The Biggest Job Creation Event Since the Internet Is Hiding in Anthropic’s Data
92 million jobs are about to disappear. But 170 million will replace them. In fact, 69 percent of CEOs say their AI investments will lead to maintaining or increasing headcount.
NOTE: I have long maintained that AI can’t replace you. But depending on what you do, it may very well replace your job.
That sounds like a contradiction. It isn’t. To quote Charafeddine Mouzouni, we’re entering a world where AI replaces tasks… but makes human accountability more valuable than ever.
To put that another way, AI has made me about five times more productive than I was two years ago. What it hasn’t done is replace me. It can’t. What I add is not just piecework, or even information: I add judgment. I’m creative in a way the software cannot be, because though it can execute, I must provide vision and direction. And at the end, I sign my name to the work, just as I would if I were assisted by human employees.
And in fact, now that I have these AI tools, I need to hire more humans to use them!
When farm automation began, we faced the same challenge. Legions of sharecroppers could be replaced with a few tractors. So where did the sharecroppers go? They learned new skills and got higher-paying jobs in tractor plants. Their kids went to college and became doctors and professors. The old world ended, yes. It was replaced by a far better one.
That’s happening again. There will be disruption. There will also be opportunity — for everyone — to a degree it’s difficult even to imagine today. — RDM
The Biggest Job Creation Event Since the Internet Is Hiding in Anthropic’s Data
by Marco Kotrotsos
April 23, 2026
So last week, Anthropic published its labor market impact study and the internet did what the internet does. Headlines about a “Great Recession for white-collar workers.” Lists of the ten most exposed occupations. Think pieces about whether your CS degree was a waste of money. I wrote one of those pieces myself. It’s easy to get caught up in the flow of opinion if you don’t stop and really think about the other side of the messaging.
There’s a thing about the Anthropic data that got buried under all the alarm: the same numbers that measure displacement also measure opportunity. And the opportunity side of the ledger is potentially enormous, if you know how to read it.
The 61-Point Gap Is Not Just a Threat
Anthropic’s key finding was the distance between what AI can theoretically do and what people are actually using it for. In computer and math occupations, theoretical capability sits at 94 percent. Observed usage is 33 percent. That’s a 61-percentage-point gap.
The doom reading: that gap is closing, and when it does, jobs disappear.
The other reading: that gap represents trillions of dollars in unrealized productivity. And closing it requires people. Lots of them, doing work that mostly didn’t exist three years ago.
Think about what the gap actually consists of. It’s not that organizations looked at AI and said “no thanks.” It’s that deploying AI at scale requires integration work, compliance frameworks, quality assurance pipelines, change management, training programs, security audits, and a hundred other things that someone has to build, maintain, and operate.
And every percentage point of that gap that closes creates work before it eliminates it.
We’ve Seen This Movie Before
In 1970, the United States had roughly 250,000 bank tellers. Then ATMs arrived. By the early 2000s, there were over 400,000 ATMs installed across the country. The number of tellers per branch dropped from 20 to 13.
And the total number of bank tellers roughly doubled.
What happened? ATMs made it cheaper to operate a branch. So banks opened more branches. Urban bank branches increased 43 percent. The tellers who remained did different work, more relationship-oriented, more advisory, less cash-counting. The technology didn’t eliminate the category. It transformed it and expanded it.
Spreadsheets tell the same story. When VisiCalc launched in 1979, there were 299,000 accountants and bookkeepers in the U.S. Excel arrived in 1985. By 1989, that number had grown to 524,000, a 75 percent increase in a decade. Today there are 1.28 million, and the profession is still growing at 4 percent per year. Spreadsheets didn’t kill accounting. They made it so much more accessible and powerful that demand for accountants exploded.
Electronic discovery software was supposed to eliminate paralegals. Paralegal employment grew. Scanning technology was supposed to eliminate cashiers. Cashier numbers increased. Every previous abstraction layer in software, from compilers to frameworks to Stack Overflow, was supposed to eliminate developers. Developer headcount grew through all of it.
The pattern is so consistent that economists have a name for it. The Jevons Paradox: when technology makes a resource cheaper to use, total consumption of that resource tends to increase, not decrease. William Stanley Jevons observed this with coal in 1865. It applies to knowledge work in 2026.
