It’s easy to doomscroll these days. AI, it appears, is coming for our jobs. Even occupations that were previously considered an easy path to a middle-class lifestyle, like lawyer and radiologist, may be subject to the AI chopping block. Yet these stories, despite their flashy headlines, are missing nuance. They examine the seen (and likely) consequences of the AI revolution, but are missing the unseen “what comes next” part of the story. Every historical episode of creative destruction involves both creativity and destruction. Yet current news stories are focusing only on the destruction. We might not know how AI will revolutionize the American workforce, but past episodes of similar technological upheaval suggest that the future will be brighter than we can imagine.
Recent headlines are, indeed, scary. Consider the following:
- May 12, 2025: “For Silicon Valley, AI isn’t just about replacing some jobs. It’s about replacing all of them” – The Guardian
- June 18, 2025: “AI Will Replace Amazon Jobs. CEO Andy Jassy Confirms Workers’ Worst Fears.” – Barrons
- July 3, 2025: “Ford’s CEO is the latest exec to warn that AI will wipe out half of white-collar jobs” – Business Insider
- July 19, 2025: “AI will take your job in the next 18 months. Here’s your survival guide.” – Market Watch
These headlines aren’t from some alarmist blogger, sheltering in a tin-hat corner of the internet. These are from reputable news sources with large readerships. And they’re causing an artificial panic.
Consider the Amazon headline. Amazon has been an industry leader in automation, yet employment at the company has continued to grow unabated. Currently, Amazon employs more than 1.5 million people. That’s up from 17,000 in 2007, and nearly double its 2019 employment figure. This employment growth has happened despite the fact that the company currently has more than a million robots in its workplaces. The jobs those robots have replaced are primarily those involving menial work or repetitive tasks, freeing up labor for more valuable pursuits. While CEO Andy Jassy recently announced that AI will likely lead to future job cuts at the company, similar claims were made in 2012 when Amazon acquired robotics company Kiva Systems. Employment grew unabated after this acquisition.
These headlines also sound suspiciously like those circulating during a previous public conversation in which technology threatened to take all the jobs away. In the mid-1990s, the internet began to move from the plaything of tech hobbyists to a central part of work and education. Jobs that had previously been done by human processors were increasingly outsourced to data processors.
In 1995, Jeremy Rifkin published his book The End of Work, which argued that the dawn of the information technology age would create a massive and structural decline in jobs. He suggested that as many as two-thirds of all existing jobs could eventually be eliminated by machines. Jobs in manufacturing, agriculture, and clerical work were particularly vulnerable to this type of technology-based outsourcing.
To be fair, machines did take over many of those jobs. But we didn’t have massive, enduring, structural unemployment as a result. Instead, new jobs emerged.
Because I’m writing a piece on how AI won’t replace all our jobs, I asked ChatGPT to help me figure out how to identify some jobs that didn’t exist in 1990 and now have a significant number of employees. It very helpfully pointed me to the Bureau of Labor Statistics’ Occupational Employment and Wage Statistics. Here are a handful of new job categories and their current employment figures from that database:
- Software and Web Developers, Programmers, and Testers: 2,154,370 employees
- Database and Network Administrators and Architects: 633,540 employees
- Computer and Information Analysts: 677,230 employees
Indeed, the full set of “Computer and Mathematical Occupations” has exploded since internet adoption began accelerating in the late 1990s. The entire category of “Computer Occupations” currently has an employment figure of 4,786,660.
These broad categories include a range of fulfilling jobs and occupations, including app developer, social media manager, cloud architect, cybersecurity analyst, and influencer. In past eras, many of the individuals pursuing these opportunities would have been good candidates for once-stable jobs in law, accounting, or manufacturing.
In 1897, Mark Twain heard a rumor that he’d died. He sent a letter to the New York Journal to clear up the matter, stating that “the report of my death was an exaggeration.” Not only are the reports of AI’s employment “death toll” an exaggeration, but they’re missing information about the critical second act of the play. After the destruction comes the creativity, and the story of the internet can give us clues about the future of work in this technological episode as well.
READER COMMENTS
Craig
Sep 19 2025 at 10:02am
In the short to medium term I suspect you are quite right I think the fear is a bit more exitential than that, ie that AI is itself potentially an agency and if the progress is as some suspect, not all suspect this, the AI marries itself with the robots over at Boston Dynamic and these robots will be faster, stronger, smarter than you or I in literally every aspect. There’s literally no reason you would need to ask a person to do anything because there’s a robot to do it. That will be a world where humans are pampered by robots or eliminated as useless eaters, I can’t be sure which.
David Q
Sep 24 2025 at 9:09am
Craig wrote:
Hi Craig,
My assignment for you is to try to disprove the theory of comparative advantage by finding a set of overall productivity numbers where A / B > C / D but where B / A is NOT < D / C.
I tried to disprove it.
I set up a spreadsheet to make the math faster when I tried to disprove it. I tried lots of overall productivity numbers for A, B, C, and D. I couldn’t disprove it, but I found a special case. When A /B = C / D, then there was no “greater than” sign to flip.
Because of that, it’s true as can be, so long as ratios exist, that the theory of comparative advantage strongly implies that one trading partner can be worse at EVERYTHING, and yet there will be something that the highly skilled trading partner can let the unskilled trading partner do, and BOTH sides will be better off for the trade. It doesn’t make a difference if one of the trading partners is a person, a city, a country, or bees, or robots. So long as they can trade, there will be something worth having the less skilled trading partner do. (If this were not true, you would be able find a set of numbers to assign to A, B, C, and D that don’t follow the rule in my first sentence.)
So even if the Al Robots are better than us at everything, we could still search for our comparative advantage, that is, the thing that we can do “least worst” at, and trade with the robots with that.
That’s why, “There’s literally no reason you would need to ask a person to do anything because there’s a robot to do it,” need not be a big fear.
Craig
Sep 26 2025 at 10:06am
One of the better lessons indeed was learning that a person who can make 6 shirts or 3 hammers can benefit by trading with a person who can make 24 shirts or 5 hammers. I remember that well and I’d suggest in a world of general scarcity that makes a fair amount of sense, but this is a world where 300mn people will be served by 300 billion robots. The vast majority of people with a million dollar UBI will not choose to work for less than the operating cost of a robot. Everybody’s reservation wage would be so ridiculously high the fact people are at relatively or even potentially absolute disadvantages to other people won’t really matter like it did. Its a world where everybody won the lottery. If the pampered existence plays out, well, that’s ok, the job is the cost of doing something.
nobody.really
Sep 19 2025 at 11:12am
Early in the 1900s people worried that the growth of engines would leave people unemployed. It didn’t. People trained for different jobs; that’s creative destruction.
But horses weren’t so lucky. The US went from having 27.5 million horses and mules to about 3 million; the rest got sent to the glue factory. That also is creative destruction.
Draw your own conclusion–or have AI do it for you.
john hare
Sep 19 2025 at 6:53pm
I can see AI creating new opportunities. Consider a 3 hour commute today which reaches 6 hours of daily driving plus the work done. 14 hour days are tough, and I speak from experience. An autonomous vehicle may allow the same commute without the fatigue of driving. The commuter vehicle could have a bed so the commuter could get some sleep going and coming. This would open up some jobs to people that don’t have to buy million dollar plus homes to be in commuting range.
Or in my concrete construction business, automatic concrete trucks could do deliveries around the clock to automated job sites. More building per unit of time and equipment as this could apply t other trades as well.
Probably more relevant is the eventual improvement in standard of living such that I can’t imagine now. 30 years ago I didn’t have 4 wheel drive and lugged a lot of materials across sandy jobsites. 40 years ago I didn’t have cellular communications which was an expensive lack in hindsight. Cable television was barely available in my area 50 years ago. Fast food and ubiquitous food delivery wasn’t here 60 years ago.
What will we consider necessities in another half century that we don’t even know about now?
Matthias
Sep 20 2025 at 7:12am
The examples of new jobs you mentioned are good, but a bit predictable.
My favourite new job is: barrista.
The number of barristas has exploded in the last few decades. And it’s not someone anyone would have predicted the information age would produce.
Mark Barbieri
Sep 21 2025 at 8:54am
I agree with you. But let’s assume that you are wrong and that we can enjoy the same standard of living (or probably higher) with an order of magnitude less labor. Why are people afraid of that? If it happens, we may need to adjust how we allocate resources, but we’ll be much better off. But let’s not panic about jobs disappearing while unemployment remains low and median incomes continue to rise.
Comments are closed.