In January, Amazon opened Amazon Go, a high-tech, cashierless convenience store in Seattle. There are no checkout lines and few employees. The only requirement to shop is downloading an app. Customers just walk in, load up their bags, and go. There’s no need to even scan purchases; cameras positioned overhead take note of items in customers’ carts and add them to a virtual bill. Amazon Go is both an interesting novelty — and a profound challenge to the livelihoods of the more than 3.5 million Americans who work as cashiers.
Rumors of a coming wave of similar stores and robot-run factories have provoked apocalyptic predictions of mass unemployment among pundits and politicians. Doomsday headlines such as “You Will Lose Your Job to a Robot—and Sooner Than You Think” reflect fears that artificial intelligence and robots will replace human labor on a mass scale and computers will become so intelligent that people will simply be unable to compete.
But such a gloomy outlook is unwarranted. Recent analyses from the Organization for Economic Cooperation and Development (OECD) and the McKinsey Global Institute paint a very different picture. Yes, these reports conclude, automation will displace some people from some jobs, but there will still be work for the foreseeable future. The total number of jobs may not even decline significantly, especially in more advanced economies.
[soundcloud url=”https://api.soundcloud.com/tracks/470080494?secret_token=s-NVn22″ params=”color=#ff5500&auto_play=false&hide_related=false&show_comments=true&show_user=true&show_reposts=false&show_teaser=true” width=”100%” height=”166″ iframe=”true” /]
Listen to this story and other feature stories from FP and other magazines: Download the Audm app for your iPhone.
Technological progress will create both winners and losers. Some workers will lose their jobs. A large share of workers will find their work changed, sometimes dramatically; others will discover that their skills are outdated. The cost of this adjustment will not be distributed equally across countries, communities, occupations, or skill levels. The transition will be especially painful for the least educated. Job growth will continue, and incomes will rise for those at the top, but wages for those at the bottom will suffer as many occupations are automated and the demand for lower-skilled routine labor, such as that of cashiers and fast-food workers, gradually decreases.
No country will be immune from the upheaval. The Economist Intelligence Unit recently released what it called the Automation Readiness Index. The key finding: Not a single nation included in the study was fully prepared to address the challenge. A handful of countries with strong education, worker training, and research and development sectors — such as South Korea, Germany, and Singapore — were found to have substantial leads. But even they, along with the rest of the world, will need to take bold action to prepare for the coming automation wave.
Technology may destroy work, but it can also create it. Jobs such as app developer, social media manager, and drone operator did not exist until recently; today, millions of workers hold such titles. According to LinkedIn, data scientist positions in the United States increased 650 percent between 2012 and 2017. As recently as 2015, there were just over 2.3 million data science and analysis job openings in the United States, each boasting salaries of upwards of $80,000, according to a joint report by IBM, the Business-Higher Education Forum, and the data analytics company Burning Glass Technologies. In 2017, researchers from the Massachusetts Institute of Technology and Boston University showed that half of all U.S. job growth from 1980 to 2007 came from the creation and expansion of brand-new job categories. Historically, technology’s positive impact on productivity has also generated broader employment across the economy. Accenture estimates that AI could double economic growth rates by 2035 and boost labor productivity by up to 40 percent.
Despite fears of humans losing out to machines, there are plenty of recent cases where people have worked with machines and, in the process, become more productive, more skilled, and, theoretically, better compensated. The best-known example of this phenomenon comes from banking. As ATMs proliferated during the 1980s and 1990s, bank tellers simply shifted their focus toward customer service. The teller job evolved into a higher-skilled position, which may have resulted in rising wages. More tellers with college degrees were hired. (Of course, tellers are not immune to the move to online banking and digital innovation, and their numbers are now declining.)
Many of today’s most alarming headlines reflect concerns over whether a specific occupation could be automated but do not consider whether it will be. Legal, policy, and practical infrastructure, as well as cultural and cost barriers, could stand in the way of the widespread adoption of AI, thereby blunting, or at least slowing, its impact on human labor. To take a recent example, deadly road accidents involving self-driving cars could trigger more regulatory or legislative constraints or simply weaken interest among consumers, keeping taxi drivers behind the wheel for another few years.
Looking only at the net number of jobs created or destroyed by AI, however, obscures the wider distributional challenges it will create. In the future, there may still be enough jobs overall — but not necessarily in the same place, in the same occupations, using the same skills, or offering the same pay and stability as today. For highly skilled workers, automation may be life-enhancing: Virtual assistants will take over scheduling and administrative tasks, and algorithms will help simplify routine tasks. But AI poses a much greater threat to the millions of low-income workers tasked with routine, predictable, and repetitive work. Office clerks, cashiers, retail sales workers, administrative assistants, waiters, fast-food cooks, and assembly workers will all be at risk.
That’s not to say all low-skilled work will go away. As the world’s population ages in the coming decades, job opportunities in the health care sector, such as home health aides and nursing assistants, are expected to rise considerably, offsetting some of the lower-skilled job losses. As Sarita Gupta and Ai-jen Poo argue in “Who Will Care for the Carers?”, work that requires empathy remains hard to automate. Professions that tend to require flexibility, creativity, and judgment will also fare well. These jobs require workers to create their own strategies to solve complex and unpredictable problems instead of following fixed rules. For now, computers simply cannot perform the core, human-oriented tasks of clergy and hospice workers. Nor, for that matter, can they undertake the complex problem-solving done by CEOs, civil engineers, and public defenders.
Far more people will find their jobs changed rather than lost because of automation. Uncertainty will replace stability as skills become obsolete. A recent OECD report suggests that roughly a third of jobs across member countries face some risk of automation and could be changed substantially. McKinsey estimates that between 75 million and 375 million people globally will need to switch occupations by 2030. A far greater number will need to develop new skills in order to adapt. More than a third of the tasks performed in over half of current jobs could be automated with today’s technology, which means most workers will need to adjust soon either by learning to work with robots and technology or retraining for new jobs.
Technology will also exacerbate inequality along geographic lines. In wealthy countries such as the United States and Japan, automation is likely to change more jobs than it eliminates, and inequalities will grow. In September 2017, researchers from MIT, Northwestern University, and Australia’s Commonwealth Scientific and Industrial Research Organisation found that smaller cities will face greater displacement than large cities as economic growth and tech talent concentrate in bigger, prosperous urban areas. This is already happening. The vast majority of U.S. job creation since 2000 has been in wealthy zip codes with a large base of highly educated workers.
The automation risk to employment is highest in middle-income countries where work requirements are less complex, cognitive, creative, and interpersonal. The OECD found this year that nearly a third of all jobs in Slovakia were at high risk of automation, compared with only 12 percent in the United Kingdom, for example, due in part to the country’s dependence on manufacturing jobs. Similarly, countries such as Lithuania, Greece, and Turkey face a substantially higher risk of job losses than wealthier countries such as Norway and the United States. Countries in Southeast Asia that are heavily invested in the low-cost manufacturing of textiles, cars, and electronics could face the greatest risk of all. The International Labour Organization estimates that just over 137 million people — or some 56 percent of salaried workers in Cambodia, Indonesia, the Philippines, Thailand, and Vietnam — are at high risk of being replaced by machines.
The larger question is whether these countries will be able to exploit the same pathways to prosperity that the so-called Asian tigers (Hong Kong, Taiwan, Singapore, and South Korea) once followed to produce meteoric economic growth between 1960 and 1995. In 2015, the Harvard University economist Dani Rodrik wrote a report for the United Nations Development Program analyzing how the manufacturing sector has historically served as an escalator, moving countries out of poverty to middle- and eventually high-income status. Leveraging their abundant low-skilled labor, poor countries have manufactured clothing, electronics, toys, and footwear for export, which has boosted growth and development and provided employment for their citizens. As technology automates much of that kind of production, jobs will vanish, and a few wealthy countries will even pull some manufacturing back home.
Today’s iPhone factories, for example, are robot intensive and require very few low-skilled production workers. Apparel and footwear manufacturers are experimenting with fully automated facilities, and sewing robots could eventually displace low-skilled garment workers. As Christina Larson writes in “Closing the Factory Doors,” this sort of decline in manufacturing employment could knock out what has long been a reliable first rung on the ladder to growth and development. If these and other countries like them can’t improve their education systems in a way that increases human capital fast enough to move hundreds of millions of young people into higher-skilled service and knowledge economy jobs, these nations could soon face a mass youth unemployment crisis.
Higher-skilled jobs in Asia could also face disruption. Automation will impact offshore customer service call centers and companies that handle outsourced IT and office work, which currently employ millions of workers across South and Southeast Asia. Technology has not yet automated all customer service, but experts predict that it soon will. In the Philippines, where more than a million people work at call centers, preparing the workforce for automation through training for complex and advanced work will be essential. In 2016, call centers and the broader office outsourcing industry generated nearly $23 billion in revenue, equivalent to 7.5 percent of the country’s GDP.
Helping millions of workers to gain new skills and adapt to rapidly changing and new occupations will also be a daunting task in wealthier countries. According to the U.S. Bureau of Labor Statistics, nearly 3 million people in the United States work as general office clerks, a position highly vulnerable to automation. The vast majority are women, with a median age of 43.5 and an average salary of just over $30,000. Fewer than one in five has a bachelor’s degree. Unless this population is able to acquire a new set of skills, additional education, or better credentials, many of them will find it impossible to sustain their current standard of living. But the chances of them finding ways to adapt don’t look good at the moment. U.S. colleges and universities spend hundreds of billions of dollars on students annually — while the federal government is currently set to spend a mere $17 billion on job training. Yet for the most part, higher education is not designed to properly address the evolving and dynamic needs of mid-career workers seeking new skills. A four-year, full-time university degree may not be practical, let alone possible, for a 42-year-old office clerk who needs to keep paying her bills. Working adults with families looking to retrain require greater flexibility, shorter course duration, and skills that can be applied immediately.
To make matters worse, the skills training programs that do currently exist in advanced economies for the most part skew to the highly educated and highly skilled workers at one end of the employment spectrum and the most vulnerable and least skilled at the other. Very little of this programming is focused on helping current mid-level workers at risk of automation.
That’s a problem because the jobs that will be created through technology and automation require more education and skills than those that will be lost. The recent changes in manufacturing employment in the United States illustrate this trend. According to the Congressional Research Service, the 84,000 people working in the U.S. steel industry in 2016 produced 3 percent more steel than nearly 400,000 workers did in 1980. The reduced need for workers in factories means that today there are nearly 5 million fewer manufacturing jobs in the United States than in 2001. The jobs that have disappeared are disproportionately those at the bottom of the skill spectrum and do not require more than a high school degree (jobs such as welding, removing parts, and working on an assembly line). U.S. workers with high school degrees have seen the greatest decline in employment, while the number of workers in manufacturing with a graduate degree grew 35 percent between 2000 and 2016.
Yet data from the most recent OECD report suggest that workers at greatest risk of displacement are more than three times less likely to have already engaged in training than those in nonautomatable jobs. Innovations in credentialing and training could help these workers. In the United States, for instance, such innovations include massive open online courses, career-oriented nanodegrees taught by experts that provide industry-recognized credentials, and coding boot camps that teach career-ready IT skills in a few months. At present, however, these programs are used predominantly by those who are already highly educated or digitally savvy and looking to further enhance their employability by mastering cutting-edge technology. Few workers at the lower end of the labor market are taking advantage of such programs.
One of the main challenges for education systems around the world is to ensure that more future workers graduate with postsecondary credentials that prepare them for the job market of the future.
In their 2010 book, The Race Between Education and Technology, the Harvard economists Lawrence Katz and Claudia Goldin outline how education investments on a mass scale can create a virtuous cycle — providing the human capital that the market demands, bolstering productivity, increasing incomes, and driving growth and shared prosperity. In contrast, when education levels have not kept pace and workers don’t graduate with the skills employers need, which has been the case in the United States since the 1970s, the demand for high-skilled workers outstrips supply, and inequality sharply increases as the highly educated command growing salaries while the rest are left behind.
A few standout countries are leading the way in investing in education. Fifty years ago, South Korea’s educational institutions and economy were on par with the world’s poorest countries. Today, South Korea is one of the world’s top performing countries in educational attainment across socio-economic backgrounds. The Economist Intelligence Unit ranked South Korea first for education in its Automation Readiness Index, citing the government’s reforms in teacher training and curriculum and its emphasis on soft skills. South Korea’s high school graduation rate — above 90 percent — is extremely high, especially compared with that of the United States. The commitment to education is evident from households to the heights of government; per student spending is higher than in some comparably wealthy countries.
But South Korea is an outlier; in recent decades, educational attainment has slowed in many advanced economies, and lower-income countries still struggle to lift much of their populations to even a basic level of literacy. In the United States, for instance, intergenerational educational mobility has stalled. According to a 2014 OECD survey, the number of 25- to 34-year-olds who were less educated than their parents was equal to the number of those who were more educated; men in this age group were even more downwardly mobile than women. Automation will only worsen this problem.
Rethinking educational systems and curriculums to ensure today’s students gain the skills that will equip them for a rapidly changing labor market is thus essential. Some places have already begun: The Canadian province of Ontario is training its young people in critical thinking, communication, creativity, collaboration, and entrepreneurship. Even China is making a push to incorporate creativity into its notoriously rote curriculum.
Another challenge for educational systems is not only how well they teach students but also when they teach them. Singapore has embraced technology and automation while taking proactive steps to prepare its workforce for changes, including easing access to continuous education. The government launched a lifelong learning initiative, offering $370 subsidies to all Singaporeans aged 25 and over to study in hundreds of career-oriented courses. Singapore’s national university also adapted to offer more worker-friendly educational opportunities, including part-time degrees, modular certificate courses, executive education, and free classes for alumni.
There are some bright spots in the United States. In 2010, Chicago’s then-mayor, Richard M. Daley, revamped the city’s seven community colleges in an effort to improve graduation rates and tie learning directly to in-demand careers. The city government forged partnerships with more than 100 companies to influence curriculum development, advise on what credentials were most valuable, and help make the degrees more commercially relevant. Six out of seven of these campuses have reoriented their curriculums to align with marketable specializations that are in demand. Governing magazine recently reported that the graduation rate at these seven community colleges nearly doubled between 2010 and 2014 and the number of degrees awarded climbed significantly. And Chicago’s model has attracted the attention of the World Bank as a program that could be replicated globally.
Not surprisingly, the countries with the most developed social safety nets are best positioned to respond to automation. In Sweden, job security councils, jointly managed by the private sector and unions, retrain workers who have been made redundant by automation. The councils teach them in-demand skills that are less vulnerable to automation, and unemployed workers are given generous benefits while they train. The Swedish government has also backed a nonprofit pilot program that provided career counseling to workers over 30 to assess their skills and career prospects, which helped increase participation rates.
Sweden’s neighbor Finland piloted a more radical solution to revamping the safety net: a universal basic income. In the Finnish experiment, launched in 2017 and due to wrap up this year, the government is providing 2,000 unemployed people with a $660 monthly payment, no strings attached, to examine the grants’ impact on employment. Although Finland won’t release the results until after the study is complete, other countries, such as Canada and Kenya, are also experimenting with universal basic income.
Scandinavian welfare states are exceptional in their long-term planning. Outside of Northern Europe, many workers displaced by automation will lack the kind of government support that could cushion the blow. In the United States, pensions, workers’ compensation, and health benefits are still largely tied to employment. This system risks accentuating existing inequalities, creating an even larger class of unemployed workers with no government programs in place to assist or retrain them.
Properly preparing for automation will require a long-term vision, bold goals, and immediate action. Workers themselves should be at the center of this planning and policymaking. Expecting millions of workers to change their careers, go back to school, retrain, and adjust to a rapidly changing labor market is a lot to ask. Improving the supply and affordability of training options will help, but it will not be sufficient to ensure that workers adapt. Understanding the expectations, preferences, and mentalities of a broad range of vulnerable workers will be critical to designing effective solutions. Success will require addressing the broader structural barriers that stand between these workers and new opportunities, such as financial insecurity, lack of affordable child care, and poor access to transportation.
Getting this wrong could deepen society’s divisions and exacerbate inequality, political polarization, instability, and even global insecurity. It will also negatively impact millions of workers — profoundly and personally. This is especially true for workers at the lower end of the income spectrum, from the 18-year-old American cashier trying to earn a living in her first paying job to the 35-year-old Bangladeshi garment worker striving for a better life for her three children.
As technology transforms their occupations, their futures could be filled with risk or opportunity, depending on the policies their governments adopt. The cashier displaced by new innovations like Amazon Go could fall into long-term unemployment, or she could develop the interpersonal, digital, and problem-solving skills that the economy increasingly demands. The government could leave her to fend for herself in a new and unfamiliar job market, or it could provide the financial support and educational guidance she needs to retrain and adapt.
Today, most systems, policies, and programs are not yet set up for her success. That must change — now.
This article originally appeared in the July 2018 issue of Foreign Policy magazine.
Molly Kinder is a senior advisor on work, workers, and technology at New America and a research fellow and adjunct professor at Georgetown University’s McCourt School of Public Policy. (@MollyKinder)