AI threatens 'jobs with dignity,' says Nobel-winning economist

ENDA CURRAN: There's no shortage of warnings that AI will have a significant impact on the jobs market. Do you agree with that?

SIMON JOHNSON: It's going to have a big effect on the economy and the labor market. It's hard to know exactly the form it takes. The current trajectory we're on is one in which there's going to be a lot of automation, and many good jobs will be lost. Any technical transformation also creates jobs, but what we're looking at on this path for the foreseeable future is a net loss of good jobs, jobs with dignity, jobs with good pay. There's an alternative path. So it's not a done deal, but it's looking increasingly dark.

EC: Is that mainly a near-term disruption? Do better jobs ultimately come on line?

SJ: Think about Waymo, the self-driving car company. It's pretty successful, because it's good. People enjoy it. It's got 20% market share in San Francisco and is spreading across other U.S. cities. When Uber and Lyft came along, they disrupted the taxi business; bad news for people who own medallions, bad news for some taxi drivers. But they create a lot of new jobs for other people driving passengers in urban areas. But Waymo doesn't replace one type of job with another. It replaces primarily people driving people with machines driving people. And then there's a bit of backup by people. You know, if the Waymo taxi door gets stuck, someone has to close the door, and they put out [an online bid for a gig worker] for that. And somebody gets paid 20 bucks to close the door. You're not going to get health insurance for closing Waymo doors, are you?

EC: How bad will it be?

SJ: I don't think it's mass unemployment, Enda, because we don't do that in the American economy. Not unless you have the Great Depression, which is not in the cards. What it is, is the loss of that middle again, it's the pushing of a lot of people down. The other thing that is striking is what's happening to the computer science majors. All these kids who were told since they were 4, "Learn coding, go to coding camp, build video games, become a computer science major, there'll be lots of jobs for you." Now they're being told, "No, thanks." Because Claude, or whatever AI tool is the fashion this week, is now better at coding than you are. So go get some other skills. That's for 20-year-olds and 22-year-olds. How are they going to climb up any kind of corporate, professional, vocational ladder? It's a big deal.

EC: What kind of sectors are most vulnerable to job losses?

SJ: We know that it's about tasks, right? So it's any task that has a routine element, and that looks like, in the first instance, a lot of white-collar jobs and clerical work. But we can also see real pressures on other jobs, like making TV or marketing. Somebody showed me a new product last week in the medical space. And he illustrated the point he was making with a very cute video that looked professional. He said, "Oh, I had somebody do that on AI." And then there's the Waymo-type jobs, right? 3.5 million delivery drivers is the number people usually use. How many of them are still needed? Perhaps you still need them to carry the package up the steps. That's possible. But how long until the robot can do that?

EC: Related to your Nobel-winning work on prosperity, what kind of institutions do you see being created around AI? Will any of these productivity improvements make us more prosperous as a society?

SJ: Well, I think it's a bit of a worry, to be honest. What we're getting is a lot of concentration of wealth and power, and people from that sector who don't want to be constrained in what they do. And obviously there are lots of debates about this happening in the political sphere. I don't want to paint everyone with the same brush, but to the extent that powerful people don't want regulation, just want to be allowed to get on with what they're doing, this is like the Gilded Age. That's a bit of a problem. So you'd expect a struggle over democracy and democratic control... There's nothing inevitable about these outcomes.

EC: Can you tell me about your role as British ambassador for AI?

SJ: I do like some of the stuff they're doing in the UK. Obviously, they don't have hyperscalers, but they are trying to be very active in developing applications that have safeguards, that are useful to people and that create jobs at least as much as they destroy jobs. So we'll see how they get on, and how I get on with them, and how we get on together. But I think that's a better attitude than just saying, "Oh, do it all at once. Otherwise the Chinese will eat our lunch." Well, they might. I don't think they will. I think we could be more careful.

EC: One role central banks could play in this AI boom would be to pull down interest rates in the event of a labor market shock.

SJ: Sure, we might like to cut, but what's inflation doing? How does that feed through?

EC: There's a conversation that if we're in a productivity boom driven by AI, that will keep inflation low.

SJ: From what I've seen so far, the productivity side seems a bit exaggerated. I would be careful about that. I think a lot of times it's about control. So you switch to the AI because you feel it gives you more control over the workflow. Productivity may not grow that much, but you can keep a grip on the workers.

If there's a lot of slack in the labor market, that would be lowering interest rates, but not because it's a productivity boom. More because it's a K-shaped economy [one with high earners gaining as lower earners fall behind]. And you're trying to balance the fact that AI booms and other people are literally losing good jobs. That's a pretty tough place for monetary policy to be, because the goods markets is quite tight, because of the upper part of the K [stoking demand].

EC: What about on the fiscal side? If there is a jobs shock, what do governments do? Can they go and spend money at a time when public debt is at a record level?

SJ: Well, obviously, it's all about the context, isn't it? You just put your finger on it, which is our starting place is not a particularly strong one. Fiscal policy in general is supposed to lean against the wind and make strategic investments and do it in a way that doesn't add inflationary pressure. There are lots of problems with doing that. It doesn't respond fast, and we're starting with these high levels of public debt. So it's too early to make a statement about a big macro response. I do think that if you look at what's happening in health care and education, there's room for encouraging what we call pro-worker AI. There could be some adjustment in the tax code. The tax code is a bit too favorable to machines relative to people. The major tax on lower-income workers is payroll tax, so you would need to lower that. Taxing labor at a time when you want more jobs to be created doesn't make a lot of sense.

EC: Do you think policymakers are responsive enough?

SJ: I think getting more people engaged and thinking about it is important. I think just deferring to people with a lot of money at the top of the food chain right now is generally a mistake. The thinking around the social media was "OK, let them get on with it now, and we'll fix it if there's a problem." But then later, they've already changed people's habits. Lots of people are addicted to social media. The companies are very powerful. They put money into politics. It's not that easy to fix it later. So I think there should be a debate now about how to redirect the technology and make it more pro-worker. There's going to be some automation. But it seems there's an unhealthy fixation in the corporate sector with how many jobs you can slash using AI, as opposed to what kind of top-line revenue you can build.

EC: How does this shock compare historically in terms of economic history?

SJ: Nothing's ever unique, is it? The scale feels like railways or electricity or the digital transformation. But those took decades, in part because they were analog technologies. You have to build a lot of stuff, physically build things. AI is distributed over the digital rail. When ChatGPT first came out—I think that was November 2022 — within about 24 hours, as far as I could tell at the time, talking to lots of people, it was being used in pretty much every corner of the world.

And there's some good sides to that and some bad sides. It was 35 years for electricity to have its full effects on the economy, because you have to build a new factory to get the productivity. Well, this is not a 35-year thing.

EC: On net, are you more optimistic or pessimistic about the ultimate impact?

SJ: I work at MIT. The "T" is technology. We're not Luddites destroying machines. My thinking is more like, "OK, we don't have to rerun the early British industrial revolution, because it's 250 years later." We've learned things. We've got more capabilities, and we can get better outcomes.

Enda Curran covers economics for Bloomberg News in Washington.

(Lead image credit: Aliaksandr Marko/Adobe Stock)

See also:

• AI becomes top priority for insurance execs in 2026
• AI notetaker lawsuit exposes potential business liabilities
• From buzzword to business tool: How independent agents are putting AI to work

Copyright 2026 Bloomberg. All rights reserved. This material may not be published, broadcast, rewritten, or redistributed.