How to adapt to change at work without losing your edge | Insights from economist Tyler Cowen

Most US jobs, the threat to your job is not AI. It's some other person who uses AI better than you do.

Welcome to WorkLab, the podcast from Microsoft. I'm your host, Molly Wood. On WorkLab, we talk to experts about AI and the future of work. We talk about how technology can supercharge human capabilities, how it can transform entire organizations and turn them into what we like to call frontier firms. Today we're joined by Tyler Cowen. Tyler is the Holbert L. Harris Chair of Economics at George Mason University, author and deeply influential public thinker who co-writes the Marginal Revolution blog and hosts the Conversations with Tyler podcast. Today he's going to endure being interviewed on the other side of the microphone. In the past few years, he has been tracking how generative AI is impacting the economy, labor markets, and institutions. His writing explores how technology is redefining productivity, creativity, and the skills that will matter most in the years ahead. Tyler, thanks for joining me on WorkLab.

Molly, happy to be here.

Okay, so I want to ask you first about kind of a shift from you, actually. Your 2011 book, The Great Stagnation, got you this kind of reputation as a skeptic, a technological pessimist. You characterize the U.S. as being stuck in an innovative slump. And today you're a little bit more of an energetic optimist about tech, I would say. Do you think that's fair? And if so, what flipped that switch for you?

That's fair. But keep in mind, in my 2011 book, The Great Stagnation, I predicted we would get out of the Great Stagnation in the next 20 years. And my book after that, Average Is Over, I think that's 2013, was saying AI is what's gonna pull us out. So I think my view's been pretty consistent. And I'm proud of the predictions in the earlier book. And we're finally at the point where it's happening. I would say biomedical advances and AI, which are now also meshing together, mean we're back to real progress again. It's fantastic. The funny thing is people don't always like it. They actually prefer the world of stagnation.

Ooh, say more about that.

It's more predictable. In any given year, your income may not go up as much, but you know what your job will look like for the next 10 or 20 years. And that's comforting to people. If I think of my own job, university professor, I started thinking about it when I was 14. Now I'm 63. That's almost 50 years. What I do has changed a bit, but it's the same job I thought I was getting. And that will not be true for many people in the coming generation.

Right. That is a fair point. We are gonna talk a, a fair bit about change management because as you point out, we are in a moment of extreme change. What did you see then that made you think AI would be the thing that pulled us out? And, and what do you think about where we are now?

When I was very young, I was a chess player and I saw chess-playing computers back in the 1970s and they were terrible, but I saw the pace at which they advanced. And by 1997, you have Deep Blue beating Garry Kasparov and being the strongest chess-playing entity in the world. What chess players know is how much intuition is in chess. It's not mainly calculation. So if you've seen computers advance in chess, you know they can handle intuition, which is something a lot of people at the time didn't see. But it's interesting that myself, Kasparov, who is a chess player, and Ken Rogoff, an economist and also a champion chess player, we all were very early to be bullish on AI, and it's all for the same reason, the link to chess.

I mean, that is a conversation I would like to have for an hour, uh, but won't. Let's fast forward to roughly today or around 2022, when certainly AI had existed for a long time before, but generative AI then has created this sort of new tsunami. How would you describe the economic impact now since its introduction?

Well, right now it's still very small. If you're a programmer, it's had a big impact on you. If you're good at using it, it does a lot of your work. It makes it harder for some maybe junior-level programmers to get hired. Most of the economy, I would say it saves people a lot of time. It's treated as an add-on. It's turned into leisure. Work that might have taken you 3 hours is done in 5 minutes. But in terms of GDP growth, productivity growth, we're not there yet.

Let's stay on this, this line of, of where we are and where we're going, because certainly it is the way of tech leaders to predict an immediate productivity boom from any new technology. Some economic experts predicted the same. Do you think the impact is in fact slower than predicted, or was this also part of the chess game?

It depends who's doing the prediction, but I think we need a whole new generation of firms and institutions. If you take currently existing companies or nonprofits or universities and try to get them to reorganize themselves around AI, Mostly you will fail. I mean, think back to the earlier years of the American auto industry. Toyota is a kind of threat to General Motors. Try to get General Motors to do the good things Toyota was doing. That's arguably a much simpler problem, but mostly they failed and we just didn't adjust in time. So adjusting to AI, which is a much more radical change, you need startups which will do things like supply legal services or medical diagnosis or whatever.. And over time they will cycle through and generationally become dominant firms. But that could take 20 years or more. It will happen bit by bit, piece by piece, step by step.

Yeah. This is something you've written about, that institutions, specifically these, you know, universities, big businesses, governments, are the real bottleneck to AI progress. Do you think that it is possible for a legacy institution to reinvent itself as a frontier firm, or do we have to wait for this startup cycle to play out?

I think we mostly have to wait. I think we'll be surprised that some companies will do it. They may be companies which are founder-led, where the founder can storm into the room and just say, we're going to do this, and no one can challenge that logic. But it's just not going to be that many players in the game. So I think it will be painful, hardest of all with nonprofits and especially universities and governments. You know, the IRS still gets Millions of faxes every year and uses fax machines. So you're telling the IRS organize everything around AI. How about telling the IRS use the internet more?

Yes.

We haven't even quite managed to do that. So yes, it will be slow.

Yeah. It seems like everybody sort of feels like we are in a bubble and we're headed for a crash and it's gonna come any minute now. Do you agree?

I would make this point right now, tech sector earnings are are higher than tech sector CapEx expenditures. That's good. It means they can afford to spend the money. Do I think every big or even small tech company will earn more because of this? No, I think a lot of them will lose a lot of money. It's like the railroads. We did overbuild the railroads, but the railroads didn't go away. In fact, they transformed the world. So it's not like Pets.com where one day it's gone and you say, boo-hoo, it was a bubble. AI is here to last. It's going to work. There'll be major winners, there'll be a lot of losers, a very common story. If you wanna call that a bubble because there are losers, that's fine, but it's not a bubble in the sense that when a bubble pops, it's gone. Mm-hmm.

I think you could argue that the parallel, actually the Pets.com parallel might apply now, which is that the internet didn't go anywhere. The, you know, the, this new economy didn't go anywhere. We just had a little bit of a speed bump.

And we do in fact buy our dog food at home online and it's delivered. Yeah. Not from pets.com, some other company. Yeah.

In an interview earlier this year, you suggested that colleges should dedicate at least a third of their curricula to AI literacy. Say more about how education needs to evolve to provide value in this era.

To be clear, right now we do not have the professors to teach AI literacy. So my suggestion is hypothetical. It's not something we actually can do. But over the course of the next 20 years, basically almost every job, I think even gardener, carpenter, Physical jobs. If you're in sports, San Antonio Spurs are claiming they're now, you know, the AI-ready NBA basketball team. So if all jobs will require AI, we need to be teaching and training everyone in AI skills. It's quite hard because AI now is changing so rapidly. You could learn something 6 months ago and then you've gotta learn a new thing. So it will take a lot of time and we need to teach people how to stay current. That itself changes. But what we need to do, in my opinion, is have writing be prominent in regular education at all levels. And I mean in-person, face-to-face writing, not take-home homework where GPT does it for you.

Mm-hmm.

Writing, numeracy, and AI. Those are the things you need to learn. The other things you can mostly learn from AI. If you can deal with numbers mentally, right, that means you can think well and work with the AIs. You're in great shape. We need to have further training in all kinds of specialized skills. Say you wanna be a chemist. Well, what do you need to learn about the lab? But just start off by emphasizing those 3 things: AI, writing thinking, and math.

I, you know, we talk a lot about frontier firms, but you're talking about frontier universities in a way, if you're saying that those professors simply are not there now. Exactly.

Arizona State is maybe the most forward thinking. I have some hopes for University of Austin. I think they will be trying, but you know, personnel is policy, as they say. And in universities, people feel threatened by AI. They know it will change their jobs. It doesn't have to make them worse, but I see very, very slow speeds of adaptation and adjustment. Hmm.

Adoption is just such a fundamental question. Adoption, and I think the word you just used, which is adaptation. There is this kind of mindset shift, habit shift, skills shift, even within workplaces that already have some AI collaborators or digital employees. Is that fair to say?

Absolutely fair to say. It's hard for just about any institution to make the adjustment, including major AI companies. They don't do most of their own internal tasks by AI, right? You would think, well, it should be easy for them. And, you know, I think they'll manage over time, but it's not easy for anyone.

You're saying, and there have also been a number of studies saying GenAI has had minimal impact on the labor market. But we are also seeing headlines about large-scale job cuts that are being framed as the result of embracing AI. Do you buy that framing or is there a large corporate disconnect?

I'm not persuaded by the framing. There could be some truth to it. I don't think we know how much yet. I think there's a good chance we're headed for a recession for non-AI-related reasons. And then of course people get laid off. AI is an easy thing to blame. Oh, it's not me, it's the AI. What I suspect will happen is as AI becomes more important, new hirings will slow down. I'm not sure that many people would be fired in the US. If it's a call center in the Philippines or in Hyderabad, India, there I can see people being laid off in the early stages because of AI. But most US jobs, the threat to your job is not AI. It's some other person who uses AI better than you do. And there's going to be a lot of job turnover due to that.

Do we risk making AI into the boogeyman?

We're making AI into the boogeyman right now. Just go to Twitter and you see people upset at all sorts of people, companies, AI systems. My goodness, they're so negative about one of humanity's most marvelous inventions.

Is what they're saying on Twitter or your response?

What they're saying on Twitter is so negative. There's so many villains in the story and we've done something amazing, but that's common.

People actually do enjoy stagnation and change is very scary.

That's right.

And you're talking about a realignment of business, education, and culture to try to integrate this technology. Like, is it that disruptive that that's where, that's what we have to consider?

I believe so. And you use the word scary. That's correct. I would also say it can be demoralizing. So if I ask the best AI systems questions about economics, usually their answers are better than my answers. Now, how should I feel about that? I would say I'm more delighted than upset, but I understand when people don't like it. I'm like, whoa, you know, what am I good for now? And I have answers to that question, but it requires adjustment from me.

Right. So then now put yourself into the context of somebody at work. Who is experiencing this. How do you manage for those potential feelings of inadequacy or the, the change in job style and behavior that will be necessary to adapt?

I don't think we know yet. This is a reason why I'm bullish on startups more generally. If you have a startup and you tell everyone you're hiring, well, you need to know AI, we're going to run things around AI, the morale problem is much lower. Tobi Lütke at Shopify tried to do this with his famous memo. He said, anyone we hire at Shopify from here on out has to be very good with AI. That was an excellent move, but it's hard to see that through and then actually reward the people who do a good job of it. When a lot of your people, they're probably using something like ChatGPT in their daily lives, but they haven't really adjusted to the fact that it will dominate a lot of their workday.

Let's talk about sectors for a minute. What are the, the sectors you think, in addition to maybe the types of work that are most likely to feel AI's effects most intensely in the next, let's say, year to 18 months?

Year to 18 months? I think it will take a few more years than that to really matter. So there's plenty of places like, say, law firms where AI is already super powerful, but you're not allowed to send your query to someone's foundation model out there in the Bay Area because it's confidential information. So that will work when, say, a law firm on its own hard drive owns the whole model and controls it and doesn't have to send the query and the answer anywhere. I don't see that happening in the next 18 months. I can imagine it happening in the next 3 years. But right now, I think entertainment, the arts, graphic design, Obviously programming. I think functions such as finance and customer service, they're often pretty repetitive. That to me seems more like a 2 or 3 year thing. Not so much the next year. We just don't quite seem to be there to have working agents that are sufficiently reliable that you would turn, you know, the candy store over to them right now. It'll come. I'm really quite sure less than 5 years, but not in the next year.

Let's talk about humans for a minute. In your recent writing, you have been exploring this sort of question about How AI might change what it means to be human. We've talked about that a little bit in the context of learning, but what do you think humans most need to protect or preserve?

You'll need to build out your personal network much more. Everyone now has a perfectly written cover letter, so that does not distinguish you. Who can actually vouch for you, recommend you, speak to what you've done with them or for them? That was already super important. Now it's much more important. I think your charisma, your physical presence. I hesitate to use the word looks because I don't think simply being attractive looking is necessarily the main thing, but to have looks that fit what people think your, your looks ought to be like and to be persuasive face to face again has always mattered a great deal.

Mm-hmm.

But that's just on the ascendancy. Uh, the ability and willingness to travel. And some sort of interpersonal skill that reflects knowing how you can fit into a system and when you should defer to the AI, when you should not defer to the AI, and how can you persuade the people you're working with to actually learn more about the AI. That's a pretty tricky skill. We're not used to testing for that, but we will need to test for it in the future.

Do you mean literally test for it? Like an app that—

We can test for it. Yeah. I asked Sam Altman recently in my podcast with him, well, how do you make sure you're hiring people who, when the time comes, are willing to step aside for AI and perform some other function in their jobs? I don't think any of us know how to do that yet. We're used to asking other questions.

Yeah, I mean, certainly people aren't good at doing it when a human comes along.

That's right. Maybe the AI is less threatening, right? Maybe.

You could sort of say, well, it is, it does have access to all the human knowledge in the world, so I guess. It's doing a better job than me. I mean, I guess on that note, then what will we need to let go of in addition to, sounds like a little bit of ego.

I'm a university professor. I give students comments on their papers all the time. I now give them AI comments and my own comments. The AI comments are typically better than my comments, but my comments see things the AI doesn't, like how the paper will play to certain human audiences. Maybe what journal it should be published in, something maybe temperamental about the paper that the AI is in a sense too focused on the substance. But on the substance, the AI just wipes the floor with me. So I need to invest more in learning these other skills and just skills of persuasion. How can I convince the students to take the AI seriously?

Going back to sort of amplifying our essential humanness and making that maybe part of our education. I feel like when we say something like that, it feels like such a big lift, like, oh, how will we ever retool our education system to focus on humanity and the arts and interpersonal skills? But we really did used to have that, you know, it's, it's sort of fair to say we overemphasized on technology. And I mean, you're an economics professor, but numbers and pure economics and data-driven decision-making. And, and it feels doable to me that we could sort of return to a different educational model. If we think the value will be there in the end.

It's not intrinsically hard to do at all. It's a matter of will and that the insiders in many institutions, schools included, they just don't want to change very much. Uh-huh. So those are hard nuts to crack, but just intrinsically, uh, we've done it many times in the past, as you point out, and we'll do it again. But Alec, how long and painful will the transition be? That's what worries me.

How long and painful do you think the transition will be?

I don't know, dragging everyone kicking and screaming. It's decades in my view.

Yeah.

But some places will adjust pretty quickly.

Obviously we cannot prove a negative. We can't prove an ROI that doesn't yet exist. But it seems that the fundamental question in, in terms of overcoming the kicking and screaming, if it were a toddler, you would say there's a cookie at the end. And I, I wonder how we attempt to measure the future value that makes this transition and this upheaval worthwhile?

My personal view, which is speculative to be clear, is that at some point younger people will basically not die of diseases anymore and will live to be 96 or 97 years old and die of old age. That's a huge cookie. Am I going to get that? I'm not counting on it. There's some chance, even Moneybets would say no. And you will have a life where anything you want to learn, you can learn very, very well. People will do different things with that. Some people will do cognitive offloading and actually become mentally lazier, and that will be a larger social problem. But the people who want to learn, it will just be a new paradise for them. I would say it is already. I consider myself one of those people. I'm happier. I use it to organize my travel, where I should eat, what book I should read, how to make sense of something in a book I don't know. I'll use it dozens of times every day, and it's great.

Yeah. Do you think that there's a role for government or public policy one way or the other in smoothing the economic adjustment, let's say, to widespread AI adoption?

There will be. I'm not sure we know yet what it is. The immediate priority for me is just to get more energy supply out there so we can actually have a transition. Since the transition now is so small, I'm a little wary of proposals to imagine 13 different problems in advance and solve them. I think we do better in the American system of government, often by waiting till problems come along when there's high uncertainty and then addressing them piecemeal, whether it's at federal, state, local level. So I'm skeptical of people who think they know how everything's going to go and they have the magic formula for setting it right.

I mean, I'm skeptical of that too, but that is our whole podcast, Tyler Cowen.

I know, I understand. But the role of government, government laws are not easily repealed, right?

Yeah.

If you pass something, you, you tend to stick with it. So I think we need to be a little cautious, recognize things definitely will be needed. The first imperative will be better and different national security policies, which we've done some of that already, but those are tough, complicated issues, no matter what your point of view. And then over time, well, a lot of our K through 12 schools are governmental. So they should change. But it's a big mistake to set out the formula. What you really want to do is create an environment where different states, counties, cities experiment, and over time you figure out what is best. And government schools should do it that way rather than all top-down.

Mm-hmm. It feels like that is advice that could apply to businesses too, as we try to advise leaders on how to prepare, even if this change comes from the outside in, even if it comes from those employees who are are doing everything that you just said, you know, using an LLM 50 times a day, so much so that it becomes an ingrained thought partner. How should leaders think about adapting, adopting without getting locked into something that, like you said, might change in 6 months?

It's going to change in 6 months. I think you have to teach your people how to stay current. Which is different from many other things you need to teach them. So, oh, you had to teach people Excel. It's true, Excel changes over time, but it's been a fixed skill for a while. Using a smartphone, at least up until now, basically a fixed skill. We now have to teach people something that's not a fixed skill.

What about a college grad? Or, you know, like my son just started college, so they're going to go, the ones who are leaving or just starting are in this very liminal space. Where the curricula has not changed, but they will potentially come out into or are coming out into a changed environment. Is it enough that they have, you know, maximum neuroplasticity?

They have to work much harder on their own. And there's a kind of status adjustment many people will need to make. So the previous generation, there's a sense if you're smart, your parents have some money, you went to a good high school, that there's a path for you. You can become a lawyer, go to finance, do consulting, work at McKinsey. And you will be like upper, upper middle class or maybe better if you're really talented. And I think that world is vanishing rapidly. The notion that these talented kids might have to forget about those paths and do something like work in the energy sector, which might even be for more money, and move somewhere that's not just New York or San Francisco, maybe, you know, move to Houston or Dallas or, you know, in, in energy center in Louisiana. It's gonna be a very different world, and I don't think our expectations are ready for that. So the people who have invested in the current system and, and ways of succeeding, I think it will be very difficult for them. And a lot of these upper, upper middle class parents and kids will maybe have a harder time than say maybe immigrants who came to this country when they're 8 years old. We're not quite sure what to expect and we'll, you know, grab at what the best options are wherever they are. Hmm.

A lot of status adjustment across the board, it sounds like you're saying, you know, from what your future is to the work that you do in an organization to how and who you manage.

Like in economics, it used to be if you were very good at thinking like an economist, you knew that if you worked hard, you would do well.

Uh-huh.

I suspect moving forward, the AI can do that for you and you have to be good at managing and prompting the AI. And setting up research design in a way the AI can execute. That's a very different skill. It's not the end of the world, but when it, that change happens to you in the midst of your career plans, it's very disorienting.

I just wanna put a finer point on what I hear you saying a lot, which is, and I think we know this, that energy, the energy conversation is a huge part of the AI future in a way that maybe we just have not been quite explicit enough. About?

I think we've not been very good in understanding it will need to expand a lot. We've made it very hard to build new energy infrastructure. I think that will again change over time, but probably a slow and somewhat painful process. But it's also a place a lot of new jobs will be created there. A lot. Also biomedical testing. There'll be all these new AI-generated ideas. Of course, most of them won't work. Someone has to test them. Not just people working in research labs, but just actual people in the clinical trials. So that will be a much larger sector. That's what will help us live to be 96 or 97. But again, you have to do the hard work in the meantime, and that will stretch on for decades. So people often ask me, where are the new jobs coming? In part, I just don't know. But those are two areas where to me it seems obvious there'll be a lot of new jobs and they'll be super productive.

Right. And then you're— I don't want to put you on the spot further. You're saying it's— we're early enough that we don't even yet know what other I guess cottage industries, so to speak, are going to be born as a result of this.

That's right. In which things can the AI do, in essence, all the work? And in which things will it still need a lot of human guidance? You know, that's going to be trial and error, but that will determine a lot where the new jobs go. And in part, we just don't know yet.

At the risk of asking you to tell me we don't know yet. What I'm—

which I will do.

I'm going in anyway. Exactly. I trust that you will be honest with me on this point.

But it's how I tell you we don't know that matters, right?

Exactly. That's the genius. As we look forward 3 to 5 years, I'm not gonna ask you open-ended what you think things will look like. That is unfair and nobody knows. What would be the measuring sticks? Like what kinds of things should we be measuring to truly understand the scale of economic impact or change related to AI?

This may sound counterintuitive, but I think under a lot of scenarios, the more unhappy people are, actually the better we're doing, because that means a lot of change. Now, there are scenarios where people are unhappy just because matters are terrible. Okay. But for the most part, I think we'll see this negative correlation between sentiment and actual progress, because progress is disruptive and there will be a lot of change. That's what I would look for. So I'm not going to say I don't know.

All right. Look for misery, everyone. That will tell you that it's working.

But not quite misery. It's more like confusion and disorientation and some demoralization. People, especially poorer people, I mean, already have access to free, very high-quality medical diagnosis. If they need legal advice to start a small business, they don't have to pay a lawyer. Around the world in poorer countries, this is starting to matter a great deal. There'll be so much more happiness on net. But if you just look at the complaining, As we said before, the complaining on Twitter about all these different AI-connected personas and companies, the complaining is going to go through the roof. And that's what I would measure.

That is outstanding. Tyler Cowen is a professor of economics at George Mason University, author and host of Conversations with Tyler. And for more of his insights, check out his blog, Marginal Revolution. Tyler, thank you so much.

Dolly, thank you.

That was Tyler Cowen, professor of economics at at George Mason University, author and host of Conversations with Tyler Cowen. For more of his insights, check out his blog, The Marginal Revolution. And for more conversations like this one, follow our show and find every episode at microsoft.com/worklab. I'm your host, Molly Wood. Thanks for listening. If you've got a question or comment, drop us an email at worklab@microsoft.com and check out Microsoft's Work Trend Indexes and the WorkLab Digital Public where you'll find all of our episodes alongside thoughtful stories that explore how business leaders are thriving in today's AI era. You can find all of that at microsoft.com/worklab. As for this podcast, rate us, review us, and follow us wherever you listen. It helps us a ton. The WorkLab podcast is a place for experts to share their insights and opinions. As students of the future of work, Microsoft values inputs from a diverse set of voices. That said, the opinions and findings of our guests are their own, and they may not necessarily reflect Microsoft's own research or positions. WorkLab is produced by Microsoft with Tryfilm. Sharon Callander and Matthew Duncan produce the podcast with editorial support from Jon Eilenberg. Jessica Volker is the WorkLab editor. I'm your host, Molly Wood. Thanks for listening.