01/13/2017

As Automation Advances, What’s Next for Human Jobs?

11:34 minutes

Almost half the activities people are paid to do globally could be automated using current levels of technology, according to a new report from the McKinsey Global Institute.

But before we start fretting about job security — or daydreaming about lives of leisure surrounded by “Jetsons”-style robot staff, the report includes a few other key details.

For one, all that automation won’t happen overnight: The researchers think half of today’s work activities could be automated by 2055, but say it could take up to 20 more years beyond that. And Michael Chui, a partner at McKinsey and one of the report’s authors, says the findings don’t mean mass unemployment is coming. Instead, he says, given the demands of the global economy, we’ll need to find new ways to work alongside machines.

In all, less than 5 percent of individual occupations in the United States can be completely automated by current technologies, the report estimates. But in our daily work, most of us do something that could be automated — and some of us more than others.

“In about 60 percent of occupations, over 30 percent of the things that people do could be automated — either using robots or artificial intelligence, machine learning, deep learning, all of these technologies that we’re hearing more and more about,” Chui says.

So far, the conversation about automation has focused on machines that replace manual labor. But Chui explains that with developments in artificial intelligence, cognitive work like collecting and analyzing data can be automated, too.

“Some of those activities are things that we’re paying very high-wage and high-skill folks to do,” Chui points out. “If you look across the professions — again, whether it’s a marketer, whether it’s a physician, whether it’s an attorney or an accountant — a significant percentage of what they do is, in fact, collecting data and analyzing data. And that’s increasingly susceptible to automation, as well.”

Rather than leading to mass unemployment, the report’s findings suggest that increased automation could actually plug crucial productivity gaps, as the world’s population ages and birth rates decline. Researchers estimate that automation could eventually raise global productivity by as much as 1.4 percent annually.

Looking just at the United States, about half of economic growth in the past 50 years has been tied to increases in the labor force, Chui says. “We have a growing population thanks to some immigration, as well as more people being born … more women in the workforce, etc. But as the country ages, that demographic dividend is basically going to disappear.”

In the future, then, we may need all human and robot hands on deck to sustain economic growth — but that will bring its own challenges. “What we ought to be doing is trying to solve the problem of ‘mass redeployment,’” Chui says. “How can we continue to have people working alongside the machines as we go forward?”

As it turns out, in the United States, mass redeployment isn’t an entirely new phenomenon. According to the report, in 1900, 40 percent of Americans were involved in agriculture. Today, that number is less than 2 percent, thanks, in part, to technology. “We don’t have 30 percent unemployment because, in fact, we found new things for people to do in the economy,” Chui says. “So we have … historically been able to do that.”

When it comes down to it, it may take time to retrain workers affected by automation. But rolling out large-scale automation will take time, too, Chui says.

In the report, the researchers did some math on how long it takes people to adopt new technologies. They found that technology can take anywhere from eight to 28 years from the time it’s commercially available to when it’s co-opted throughout the economy. And when it comes to widespread automation of our work, Chui says there are other factors that could affect the timeline.

“There’s a political question, there are also questions about what one finds acceptable,” he says. “I mean, there was a time when people would say, ‘You know, if I go to a bank, I want to talk to a person, because this is an important financial transaction. I don’t want to deal with a machine.’ Since then, a lot of people are now transacting either [with] ATMs, online or on mobile, etc.”

“And so that you know again, as we talked about that adoption curve, different people will choose to use automation in different ways over time.”

—Julia Franz (originally published on PRI.org)

Segment Guests

Michael Chui

Michael Chui is a partner at the McKinsey Global Institute in San Francisco, California.

Segment Transcript

JOHN DANKOSKY: This is Science Friday, I’m John Dankosky. Ira Flatow is away. One of the big themes during the election campaign was jobs– bringing jobs back to America versus outsourcing jobs to nations overseas. But many of the jobs lost in recent years aren’t exactly moving, they’re being replaced or modified by automation.

And while people may be able to imagine a robot welding car frames, for instance, it’s a bit hard to wrap your mind around how advances in automation and artificial intelligence could affect other industries like workers in a grocery store. My next guest has been studying automation in the workforce, and in a new report issued today, he and his co-authors estimate that almost half of the activities people are paid to do around the world– yeah, half– could be automated using current levels of technology. But it might not happen right away.

So you think your job has a future? Our number is 844-724-8255– that’s 844-SCI-TALK– or tweet us, @SciFri. Michael Chui is a partner at the McKinsey Global Institute. That’s the research arm of the McKinsey Consulting firm. He’s based in San Francisco and joins me today from KQED. Welcome to the show.

MICHAEL CHUI: Hi there, John. It’s terrific to be here. I’m a longtime Sci Fri fan.

JOHN DANKOSKY: Oh, well I’m glad that you’re here. Well, let’s talk about this. What do you mean when you’re talking about automation? What exactly are we talking about here?

MICHAEL CHUI: Well, we are technically talking about the use of technology to do things that people are currently paid to do in the economy today. One of the things that we did in our research is not only look at the level of individual jobs, because we think it’s actually quite rare that you’ll have a robot that comes in and can do everything that a person does or an AI program and does everything that someone does in their job. In fact, less than 5% of occupations is that true for. We looked at things at the level of individual activities because everybody’s job is made up of different activities, which have different potential to be automated.

JOHN DANKOSKY: So you’re looking at the job itself, but then you’re breaking down all the various activities and you’re saying, some of those activities could be handled by a robot helper of some sort.

MICHAEL CHUI: That’s exactly right. In fact, almost everyone’s occupation has a significant percentage of activities which potentially could be automated using technology. Whether it’s physical activities or cognitive activities, some of the things that we find so amazing are that artificial intelligence now is opening up the possibility for some of the things that previously we thought that only people could do with their brains, now can be done with machines as well.

And what we found is that the potential is really widespread. In about 60% of occupations, over 30% of the things that people do, these activities could be automated either using robots, or artificial intelligence, machine learning, deep learning, all of these technologies that we’re hearing more and more about.

JOHN DANKOSKY: Well, we’ll talk about some of the AI technologies. Maybe you can give us an example, because the history of automation, so far, has been a lot of blue collar jobs being taken away by machines that can do that type of work with artificial intelligence. Now you may be talking about white collar jobs, things that humans thought that only they could do.

MICHAEL CHUI: That’s right. Over history, we’ve usually thought about automation affecting frontline workers, low wage and relatively lower skill activities. And you know, that’s one set of type of activities that we found is relatively automatable using the technology that we’ve seen demonstrated today. But another set of activities, those that are around collecting data and then analyzing that data, are also things that we’re starting to see some of these artificial intelligence techniques be able to take over as well.

And some of those are lower wage activities, things that a clerk might do, somebody who is processing financial transactions in the back office of a financial institutions. But in fact, some of those activities are things that we’re paying very, very high wage and high skill folks to do as well, some professionals. And so again, if you look across the professions, again, whether it’s a marketer, whether it’s a physician, whether it’s an attorney, or an accountant, a significant percentage of what they do is, in fact, collecting data and analyzing data. And that’s increasingly susceptible to automation as well.

JOHN DANKOSKY: When you talk about wages, that’s going to be one of the key factors as to how much automation is actually used. So if you’ve got someone, either in the United States or somewhere else in the world, making a relatively low wage, the calculus has to be, will building a robot to do some of that work actually be more cost efficient than just paying somebody not very much money to do that same work?

MICHAEL CHUI: That’s absolutely right, and that’s what we think is part of the contribution of the research that we’ve done– which you can, by the way, just download off the web. Because, again, what we hear about these amazing things, whether it’s self-driving cars or computers that can read lips better than a professional lip reader, and you say, gosh, that’s incredible. Won’t suddenly all these activities we pay people to do just be automated tomorrow? And we try to break out the fact that it’s not just having technological feasibility, but in fact, you need the economics to work out. You need, in fact, the cost of automation to be outweighed by all the other benefits. And only some of those are reducing labor costs.

And then, by the way, even when you do have a positive business case, it still will take time for all of these automation technologies to take place. When we looked at history, we looked at any type of technology that’s been adopted over time. Usually, it’s taken something like eight to 28 years between the time it was commercially available and the time it was fully adopted throughout the economy.

JOHN DANKOSKY: We’ve talked a bit about physical work and about cognitive work. I want to go to Lisa, who’s calling from Fayetteville, Arkansas, because you’re asking about something else that humans do. Lisa, go ahead.

LISA: Yes, sir. I’m a registered nurse, and I do believe that nothing can replace that human touch. In health care, I know that there are many things that machines take over. There are robots that can do things much better than a human can do, surgically. But part of nursing is that holistic approach, and I don’t see how human interaction can ever go away from that.

JOHN DANKOSKY: Lisa, it’s a great question. Michael, what do you say to her?

MICHAEL CHUI: Well, a couple of things. Health care is, as we talked about, so many different activities incorporated into health care. And as we studied it, absolutely, some of the things that will be hardest to automate over time are, in fact, interacting with people, being able to understand their emotional state, being able to understand the patient holistically. Ironically, some of the things that are perhaps more automatable are some of the higher paid activities in health care.

For instance, again, if you think about a physician, some of the things that we think about them as being their highest value is doing differential diagnosis– that is, collecting data and analyzing data. And so ironically, there are a number of people who think the practice of radiology might be easier to automate than the practice of being a nurse, like our caller.

JOHN DANKOSKY: Well, and another thing that your study looks at that I think is fascinating, it gets to the question of what’s going to happen as more automation takes hold and potentially more people are forced out of jobs? You say that the health care industry is one place where we’re going to have such a need because we’re getting so old as a society, people are living longer, that maybe more people will get jobs in health care, but leave a lot of other jobs to the robots.

MICHAEL CHUI: That’s exactly one of our other findings of the research, because what we’ve discovered is, that over the past 50 years, about half of the sources of our economic growth have come about because we have more people working. We have a growing population, thanks to some immigration, as well as just more people being born, as well as more women in the workforce, et cetera. But as the country ages, that demographic dividend is basically going to disappear. And basically, what we’re saying is we need not only everyone working, but all the robots working too to make sure that we actually have enough economic growth.

Now, that does mean that people’s jobs are going to change. Like I said, a significant percentage of all of our jobs could potentially be automated. And that means we’re all going to have to retrain, we’re all going to have to understand how to adapt as we adopt technology. And so sometimes we’ve described it as, we worry a lot about mass unemployment. What we ought to be doing is trying to solve the problem of mass redeployment. How can we continue to have people working alongside the machines as we go forward.

JOHN DANKOSKY: Yeah, Hayes has a question along those lines. He’s calling from Cedar Rapids, Iowa. Hi there, Hayes.

SPEAKER 1: Hello.

JOHN DANKOSKY: Hi, you’re on the air. Go ahead.

SPEAKER 1: I think you’ve got the wrong person.

JOHN DANKOSKY: Oh, I do? Go ahead, Hayes.

SPEAKER 1: Sorry. I think you have the wrong guy.

JOHN DANKOSKY: Oh, I think we had two people on the line at once, and that was probably my fault. I want to ask you about the timeline here, Michael. What are we talking about? If so many jobs that we have right now or tasks at our jobs could be replaced right now, well, are we looking at the next two years, five years, 20 years? What’s the timeline look like?

MICHAEL CHUI: Well, that is one of the things that we try to contribute. And again we can’t predict the future. We don’t have a crystal ball. But we tried to lay out some of the scenarios for how long it might take for some of these things to take hold. And as you mentioned before, we talked about 50% of the activities that we currently pay people to do could potentially be automated by adopting those technologies that already exist.

But then, how long might that take? And when you factor in the additional technology development that has to occur, when you factor in the economic business cases that we talked about, and then just the natural curve of adoption, that 50% might not happen until 2055. We also modeled out scenarios that occurred 20 years earlier and 20 years later than that, but that’s roughly a work generation. So it might take some time.

Now, two other points about it. One is, if you look at history– and actually, this level of change in what people do is not unprecedented. Around 1900, about 40% of the US population was involved in agriculture. Now, less than 2% are. And we don’t have 30% unemployment, because in fact, we found new things for people to do in the economy. So we have, in fact, historically been able to do that.

But what that does mean is that everyone is going to have to be able to shift, in terms of the activities they do, and that’s part of the challenge going forward.

JOHN DANKOSKY: Yeah, there’s the economic business case, but there’s also the political case too that might affect your timeline. If there’s a push to keep more Americans in the types of jobs that they have right now, perhaps that means that we push off some of the automation that many industries are probably clamoring for.

MICHAEL CHUI: I think a lot of those factors come into play. There’s a political question, there are also questions about what one finds acceptable. I mean, there was a time when people would say, if I go to a bank, I want to talk to a person because this is an important financial transaction. I don’t want to deal with a machine. And you know, since then, a lot of people are now transacting, either ATMs, online, or on mobile, et cetera. And so that, again, as we talked about, that adoption curve, different people will choose to use automation in different ways over time.

JOHN DANKOSKY: And we’re getting used to these electronic personal assistants that are in our pockets right now, and they seem pretty helpful. I don’t know. I want to thank Michael Chui, who’s a partner at the McKinsey Global Institute. Thanks so much for this interesting study and for being with us today.

MICHAEL CHUI: Thanks very much.

Copyright © 2016 Science Friday Initiative. All rights reserved. Science Friday transcripts are produced on a tight deadline by 3Play Media. Fidelity to the original aired/published audio or video file might vary, and text might be updated or amended in the future. For the authoritative record of ScienceFriday’s programming, please visit the original aired/published recording. For terms of use and more information, visit our policies pages at http://www.sciencefriday.com/about/policies/

Meet the Producer

About Charles Bergquist

As Science Friday’s director and senior producer, Charles Bergquist channels the chaos of a live production studio into something sounding like a radio program. Favorite topics include planetary sciences, chemistry, materials, and shiny things with blinking lights.