Bloomberg Economics put a number on the top end of AI's labor-market effect: as many as 380 million workers worldwide whose jobs are significantly disrupted.
Most estimates of AI and employment start from the occupation. Tom Orlik's team started from the task, broke every job in the economy into its component tasks, and counted only the jobs where more than half of them are things a model can do.
"Are we going to see a wave of white collar redundancies as Claude and other models put us out of a job?"
Orlik is the chief economist at Bloomberg Economics and published the research the day this aired, covering the US and a range of other countries.
The full segment is covered here so you can skip it.
Here are the 6 takeaways that matter.
👤 Guest: Tom Orlik, Chief Economist at Bloomberg Economics, speaking from Washington
🎙️ Hosts: Jess Minton and Alexis Christophorus, anchoring Bloomberg Businessweek Daily in place of its regular hosts
📰 Published: 14 September 2026 on the Bloomberg Businessweek Daily podcast
🟣 Apple Podcasts | 🔗 Episode page | ⏱️ length not available
Key Takeaways
The headline figure is a ceiling, not a forecast: 380 million workers worldwide at the top end of what AI could do
It counts jobs where more than half the tasks are ones a model can perform
The exposure runs the opposite way to the last two disruptions, which hit blue-collar work
Accountants, coders, traders and economists are the exposed group this time
There is no visible displacement in the data yet, and in some places AI is adding jobs
Model-building, model-training and data-center construction are all hiring
A report from Anthropic puts possible US unemployment above 10% as knowledge work is displaced
The news about graduates and the experience of graduates do not match
Orlik's example is a computer-engineering graduate reading that he has no future while his inbox fills with interviews
The risk he names is not that AI fails but that it repeats the dot-com sequence: capital in, valuations up, profits late, crash, then the real thing
1. The 380M Number
Orlik framed the research as an attempt to settle which of two histories AI is going to rhyme with — the one where a technology makes workers more productive and better paid, or the one where it replaces them.
The choice he set up is between a productivity boom and a repeat of factory automation, which he said put a lot of blue-collar workers out of a job in the 1990s and 2000s
"Are we going to see a wave of white collar redundancies as Claude and other models put us out of a job?" he asked
He was explicit that the work does not produce an answer, only a range: "We don't have definitive answers, right? This technology is still evolving. But what we do have is a kind of a range for what the estimate might be."
The top of that range is the number the segment was built on: "At the top end, if we look at AI hitting its full potential, we're looking at hundreds of millions of workers, perhaps as many as 380 million workers worldwide, who are going to see their jobs significantly impacted, significantly disrupted by this new technology."
2. Jobs Split Into Tasks
Jess Minton asked what "impacted" actually means — using AI at work, or being replaced by it — and Orlik answered by describing the method rather than the outcome.
The unit of analysis is the task, not the occupation: each job is broken into the set of tasks it contains, and each task is sorted into what AI could do and what it could not
The threshold is explicit, and it is what makes the total countable: "And for jobs where more than 50% of the tasks, more than 50% of what's done in that job, AI could have a significant impact. Well, those are the ones which we include in our calculation."
Adding that count across every country in the study is where the 380 million comes from, he said
3. Coder Out, Hairdresser Safe
Orlik used two stylized examples at opposite ends of the spectrum to show what the task test does to a real occupation.
The coder is the clearest case of a job already being reshaped: "Think about a computer coder. Well, most of the tasks a computer coder does, AI can do. So computer coders are one of the jobs which has already been pretty significantly disrupted by AI."
The hairdresser is the control: "Could AI help with the bookings? Sure, a little bit. Could it help with keeping up with latest trends, latest styles? Yeah, a little bit. But most of it, AI is not going to be able to do, right? So that hairdresser, their job's pretty secure."
Minton's reply was that everyone listening was now going to retrain as a hairstylist, and she asked what other sectors come out better
Orlik's answer inverts the last thirty years. Blue-collar workers, he said, took two shocks — robots arriving on the factory floor, and then what he called the rise of the rest, meaning China, globalization and outsourcing — while white-collar workers stayed secure
"The AI revolution is gonna be hitting the white-collar workers hard, right, you're an accountant, you're a computer coder, you're a trader, you're an economist, unfortunately, AI is going to bring some pretty significant disruption," he said, putting his own profession on the list
"If you're in the blue collar world, if you're working in a factory, if you're working in, I guess what you might call the sort of the human touch services professions, you're a hairdresser, you're a sports coach, well, AI ain't coming for you. That's where the greatest security is gonna be."
4. AI Is Hiring, Not Firing
Asked how someone acquires AI skills — whether that means a science or engineering degree — Orlik went instead to what the labor-market data currently shows, which is close to the opposite of the forecast.
He gave the bear case its strongest outside citation: "Anthropic just put out a report saying that potentially unemployment in the United States could move above 10% as AI displaces knowledge workers."
The evidence today does not support it yet: "But right now, if we look at the impact of AI on the labor market, well, first, it's difficult to see much evidence of significant displacement, not many people losing their jobs because of AI right now."
The current effect is additive, and he named three sources of demand: people at the very top of the skill range to build the models, sector specialists hired to train them, and construction workers to build the data centers he called the brains of the AI universe
The specialists being hired to train models are, on his account, training their own replacements — the demand exists to make people in a sector smarter now and perhaps to replace some of those workers in the long run
The timing he put on the turn is three or four years out, and he framed it symmetrically: more productivity if you want to be positive, more unemployment if you want to be negative
5. The Grad Who Gets Emails
Alexis Christophorus raised the entry-level question — graduates who studied computer programming or finance and are struggling to find work because employers are using AI for junior tasks — and Orlik answered with an anecdote he was careful to label as one.
He said he had asked a friend graduating in computer engineering over the weekend whether he was worried, and the answer described two contradictory realities at once
The graduate told him that "everything I read in the newspapers, everything I read online tells me I have no career. I have no future. Everything I see in my email as I apply for jobs, as I apply for internships, tells me I'm going to get a job next week."
Orlik would not generalize from it: "Certainly a lot of news out there about young grads finding it harder to get that first step on the jobs ladder. Are we seeing that so much in the data yet? Well, something to pay close attention to."
6. The Dot-Com Pattern Again
The last question was about an AI boom and bust, and Orlik answered it with the history of earlier general-purpose technologies rather than with a market view.
The pattern he described repeats: "So if we think about the history of technology, game-changing technologies like railways, like the internet, what we see very often is a pattern where there's huge enthusiasm for the new technology, capital piles in, valuations go sky high." Profits, he added, are not so quick to materialize, and then comes a moment of pessimism and a crash
The crash is not the end of the story, which is the part he stressed: "We saw that most vividly with the dot-com boom bust cycle in the late 90s, early 2000s, right? The market roared, the market crashed. Did that mean it was the end of the dot-com story? No, in some ways it was just the beginning." The internet revolution, he said, was very much the story of the 2000s
His concern is the order of events rather than the destination — that the boom and bust in the market arrives before the broad application that makes everyone more productive
"And with valuations for AI companies in the U.S. extremely high, well, some people think that boom-bust cycle is actually something that we're looking at right now," he said
Bonus Insights
Minton opened the hour with the day's tape off the Bloomberg terminal: the basket of Magnificent Seven stocks was up four-tenths of a percent, which she called a surprise given the session
On a year-to-date basis she put the Nasdaq 100 up about 16% and the Magnificent Seven basket up only about 6% to 7%, and attributed the gap to gains concentrated in memory and other chipmakers rather than the largest names
Orlik joined from Washington and the report was published the same day the segment aired
Orlik's bottom line is that AI's labor-market damage is a few years out rather than visible now, that it lands on the white-collar occupations the last two waves of disruption left alone, and that the more immediate risk is a market cycle that runs ahead of the technology's actual usefulness.
Products, Companies & Tools Mentioned
Bloomberg Economics (Orlik's team, and the source of the task-based model behind the 380 million figure)
Anthropic and Claude (The report Orlik cited putting possible US unemployment above 10%, and the model he named when asking whether white-collar work gets automated)
Books & Resources Mentioned
Bloomberg Economics' report on AI and jobs (Published the day of the segment; the source of the 380 million estimate and the task-level method behind it)
Anthropic's report on AI and US unemployment (Cited by Orlik as the case that unemployment could move above 10% as knowledge workers are displaced)
If this was worth your time, send it to someone closer to the industry than you are.
Get the latest market chatter as it happens: