-checkout users admit to stealing, which is pretty bonkers—that’s one out of every 6/7 (sigh, inadvertant 6-7 joke...) people. So the displacement is happening, but it’s messier and slower than the headlines suggested. It probably seemed messier and more chaotic in the past than what we see now, because we have a clearer snapshot with all the data. Hindsight bias, and all that.
Call centres
700,000 jobs outsourced between 2000 and 2005 alone—before AI automation was even in the picture. Those jobs moved to the Philippines and India. Now AI agents are displacing the workers who replaced those workers. Double displacement (whoa, double-rainbow!). Where did the original domestic call centre workers go?
Knowledge work
This is the newest front. Paralegals and legal assistants are seeing lower employment demand as firms adopt Large Language Models (henceforth to be LLMs) for document review. Data entry, back-office processing, accounting support—all declining. The Bureau of Labor Statistics has started incorporating AI impacts into employment projections and the picture for legal support staff isn’t good. What’s being created instead? AI-related roles? But 77% of those require a master’s degree, apparently. The pipeline from displaced paralegal to AI specialist doesn’t exist for most people.
The One Sector Where We Have Real Transition Data
Tech is the only sector where we have decent numbers on where displaced workers actually ended up, and it’s worth looking at closely.
In 2025, 245,000 tech workers were laid off globally, roughly 70% from US-based companies. Of those, around 55,000 layoffs explicitly cited AI automation in company filings. By Q1 2026, 23% of tech layoffs cited AI as the reason—up from 14% in Q4 2025. Acceleration!
Where did they land? About 41% found roles at other tech companies—but typically at junior or mid-level despite them being senior previously. 23% moved into fintech or adjacent sectors. 18% went into management consulting. 15% into non-tech corporate functions like project management.
Only 56.4% of displaced tech workers who found new roles secured higher pay than before—down from 60.8% just one quarter earlier. 27.3% accepted pay cuts. Workers over 40 had longer job searches—5.2 months on average versus 2.8 months for under-40s—and saw steeper salary declines.
The telephone operator parallel is almost exact. Workers moved, yeah, and most of them moved sideways or down. Only a minority moved up. The ones who moved up tended to be younger, more credentialed, with skills that mapped directly to growing sectors.
The Unemployment Rate Problem
I was a bit thrown when I started digging into the broader picture.
Official unemployment—the number you see reported—stayed relatively flat across most of the period we’re looking at:
2024: 4.1%
2025: 4.2-4.5%
April 2026: 4.3%
Looks fine, right? Nothing to see here.
But there’s a broader measure called U6 that includes discouraged workers, marginally attached workers, and people who are working part-time but want full-time work. In 2024, U6 was 7.8%—nearly double the official rate. That’s an additional 8-10 million people not reflected in the headline number.
And then there’s the labour force participation rate. In 2000, 67.3% of the working-age population was either employed or actively looking for work. By April 2026, that figure is 62.5%. A 4.7% age point drop. Roughly 7-8 million people who simply stopped being counted.
You only count as “unemployed” if you’re actively looking for work. Stop looking, and you’re gone from the statistic. A warehouse worker displaced at 45, unable to find equivalent work, eventually stops looking—maybe takes early retirement, maybe claims disability, maybe just gives up. They’re no longer unemployed by official definition. They’re out of the labour force. Invisible to the headline number.
At the same time, involuntary part-time work increased by 40% between 2000 and 2023. Gig work now accounts for 16% of total employment. Someone who used to work 40 hours a week as a warehouse picker at $20 an hour, now doing 15 hours across three gig platforms averaging $12 an hour? Officially employed but statistically indistinguishable from someone thriving.
I’m not saying governments are deliberately hiding the full picture—both things can be true at once. The displacement is real and documented, and the unemployment rate isn’t the right tool to measure it. The numbers are murkier than the official narrative suggests. What I am saying is that the unemployment rate staying flat during a period of documented large-scale displacement tells you that workers aren’t becoming unemployed, they’re simply becoming invisible.
Real wage growth for median workers from 2000 to 2020 was 0.3% annually. Flat as a board. For the bottom 10%, real wages declined. The productivity gains from all this automation went somewhere—and it wasn’t to the workers who got displaced.
Applying the Pattern Forward
So where does this leave us?
The historical pattern—telephone operators, textile workers, carriage makers—is consistent. Displaced workers move into sectors with low barriers to entry, chronic labour shortages, and below-average wages. They move because they need income, they’ve gotta eat. The pipeline into better work doesn’t exist for most.
Apply that to a warehouse worker displaced today. Adjacent roles such as transport and delivery coordination often require certification they don’t have. Retail pays less. Food service pays less. Gig delivery is flexible but earns less per hour once vehicle and other costs come into it. The historical parallel says this person moves into one of those roles, takes a 20-30% wage cut, and doesn’t move back up. They’re now competing in those absorbing sectors alongside other displaced workers, which puts additional downward pressure on wages in sectors that were already at the low end.
Apply it to a call centre agent displaced by AI automation (previously off-shoring). In-person customer service, retail, hospitality—all available and pay less. The telephone operator became a waitress or a secretary. The call centre agent becomes a barista or a retail floor worker.
Apply it to a paralegal. Document review tasks handled by an LLM. The adjacent roles are administrative assistant, office manager, court clerk. All below their previous wage and status. LLMs replaced them with less, and the firm hired fewer support staff overall.
The “new jobs created by automation” argument is technically true in a narrow sense. AI engineers, robotics maintenance techs, data scientists—these roles exist and they pay well. But they require credentials most displaced workers don’t have and can’t easily acquire. The new jobs don’t absorb the displaced workers. They employ a different, smaller, more credentialed group of people and the gap between these two groups has been widening for 20 years—technically the last 100 if we want to include tech advancements, the pattern is the same.
The Data Gap
The most striking thing I found in all this research is what isn’t there.
We have detailed data on how many jobs were lost. We have almost no systematic data on where those workers actually ended up—except in tech, where some surveys exist, and in the telephone operator case, which is a century old.
The telephone operator study tracked individual workers across a decade using census records. It’s more granular than anything available for current displacement. We haven’t built better tracking for the current wave. The labour force participation decline, the U6 gap, the involuntary part-time increase—these tell us something is happening but not exactly what happened to specific workers.
The pattern from the historical data is consistent, the current data we do have (tech sector transitions, wage compression numbers) aligns with it, and the aggregate signals (participation rate, U6, gig economy growth) are consistent with large numbers of displaced workers moving into lower-quality employment or out of formal employment entirely.
The “upskilling pipeline” that companies and governments talk about—the idea that displaced workers will retrain and move into better roles—doesn’t appear in the data at scale. Only 11% of displaced warehouse workers accessed formal retraining programs. 77% of new AI jobs require a master’s degree. The pipeline exists in press releases more than in practice.
Workers move. But they almost never move up.
That’s all for now.
As always,
Good luck,
Stay safe,
and Be well.
See ya!
If you liked this and like inconsistent, hap-hazard posts about whatever is on my mind at a given moment, then you should:
One day I’ll set up a consistency pipeline.
Sharing is caring!
Let me know what you think, and