MIT Professor: How to AI proof the job market
Paul Osterman, professor emeritus at MIT Sloan and author of "Disposable Workers", argues that the bigger effect of AI on work is not mass job loss but the spread of contractors, freelancers and other marginal workers. He talks to Autria Godfrey, Laila Rizvi and Stephen Horn about who wins and loses, what companies owe displaced workers and how to help older workers who lose their jobs.
Transcript
It's the AI threat that we just can't escape, that artificial intelligence will take our jobs.
Do you think the fears around job apocalypse and this permanent underclass forming are overblown?
Firms don't want to have real employees. They want to have contractors or freelancers or what I call marginal workers.
Is becoming a disposable workforce the real downfall of the AI uprising?
You will see people not getting a job because of AI and innovation. And I wonder how the social contract extends to people who aren't in the workforce.
There's a practical argument, which is if you don't want your country to blow up and you don't want to see widespread discontent, you have to protect people.
Paul Osterman, one of the leading scholars researching employment and labor markets. He's Professor Emeritus at MIT Sloan, and in his new book, "Disposable Workers," he argues that companies have spent decades quietly trading loyalty for flexibility, building a workforce that's easier to hire, easier to cut, and easier to forget. So today we are asking him a harder question than whether AI will create or destroy jobs overall. Instead, we want to know who benefits most from this technology, what will happen to those whose work becomes less secure, less valuable, or less protected.
All right, Stephen, I want to get started with you before we talk with Paul and ask you, do you think there will actually be clearly defined winners and losers when it comes to the AI revolution?
So that's a great question. So what I think we are going to see, right, is huge change. We'll see seismic change in the job market like we already have over about the last 50 years or so, but this will just accelerate it. But what's quite interesting right at the moment, if you ask me, is the reality is nobody knows. Nobody knows who the winners are, nobody knows who the losers are. What we do know is we'll get both. Laila.
I agree with that. And I think it's crazy that there's this discourse in Silicon Valley where, oh, you're going to be part of a permanent underclass if, you know, you don't take action within the next five years. And I think the thing that's different with AI is that everyone kind of has it in their pocket. It's not that it's just these big companies that have these large robots or, like, computers, that they didn't start off as personal computers, that everyone has access to a free ChatGPT subscription. So that I think changes the dynamics of it as well.
That's true.
All right, we're going to see what Paul has to say.
Big central question for you, do you think there will be clearly cut, clearly defined winners and losers when it comes to the AI revolution?
Yes, obviously there will be in the sense that if you were an accountant adding up numbers and subtracting numbers, AI is likely to replace your job. If you're me doing a research paper and some academic journal, AI is going to help me. It's going to help me with search. It's going to help me with calculations. It's going to correct my spelling errors. So yeah, there will be winners and losers. What's not at all clear is what the aggregate outcome is and what is not all clear is whether we'll ever really know the aggregate outcome. What you need to know is that in the United States at least, employment is still growing. It's been growing month after month after month, and so there's no evidence of an AI having an aggregate negative impact.
So to add onto that, Paul, do you think the fears around job apocalypse and this permanent underclass forming are overblown?
Let me just bring it back to the argument I make in the book, which is that there's this trend towards disposable workers. Firms don't want to have real employees, they want to have contractors or freelancers or what I call marginal workers. As AI introduces more uncertainty, not just for workers, but for firms about their staffing levels and their balance, that uncertainty is going to lead them to want to have more disposable workers so that they're hedging their bets. From my point of view, that's the real impact of AI. The long-term impact could be quite different, but in the short term or the medium term, it's going to lead to disposability.
But when you look at, you know, surveys that Microsoft do or Bain do, they actually say do just that to companies, you know, concentrate your own staff, if you like, on what you do well and then use, you know, more flexible, more freelance workers to kind of do the rest. So aren't we sort of in that loop because it increases productivity?
Well, whether it increases productivity is a different question. What you just articulated with what the jargon is kind of core competency, and that became a popular business phrase or slogan maybe 20 years ago. And in fact, if you look at companies that at least articulated that they adopted core competency as their strategy, there's no evidence that it led to better performance in the aggregate. Almost none of these business-type buzzwords or slogans or fashion of the day are correlated with performance over time. So that's wrong.
The attitude that companies do have, which is a problem, not just about AI, but about everything, including disposable workers, is that they don't really respect their frontline workforce. McKinsey issued a report a few years ago in which they said that 95% of a firm's value is produced by 5% of the workforce. Whether we trust McKinsey research or not is a different question, but that's their attitude. And so this kind of contempt or lack of respect for frontline workers is a challenge.
But I think in white-collar work, I guess the value that you bring is your expertise. So I don't think you necessarily get the respect for just having an education or showing up. It's about the value your expertise brings. And especially with AI, a lot of people are going to be stripped of that, and it doesn't feel good and you feel disrespected and it's going to be scarring, but that is the reality of the market.
It depends on what you mean by white-collar work. So I use myself as a positive example, but if you're a doctor, that's a white-collar job. AI is going to help you. It's not going to tell you what to do with the patient. It's going to give you a pretty tight range of possibilities or options, and it's going to summarize the literature for you. You can't fully trust it, but it's going to be helpful. And then there's a whole range of white-collar jobs which basically involve interaction with people. And that, by the way, includes managers. A good manager is someone who knows how to interact with his or her, you know, subordinates and teams. AI can't do that for you. I'm not kind of cheerleading here for AI, but I'm saying we need to have kind of a less hysterical perspective on it.
So Paul, I want to ask you something that the Brookings Institute just put out, and Molly Kinder, who led their research on AI and work, said that really this won't be the jobs apocalypse that everyone is predicting, and instead there will just be a hard, messy period of pain where young people are trying to make their way into the workforce. She has suggested that we can turn some of these entry-level jobs into more, like, residencies like young doctors. Do you think that that's a viable solution?
Not everyone needs a college degree. And so you can imagine a variety of training mechanisms or skill development. The classic one is apprenticeship program. So what you just described is just the rewording of the idea of apprentices, maybe extending it to other occupations, or short-term higher education. Not every college degree needs to be four years. And again the United States, community colleges, which are typically shorter-term job training or two-year degree programs, have become increasingly popular. So yeah, we need to have a more tolerance for a wider range of kind of entry strategies. I think that is fair.
But again, I think the deeper question, at least the question I try to argue about in this book, is we need to rethink how firms approach their workforce strategies. We need to kind of put some floors on job quality. We need to put some constraints on disposability. AI is part of a larger development in terms of what firms are doing and not doing, and it needs to be put into that context too.
Sure. I mean, is there any kind of argument on the other side of this that actually says to some extent that workers, or some workers at least, are going to be empowered because the firms may not need them all the time to do things, but they will need them, they will need their expertise, and that expertise actually becomes more valuable?
I think that's right. And I think, again, if you look at the data that I work with, think about freelancers. Freelancers can be very high-end people. By the way, I'm not talking here about the person who repairs your gutters in your house or did the index for my book. I'm talking about people who are high-end freelancers. They get work because they have skills that are needed for a short period of time and companies use them and then they move on to another job. Job satisfaction among those freelancers is quite high. The only problem is it's not scalable in the sense that in America today, according to my data, about 5% of the workforce are freelancers the way I just described it.
All this kind of business school buzz about pack your own parachute, everyone's a freelancer now, that's just hot air. You know, the labor force is not going to be a freelance labor force, right? But yes, for those folks, it is empowering and they do have power. That's correct. But it's a subset of the workforce.
Good. Yeah.
Right, okay, well, so for those that are at risk of becoming collateral damage of an AI takeover, well, then what? Coming up right here after the break, Paul answers the question of what becomes of those deemed the losers in a post-AI world. But in the meantime, we want to hear from you. Has your industry seen a move toward a disposable workforce? How is that impacting productivity, morale, career satisfaction? Weigh in while we take a quick break. Welcome back. We're continuing our conversation with MIT's Paul Osterman. And so if we're accepting the notion that AI will eventually create a clear distinction between winners and losers, should anyone intervene to try to provide a safety net? And if so, who?
Is that a question?
It is, to you. That's the big question. Do you think that these companies should bear some type of responsibility for some of these workers that are potentially going to be outsourced because of AI?
Yeah, so you're asking, should there be a social contract? Should there be some understanding that we're all in it together? You could make a theoretical argument that the answer is yes, and that is that people have invested their human capital in firms. They've learned firm-specific skills about their organization. They've learned how to do their job. They've made an investment. And just like firms want to get a return on their investment for machinery, people are entitled to that investment. If you take away their job, you're taking away the returns to that investment, and we owe them something. So, that's a theoretical argument.
There's a practical argument, which is, if you don't want your country to blow up, and you don't want to see widespread discontent and extreme politics, you have to protect people. Your country and my country have always had that view. Whether we've acted on it is a different question.
And then there's the kind of hard facts of life, which is that firms are really about just doing one thing, which is maximizing profits. There's no use in pretending otherwise. That's just a fact of life. That's not bad. That impulse leads to innovation, it leads to new products, but it also leads to squeezing the workforce, and you need to put constraints on that. So when you add all that up, yes, there's a strong case to be made for having a social contract and enforcing it.
The point that you made at the beginning was a really interesting one, when you were saying that actually if you look at employment in the US or the UK, that employment is actually going up, is not going down. And so, what we're seeing at the moment, arguably, is not a lot of people losing their jobs because of AI and innovation, but arguably, you will see people not getting a job because of AI and innovation. I wonder how the social contract extends to people who aren't in the workforce?
Well, it definitely should extend to it, to those folks, and we can talk about what we should be doing for them. Think about the other shocks to the labor market. The internet was a shock. Mainframe computers were a shock. In the United States, the China trade was a big shock. And many people gained, but a lot of people lost. And employment levels have consistently gone up after all those shocks, and I don't see why that shouldn't be true in the case of AI. It's a question of making sure that the winners are generous towards the losers. We, at least in the United States, have not had a good history of that, and that's what we really need to turn around.
That's what I was thinking about. Do you think that the company should bear some sort of responsibility when it comes to, rather than replacing their workers, upskilling their workers so that they can utilize AI and retain their jobs?
Yes, we've also required companies to have responsibility. I'm not going to defend our healthcare system, but we do require that companies support health insurance. We do require that companies contribute to Social Security, or retirement income. We do require that companies have some minimum standard of safety on their job. We do require that companies, in principle, prevent racial or gender-based discrimination of various kinds. So, we've always required companies to have responsibility. The politics have shifted against actually taking that seriously, but it's not a big new revolutionary idea. It's just a question of defining the responsibility and enforcing it.
Paul, I wanted to go back on your point about potential widespread discontent, and I wanted to look at the flip side to that, where, do you think AI could be a widespread equalizer, because everyone kind of has access to free plans for ChatGPT? And Elon Musk has predicted that we'll have universal high income, Jeff Bezos thinks that there'll be too much work for everyone to do, and Sam Altman has gone back on the job apocalypse scenario. So, do you think that will actually happen, or are they making it out to be this utopia which isn't that straightforward?
I think what these people are saying is basically politics and PR. They're trying to find a way to frame AI in a way that reduces the amount of concern that people have and reduces the opposition to new data centers and so on. As far as Elon Musk being optimistic about jobs, do recall that two or three years ago he was on a stage with his chainsaw bragging about how many jobs he had killed.
That's true.
And so, I wouldn't take those statements too seriously. I'd look at the reality of what's happening inside of companies. And it's really too soon to know exactly what'll happen. I think it's clear there'll be winners and the losers. I think, on balance, there will be more jobs. Jobs, we're not going to see the job apocalypse. I can't even pronounce the word, but we're not going to see it, whatever that is.
We need to think about how to put floors under job quality and floors under risk in the labor market. I do want to just push on this point, that my story about disposable workers, that's going to intensify with AI, largely because of the uncertainty that AI introduces.
But do you think the concern exists because the people who are kind of winning now are the ones that might become the losers, and the people who are actually on the back foot because of resource constraint, like those who don't have access to education, might actually win more, because they have access to AI? It's the fact that that's actually going to flip, potentially.
Other people will benefit. But among the people who benefit are people in kind of service jobs that require interaction with people, other people. Healthcare is an example. It's a high-end service job, but there's actually a lot of low-end healthcare work. The person who comes to your home or your mother's home to take care of her when she's very old and needs help? That's a service worker. And that job will never be replaced by AI. It may be augmented by AI, but it's not going to be replaced by AI. Your doctor is not going to be replaced by AI. People in human interaction jobs, high-end and low-end, I think will benefit. That, I think, we can be pretty confident.
So two quick follow-ups for you. You said we really don't know the full outcome of what the aggregate result will be of how AI will impact the workforce. If we were looking at some sort of a timeline, when would you predict we might really start feeling that outcome?
The only way I can judge that is, look at the internet. The internet changed all of our lives in radical ways, and helped some of us and hurt some of us. Throughout that period of the introduction of the internet, employment continued to grow. There were winners and losers, but the aggregate employment continued to grow. So, I think it's a process. We're still using the internet and still changing some of our work in various ways. It's the internet that's connecting machines and factories. So, I don't think there's a date associated with it. I think it's a process.
Okay, and finally, what you mentioned kind of at the top of today's show when you were mentioning somebody like an accountant who's probably most at risk of being fully replaced by AI, for those people, what can they possibly do to not be labeled as one of the AI losers?
Right, so there we do get into a question about who you can train and who you can't train and retrain. I mean, stereotypically, a 30-year-old person not that far out of school can be trained. And we have experience with effective training programs in this country. A 50-year-old person, 30 years out of school who's been doing their job, his or her job, for decades? I think we need to find ways to get that person into a job, because work is not just a source of money, it's a source of dignity and a source of self-worth. But that job may not pay what their prior job paid, and we can find ways to top that up. And there's experience with that.
A number of European countries have what are called short-work programs in which, if someone is losing their job, they're kept on part-time, and their income is topped-up by public funds. Or public service employment is another model of finding good work for people that can't find work otherwise. So, we need to be more creative, but it's a question of political will, really. These are not easy questions, but they're not impossible questions, if the politics are right.
All right, we've got a couple of rapid-fire questions that we ask all of our guests. Stephen, you want to get us started?
The question I always ask is, when you look to the future, when you can see an answer to the ups and downs of AI, what are you optimistic about? We've heard a lot about what we need to worry about. But what are you optimistic about?
I think AI will help a lot of people do their jobs better. It helps me do my job better. It helps my doctor do his or her job better. A lot of people's work will be improved. The quality of their work will be improved by AI.
And my question is, what's something that's universally accepted in your field of expertise that you actually disagree with?
I think that there's a lot of complaints about the negative effects of labor market regulation. But to address the kind of questions that we're talking about here, and kind of reduce the risks, and help people who are losers, I think we need to expand in a smart way, or raise the floor at the bottom of the labor market, would be a better way of putting it.
And finally, where do you think people will draw the line when it comes to bringing AI into their personal lives?
Into their personal lives?
Mm-hmm.
Somebody's going to tell me to stop looking at my computer all the time. So, you know, that's possible. You know, no one wants, you know, the government in their bedroom, obviously. No one's going to want Claude in their bedroom. So, you know, there'll be a line. One thing we haven't also talked about here is something else that's happening in the labor market. One is my story about disposability. But the other is the labor supply. You know, people are going to retire in a lot of countries, including mine and including yours. Those folks are going to need to be replaced, and what's happening around immigration is going to be a real problem for companies.
All right. That's a whole different episode. Paul, thank you so much. Such a pleasure to have you with us today. We appreciate your time.
It is.
All right, so Paul actually wasn't as, I think, pessimistic as I was expecting when it comes to this notion of the disposable workforce. Am I on an island on that? He says essentially, you know, that for some people who want to be freelancers, who want to be contract workers, it's actually a very good, you know, gig.
He said what Stephen said at the beginning, which was that-
You don't say that very often.
I know. It doesn't happen often that our guest agrees with Stephen, that they get it right. That we don't know, as in it's uncertain.
Right.
And I don't think it's clear, whereas I think so many people come at it with such certainty, including the tech bros, that everyone's, you know, going to lose their jobs, you're not going to have anything to do, it's going to be horrific, AI is going to rule the world. But who knows?
Right.
I think it's interesting, because I think Laila's right, and Paul said the same thing, about, you know, we don't know yet, but clearly at some point we will.
For me, there's a huge elephant in the room, right, it all sounds lovely in theory, but he's talking about how we put floors underneath this. So at the same time, he says companies have got one thing to do, which is to make money. So who's building the floors, right? And there isn't even a conversation around that. There's no government regulatory, you know, conversation about how to deal with that, whilst at the same time we're talking about winners and losers in the future. So, until we start having that discussion about what this looks like and how you support that, how is that going to happen?
Yeah, and I also think it's not even been three and a half years since ChatGPT first came out. That's nothing. And you think governments are going to get their act together? We can't even build train tracks in 20 years, as in, there is no chance.
Right.
In the first Industrial Revolution, arguably it took 140 years before those floors were put in, in terms of workers' rights. Now, clearly we're not expecting it to be 140 years now, but it's not going to be 14 months, right? And that whole process has to happen. And he's right when he talked about political consensus, and you show me anywhere where there's political consensus, right?
I also think the 140 years, it's going to take double that time, given how government acts now.
How slow they are. Yeah.
Yeah, as in both in the UK and the US, nothing, nothing is going to get done. It was quicker back in the Industrial Revolution. So, good luck. And that's not coming from a doomer perspective. There's no chance. There's absolutely no chance, unless you tear everything up and start again.
Right.
I just thought it was encouraging that he did highlight the fact that, you know, yes, some people will lose their jobs, and I guess be considered the AI losers. But for a lot of people, it's just going to be a new way of working, and their jobs will just be augmented by AI, rather than completely replaced. So, it gives us a little something to hold onto, I guess.
All right, well, now that you have heard from our expert, we want to hear from you. Is AI already separating the haves from the have-nots? What are you seeing in your field? Weigh in below, and then be sure to like and subscribe. Thanks for joining us today. We'll see you right back here next week for an all-new episode of "Agents of Tech."