Bill Kerr: Due to advances in automation, over the next 12 years as much as 14 percent of the world's workforce may need to find new jobs or learn new skills. That's according to a recent report from the McKinsey Global Institute. New technologies will also create new jobs and those could surpass the ones lost to automation. But the jobs will be different. Millions of people in the United States and abroad will need retraining. McKinsey says, “Many countries will need to invest in training workers on a huge scale, rivaling the project that rebuilt Europe after World War II, known as the Marshall Plan.”

Welcome to the Managing the Future of Work Podcast from Harvard Business School. I'm your host, Professor Bill Kerr. I'm joined today by McKinsey partner, Michael Chui, who is one of the authors of the report “Jobs Lost, Jobs Gained.” Welcome, Michael.

Michael Chui: It's great to be here. Thanks for having me on.

Kerr: The report is rather hefty. It's 150 pages long. Maybe we can boil it down a little bit. Can you tell us the types of professions that are going to most likely benefit from this and those that will suffer?

Chui: Well, we're developing robots to be able to read our long reports, but in terms of which professions or which occupations are going to be affected, one of our key findings is actually all occupations are going to be affected, because we looked at, not only occupations, but the individual activities within [an] occupation. And what we discovered was all kinds of occupations, from the lower wage frontline occupations all the way through executives, almost every occupation has a significant percentage, and of that tens of percentage of their activities or time could be automated by adapting technologies, which exist today.

That said, there are certain types of activities which have the highest technical propensity for automation. So those are physical activities in predictable environments. Occupations that have more of those types of activities, so somebody who works on an assembly line, for instance; parts of agriculture; even parts of health care are the sort of physical activities in a room or an assembly line or a factory, et cetera ... So those have high propensity to be affected. Also collecting data and processing data, and that's a pretty interesting set of activities, because some of those activities are performed by people in a back office of a financial institution, people in an HR function, just processing transactions.

But what's interesting is we pay MBAs to do a lot of collecting data and processing data. We pay a lot of JDs to do that. We pay PhDs to do that. You know, we-

Kerr: It's getting a little sensitive here.

Chui: Well, yeah, I was about to say management consultants, but then I was going to leave that one off, yeah, so it's real. But then again, when we talk with senior executives, at first they're surprised. Then they say, “Yes, please give me my 20% to 30% of that time back. I can repurpose that.” I think that's a powerful way to think about the potential.

Kerr: Michael, as the U.S. transitioned from an agricultural economy, many people had to move to cities in search of work. How would you think about the population that's going to be most heavily affected by these upcoming changes?

Chui: One of the concerns that we discovered, or rediscovered, as we were doing this research is we know that people will have to be redeployed from some of the things they're doing now to the jobs of the future, and we think there's enough demand for work in the future. That said, we can't minimize the pain, the difficulty, the challenge, when someone loses their job. We know that that's important in order to redeploy labor, but historically, one of the things that's allowed the labor force to be redeployed is not only moving jobs, but also moving geographically.

One of the things that we know here in the United States is that labor mobility is at multi-decades low. There are lots of reasons for it. Not all of it we can explain, but now we have some people who are underwater on their real estate and their home. We have dual incomes, so it's harder to move more working members of the family, so those are challenges. That's going to retard our ability to be able to adjust, even as people are displaced partly by these technologies, to the other jobs that might be there. You know, even just personal identity, there are roles and occupations that a lot of people view as gendered, and somebody who's in a typically male occupation might think, “It's strange for me to think about being a nurse,” but because of aging, we do see increases in the demand for nursing and nurse's assistants, et cetera. And so I think there's some challenges there that we'll need to work through.

Kerr: As we look at the differences across countries, are there going to be some countries that will benefit more from this, some countries that have a darker future?

Chui: Well, I'll say a couple things. I mean, number one, all countries ought to be able to benefit from the use of these technologies, because all countries need to accelerate their productivity growth. Because of aging, basically half of the sources of economic growth over the next 50 years are about to disappear, so we need to accelerate productivity growth. These technologies have the possibility of doing that.

That said, certain countries are likely to have more of these technologies take hold sooner. So two factors that are important: One is countries that tend to have higher wage rates, that makes the business case more compelling, but also those that have more predictable physical activities, more of these collecting data and processing data, so depending on the mix of sectors and activities, some of those are likely to have more automation sooner.

Kerr: During the last year, so 2017, you released two reports on automation and artificial intelligence. Was it planned to have two reports, or did one lead to the other?

Chui: We do hear these questions that are coming about — such as, will there be jobs for people to do, and what are the impacts of these technologies? — really starting to pick up, at least in the awareness of senior executives, and so, collectively, those things together did make the first report that we published about the potential impact of automation to be a topic that we wanted to cover, and then this question about whether or not there'll be additional jobs.

Kerr: To go back to your statement about the awareness, is it that the awareness has changed, or are we at an inflection point for the technology?

Chui: People say, “Why is AI hot now?” Certainly a few trends on the technical side have come together to make some of these advances possible. We have more data than we ever did before. Thanks to Moore's Law, we have more computing power than we ever did before. Now, when you add more data and more computing power, we do start to see machines not only be able to equal human capability, but exceed human capability on things that we're surprised that they're able to do.

You know we probably would've thought it'd be another 10 years before a computer could beat the world champion in the game of Go, this strategy game, or we would've thought that the ability to have a self-driving car would be one of the hardest things possible, maybe a decade or so off. And we're starting to see these things happen already. So I think that's the other thing that's happening, which is now we see surprising technical results, and that catches people's attention. Then, you know, another funny thing, from those of us who've been practitioners, we always used to have this joke that we'd call artificial intelligence the thing that hadn't happened yet, right? The thing that was always 10 years away.

Kerr: Just right around the corner.

Chui: It's always going to be 10 years away, 10 years away, right? And then, by the time something starts working, you don't call it artificial intelligence anymore. You call it dishwasher. It was just something that works. Now, some of the stuff that works, we call artificial intelligence, so there's been sort of a funny linguistic change, as well. But certainly, I think, having studied a number of different technologies as they've moved up the hype cycle, I think we are at a little bit of an inflection point, with regard to senior executives saying the words and asking, “What does this mean for our business?” And that's where we're trying to inform some of the dialogue.

Kerr: Is the difference now that we are automating cognitive tasks versus manual ones?

Chui: I don't think it's only that. I mean, the honest truth is spreadsheets, for instance, used to be a physical artifact. You'd have this giant thing. You'd send it to your accountant. They would go work on it. It would come back the next day, and you'd be able to look at it and do a what-if, right? And now, with spreadsheets on computers, you can do that many times in an hour. So I don't think it's the first time that we've actually started to automate cognitive tasks, but again, it's a little bit about the surprise, the fact that now machines can read lips better than people, right? I mean that's-

Kerr: That's a little scary.

Chui: Yeah, it's-

Kerr: There's a lot of equipment in this room.

Chui: No, exactly. It evokes that sort of emotional response, "Oh, that's scary, right? Oh, that's a surprise."

Kerr: We talked about computers' ability to beat humans at Go coming 10 years earlier, but then we've also used the word hype. Do we ever predict things are going to happen tomorrow, and it ends up taking two, three decades before they realize, or maybe they have a sense of something that everyone believes is imminent, but in fact it's going to take a while?

Chui: Yeah, I think we get this all the time, and we do this all the time. My job's not to predict the future, but to talk about potential. But you know when you look back into the past, you look at predictions. First of all, it's always important to ask, “Is this time different?” Compared to what? One of my grad school professors had a textbook. The title was “Compared to What?” And so, again, are we adopting technologies any faster than we were before, because there's this trope, right? Things are accelerating. Things are accelerating. And in some sense, we are, right? The number of people who are affected by a technology, by gross numbers of people, has been increasing over time. Look at a country like China or India or Indonesia.

In that sense, things might be happening faster. But from a percentage basis, one of the things that we found in our research is that adoption of technologies, particularly within enterprises, hasn't really been accelerated that much when you talk about percentages from the start of availability all the way through to the plateau in adoption. So you know I think there's a lot of nuance to whether or not things are different when you look at them.

Kerr: This is a key point in your report is to say that the rate at which we make a technology advance can be different from its adoption. So we can see these scary YouTube clips or inspiring YouTube clips of robots doing backflips and all that kind of stuff, but it's going to be a long time until it comes to the neighborhood near you. Talk about that lag. What does the technology lag look like and what shapes it?

Chui: Yeah, I mean, to your original point [the] things that we saw on The Jetsons or Star Trek, I mean, now we do have Star Trek, right? I have a little computer on my hand, and all that sort of thing. On the other hand, I don't have a flying car yet, although, you know, there's some really interesting companies working on that kind of stuff, too. But you're absolutely right. Just the invention of a technology... The time between the invention of a technology and the time that it actually gets used in the economy, I think a lot of people just don't think about it.

I don't want to say that people are ignorant about it, but I think it's just not considered, right? I mean, I saw a quote once from a venture capitalist, who said, “We're going to have self-driving trucks, and in three years, there'll be no more tractor trailers.” And as you might have seen in our report, each tractor trailer costs $160,000 or something like that. If you have 2 million of them in the United States, that's $300 billion worth of rolling stock. It takes real time for things to be adopted.

So I mean the factors include the following: You need to actually not only develop the individual technology pieces, but integrate them into solutions. That takes real time, right? I mean, those in the technology industry know that takes time and takes capital. Even after you develop the solution, it's usually fairly expensive when you start off. For instance, if you're automating a human activity, you now have to net that out against the cost of human labor. And then often there are other benefits as well, but that takes some time, because the cost will come down, and that makes sense. Then, even when there's a positive business case, you know again, in our research it takes eight to 28 years between commercial availability, plateau, and adoption. That's one to three decades.

So it takes real time and I think that gives us time to adapt as we adopt. But we also talk about it being fast in micro, slow in macro, so even if it takes decades in total for an economy or a sector to adopt, if you're the company that has to compete against another company that's using this technology, or if you're the individual worker who's affected by this, it'll happen fast for you. And so again, there's a little bit of a dichotomy between how long does it take in macro versus how fast might it affect you as a company or an individual?

Kerr: Is this going to hurt the future of small business, or favor the multinational?

Chui: It does affect both of them, and there are these sort of countervailing factors, because, as you know, large multinational companies have bigger investment budgets. It's easier to make the nugget investment in order to deploy some of these capital-intensive automation technologies. On the other hand, some of these other trends that we've seen in technology, whether it's cloud or mobile — or you can buy machine learning with your credit card nowadays — make it possible for even a small company to have access to large amounts of data, to have access to large amounts of computing storage, and to take advantage of some of these technologies. You know there are some great examples on the web of, I believe it's a cucumber farmer in Japan, using machine learning in order to sort their cucumbers. So I mean this is a truly remarkable-

Kerr: It can affect our lives in many different ways, and oftentimes unexpected.

Chui: Absolutely.

Kerr: Okay, as you think about advice we can give to individuals, what could they do to become automation-proof or prepare themselves for the changes that lie ahead?

Chui: So there are a few things. I mean, number one is just awareness. I think that's one of the things I'm most encouraged by, because I think there's a trope that people don't know it's coming. And actually, I think a lot of people have some awareness of the possibility of self-driving trucks or robots in an assembly line, but also machines that will look at X-rays and diagnose disease like a radiologist. I think there's some awareness, but I think that can continue in some greater awareness of what's possible in technology.

I think we'll need to inculcate a mindset of continuous change, continuous learning. I know we've talked about lifelong learning for decades, but it really is becoming real now, and all of us realizing that we're not done with our education after the first two decades of life. We're all going to have to continue to retrain. And that's something that's important, as well.

And then one other thing that I often say, with educators, particularly, just to needle them, but you know I say, “We teach too much calculus and not enough statistics.” And I don't actually mean that, because I think calculus is really important, and the understanding rates of change and all that is really important. But I think increasingly understanding data, understanding conditional probability, particularly data which is incomplete and maybe unreliable, understanding the difference between correlation and causation and type one and type two error. That's not only important in all of our jobs, but is important as citizens, important as we just read about, you know, does this ... How do we think about this nutrition study? How do we think about what matters for our kids in education?

I think that's incredibly important. And it doesn't make us automation-proof. But I think it gives us some of the tools to be able to engage with these technologies and engage with the future, as we all change what we do.

Kerr: We've got some big tasks ahead. The recent McKinsey report, “Jobs Lost, Jobs Gained,” suggests that 400 million people may have to change their jobs or re-skill for the coming 12 years. We appreciate Michael Chui joining us today on the Managing the Future of Work Podcast from Harvard Business School.

Chui: Thank you.

Kerr: And thank you for listening in.