It was a warm April morning in Washington, D.C., when I stepped into the Executive Office building less than three months after President Donald Trump took office. The president’s schedule that Tuesday noted that he would host a CEO town hall on the business climate. I was part of that CEO gathering as a member of the Partnership for New York City, an organization that represents the city’s business leadership and its largest private-sector employers. The partnership works with government, labor, and the nonprofit sector to promote economic growth and maintain the city’s position as a global center of commerce and innovation. Through its Partnership Fund for New York City, we contribute directly to projects that create jobs, improve economically distressed communities, and stimulate new business creation.1
Presidents throughout history have had a similar focus. Three years earlier, President Barack Obama had named me to his advisory council on financial capability for young Americans. Its aim: to help ensure that all young people are equipped with the knowledge and skills necessary to make smart decisions about their finances.
On that spring day in 2017, President Trump and Vice President Mike Pence welcomed us and made a few brief remarks. Deloitte’s then CEO, Cathy Engelbert, rose to ask the president for his thoughts on the priorities of the administration around education and the future of work—or, as she phrased it, the work of the future.
“Improving the quality of our workforce, expanding opportunities for what is known as career and technical education—what we back in Indiana call vocational education—is a real passion for our new president,” Vice President Pence said. “And we look forward to partnering with you in ways that we can continue to encourage investment and create opportunities for expanded career and vocational education.”2
A few years later, as the 2020 presidential campaign progressed from state to state, candidates pushed the issue of automation and the future of work to the top of the agenda. Daily newspapers published hundreds of articles and editorials. According to the Pew Research Center, most Americans anticipate widespread job loss as a result of automation.3
The future is here.
The London School of Economics observed in a 20194 blog post that barely a month passes without forecasts of technology-induced job loss. Automation is coming for your job and mine. Two researchers from the University of Oxford, Carl Benedikt Frey and Michael Osborne, ignited the debate with their paper, “The Future of Employment.” According to Frey and Osborne, fully 47 percent of U.S. jobs were at high risk of being automated. This number was larger than many people had expected and quickly became one of the most widely cited figures in discussions about automation and jobs. Although researchers had been studying automation for years, no one had yet sounded the alarm about job displacement in such a direct way.5
The news headlines over the next several years were dire. The Atlantic foretold “A World Without Work.”6 The New York Times asked the question on everyone’s mind: “Will Robots Take Our Children’s Jobs?” 7 And CNBC declared, “AI and Robots Could Threaten Your Career Within 5 Years.” 8
Frey went on to publish, in 2019, a more sober analysis, The Technology Trap: Capital, Labor, and Power in the Age of Automation, which looked at labor market changes over the course of various industrial revolutions. Technology has a history of making workers more productive but also displacing jobs, hampering wages, and producing inequality, a dynamic known as the “Engels pause,” named for Friedrich Engels, who cowrote The Communist Manifesto with Karl Marx.9 The Engels pause is cited as preceding the Luddite uprising, which was a protest against technology and automation. Frey raises a provocative question about whether or not we could face a new technology trap, a new Luddite uprising, in order to stop AI, machine learning, and other emerging automation. The book concludes, as the London School of Economics book review notes, with the importance of education, which “should send alarm bells ringing … In the so-called race between technology and education, current budget cuts amid a general slowdown in human capital accumulation will present workers (both present and future) with even greater troubles ahead.”10
This long-simmering discussion about what technology will do to our workforce and our economy can be seen everywhere—in the media, in politics, and in the arts.
The critical question of our time is what to do about it. When I read pieces about automation and the future of work, I’m reminded of the urgent pleas of climate scientists. The first lives lost to global warming have already been counted.11 So have the first jobs been lost to labor-saving technologies such as AI. And just as there are climate deniers, there are economic history deniers. Facts are facts.
Layoffs tend to dominate the headlines, but there are many other illustrations of a slower and longer-term trend. Little by little, industries are finding new ways for automation to supplement the work of humans—and change their jobs in the process.
At Amazon’s warehouses, for example, fulfillment coordinators don’t walk the aisles filling boxes themselves. They receive orders, punch instructions into a computer, and send robots and drones to fetch products off the shelves.12 Meanwhile, in hospitals across the country, pill-dispensing machines such as Swisslog’s PillPick are filling prescriptions for patients’ medications.13 And at some Marriott hotels, guests direct their comments to Alexa instead of calling down to the concierge.14
Since the initial Oxford study was released, a number of researchers have attempted to quantify the effect of technology on the future of work, and all have arrived at different conclusions. Many of the conclusions were more optimistic than those of Frey and Osborne. But the truth is, no one seems to know what will happen as new technologies render more and more tasks automatable and certain occupations obsolete.
Experts do, however, agree on a few things. Contrary to what the most sensational headlines would have us believe, we will not simply wake up tomorrow and find that millions of jobs have disappeared. Although technology has made enormous strides in recent years, we are still a long way off from any sort of apocalypse.
Some occupational categories might disappear, but many more will simply evolve, requiring new sets of skills over time—and providing new opportunities for workers. Moreover, at the same time that automation displaces some jobs, it will create others. A comprehensive McKinsey study, published in 2017, estimates that between 8 and 9 percent of labor demand in 2030 will come from jobs that don’t currently exist, and indicates that up to 375 million workers might need to find new types of work and learn new skills.15 Meanwhile, attendant economic growth from rising incomes and increased goods consumption alone could add 250 million to 280 million new jobs.16
McKinsey also looked at the issue of gender and found only minor differences in job prospects between men and women.17
In the case of jobs lost, women may be only slightly less at risk than men of their job being displaced by automation. In the ten countries, an average of 20 percent of women working today, or 107 million women, could find their jobs displaced by automation, compared with men at 21 percent (163 million) in the period to 2030…. By 2030, women could gain 20 percent more jobs compared with present levels (171 million jobs gained) vs. 19 percent for men (250 million jobs gained). Across the ten countries in our sample, on average 58 percent of gross job gains by women could come from three sectors: healthcare and social assistance, manufacturing, and retail and wholesale trade. On average, 53 percent of men’s gross job gains could come from the manufacturing, retail and wholesale trade, and professional, scientific, and technical services sectors. Women are well represented in fast-growing healthcare, which could account for 25 percent of potential jobs gained for them.
***
A microcosm of the future of jobs has already played out in a pizza kitchen in northern California.18 As the case with most pizza kitchens, Zume Pizza was the site of much activity. There was dough being rolled, sauce being spread, and pies being loaded into industrial pizza ovens. However, unlike most pizza kitchens, those tasks were being completed by robots capable of pressing dough five times faster than humans and churning out 288 pizzas an hour.19
Zume Pizza, a startup, integrated automation into every stage of the pizza delivery process. In addition to robot food preppers, for example, the company used predictive analytics to determine workflow and demand so it could be more precise when purchasing ingredients. Still, even though Zume was tech-driven, 30 of its 150 employees worked alongside robots in the kitchen and during the delivery process.20
It might be only a matter of time before robots become proficient enough to fully take over pizza-making and delivery tasks, but this doesn’t mean there will be fewer jobs available at companies like Zume. In fact, its founders imagined a future where robots free up line cooks and delivery workers to learn new skills and take jobs on technical support teams or in programming and web design.21
Zume highlights the potential that automation holds for the business community. Technology that replaces repetitive tasks can leave more innovative and more cognitively demanding jobs in its wake. But business leaders must be willing to take that last step—to help their employees get the skills they need to adapt to these new and different roles and contribute productively.
Whether pizza makers, stock pickers, or grocery store cashiers, people are being replaced. Not far from our offices in New York City, Yotel, the small UK-owned hotel chain, offers another glimpse of the future of consumer automation and the future of work. Yotel has put machines in charge of repetitive tasks so that hotel staff can focus on higher-touch, more personal services. On the street level, Yotel guests check themselves in at kiosks and receive a room key. There the hotel robot—known as “Yobot”—can store your luggage if your room is not ready. Take the elevator to the fourth floor, and there you are welcomed into an enormous, beautifully appointed lounge with plenty of staff for concierge services, bars, restaurants, and work spaces. Think WeWork or another collaborative work space.
As academics and business leaders across the country have illustrated, we have the tools to prepare ourselves for the future of work. Indeed, we have an unprecedented opportunity to adapt to a new reality and realize the full potential of the Fourth Industrial Revolution.
But we must not only act; we must act with intelligence. David Deming, a Harvard professor focused on skill development and the labor market, writes that we need a strengthening of the social contract to manage the changing nature of work in the digital age.22
Right now we can look at projections for the years 2030, 2040, and 2050. These are helpful guides, but the worst mistake we can make at this point is to get so caught up in imagining what could happen that we forget our own agency in determining what will happen.
Automation is changing the way we work, quickly and irreversibly. To understand what we can do about it, we must first understand the technology itself—and then its effects on workers, businesses, and society.
THE QUICKENING PACE OF CHANGE
On July 20, 1969, more than 500 million people around the world gathered in front of their televisions to watch Neil Armstrong and Buzz Aldrin take “one small step for a man, one giant leap for mankind.” America had put a man on the moon.23
People everywhere marveled at the technology on display. The Saturn V rocket, which launched the Apollo 11 astronauts into space, contained five F-1 engines, weighed 6.2 million pounds, and stood at 363 feet tall.24 Even today, with 7.6 million pounds of thrust at liftoff, it remains the most powerful rocket engine ever constructed.25
Equally impressive was the computing power that helped guide the Saturn V into space, allowing Armstrong and Aldrin to traverse a total of 356,000 kilometers—to the Moon and back—safely.26 The aptly named Apollo Guidance Computer was a real-time operating system that made it possible for the astronauts to control the spacecraft by entering commands that consisted of pairs of nouns and verbs.27 It was designed by a team of some of the brightest minds at the Massachusetts Institute of Technology (MIT).
The lines of code that powered the Apollo Guidance Computer amounted to two megabytes, allowing the system to store up to 12,000 “words” in its memory.28 Though this was a significant feat at the time, our computing power today has far eclipsed it. The storage capacity required to run the top ten most installed apps on the iPhone is about 1.9 GB—950 times more than it took to put a man on the moon.29 In less than a generation, we went from using landline telephones, watching TVs with antennas, and living in computer-free homes to carrying—in our pockets—a mobile device that uses more sophisticated technology than that required for the Moon landing.
As technology has advanced, it’s also become more accessible. In the early days of computing, it could cost $200,000 a month to lease a computer.30 This was a piece of technology reserved for only the most distinguished scholars at the most elite universities. Today the average price for a smartphone significantly more powerful than those old machines is $363.31 There are several reasons that technology is moving faster than ever: more data, more computing power, and better understanding of what to do with both. AI has been around since the 1950s, but in recent years its sophistication—and therefore its insights and predictive abilities—also have grown.32 Purists will question what we mean by AI; for purposes of this book, we refer to AI as inclusive of a range of data-intense subfields such as machine learning.
The field of AI has entered a crucial, exciting, and worrying new phase, thanks in part to sets of algorithms known as deep neural networks—so called because they are modeled on the connections between neurons in a human brain. Instead of simply crunching data and spitting out a result, these neural networks can take large amounts of data, recognize patterns, and make predictions. They can run tests to see if their predictions are accurate. According to Gary Marcus and Ernest Davis in Rebooting AI, early machine learning scores of 75 percent correct were considered good. By 2017, thanks to deep neural network research, those scores had reached 98 percent correct. And from the results of these tests, for the first time, neural networks can learn what works and what does not. Not only that, in some cases computer scientists no longer need to train their algorithms; the algorithms can learn by themselves through techniques such as transfer learning and unsupervised learning.33
Deep learning has exponentially sped up technological advancement and cracked open a whole new realm of technological possibilities. The pace of change is no longer constrained by the limits of human innovation. In 2018, we saw a glimpse of what this could mean for tasks that have traditionally been performed by human beings. IBM, an early pioneer of AI—and the company behind two of the world’s most famous supercomputers, Deep Blue and Watson—staged a new intelligence matchup between man and machine.
It was called Project Debater, and it was designed to break new ground by bringing an AI system into the unpredictable world of debate, where it would have to build an argument and rebut a human opponent in real time. Project Debater was designed to understand speech, process vast data sets, create coherent lines of reasoning, and emulate natural language—all while adapting to new information served up by its opponent. This was significant because although machines are exceptionally good at rote tasks in structured environments—for instance, putting parts together on an automotive assembly line—in the past they have been exceptionally bad at functioning in real-life conditions.
At an event in San Francisco, Project Debater took the stage next to Israeli student debate champions Noa Ovadia and Dan Zafrir. Their topic was government-subsidized space exploration. The AI system had never studied it. Instead, it drew on massive troves of data to construct its argument, and it performed admirably well—opening with a joke, making a case for subsidies, and rebutting the students’ points.
The debate was an eye-opening display of where this technology could lead, and fast. The machine seemed to “think” like a human. It was a remarkable step forward for AI and foreshadowed the world we could one day live in, where issues of law, politics, and ethics might be handled not by humans but instead by machines.34
TECHNOLOGY’S EFFECT ON WORKERS
Fortunately, this foreshadowed future is not immediately at hand. Still, some core assumptions about which jobs are vulnerable to automation have been proven wrong.
Both blue-collar and white-collar occupations are at risk, including office jobs that were once considered safe. Repetitive, predictable tasks—whether it’s fitting parts on an assembly line or entering data into a computer—are the most easily automated. Meanwhile, creative, unstructured tasks—those that might involve strategic thinking, relating to others, or coming up with solutions to complex problems—will likely remain: the so-called new collar jobs.
Not all skills are created equal; and education, a fundamental building block in achieving the American Dream, remains the key to surviving and thriving in the economy of the future.
In 1979, a man with a high school degree could expect his median earnings to be $17,411 less than a man with a college degree. By 2012, this number had just about doubled, to $34,969. For women, the gap in earnings similarly grew, from $12,887 to $23,280 annually.35 Estimates of the differential over a lifetime range from $275,000 to $1 million.
As more and more tasks become automated, this gap will only widen. The people who have the means to acquire higher-level skills and expertise will have a much better shot at a comfortable livelihood.
Peterson’s, a media company focused on education and career planning, recommends a college timeline that begins in the ninth grade with the student meeting a guidance counselor, getting involved in extracurricular activities, picking the right classes, checking out colleges, and making use of every summer to build credentials. By the tenth grade, the student should take a practice PSAT, in the eleventh, take the PSAT—and remain in touch with a guidance counselor throughout in order to prepare for and choose the right college.
Though education is the answer, this is not a clear-cut, multiple-choice test. There are many complicating factors when it comes to getting the skills needed to thrive in today’s economy.
Take an employee who has recently been laid off from a retail job. She has a high school diploma, but she never pursued higher education because it was too expensive—and it’s been many years since she was last in a classroom.
Whereas once she might have been able to find stable employment at the local factory, today no such job exists in her town. So she decides to pursue an entry-level job as a receptionist at an office down the road, only to find out that the position requires a bachelor’s degree.
It would be difficult for her to balance her full-time job with a full class load and impossible for her to give up a regular paycheck. When she weighs the cost of a degree—in time, tuition, and work hours sacrificed—the opportunity hardly feels like a deal.
Even if she does find a way to get the training she needs to land a new job, perhaps through a local degree or certificate program, it might be only a few years until the job looks very different and the skills needed to succeed in it have changed dramatically. According to the National Association of Colleges and Employers, graduates with technical, job-specific skills are likely to see their skills become out of date within six years.36
Computer hardware engineers are facing this challenge right now. In the past these engineers were involved in developing, testing, and managing computer hardware components. But with the advent of cloud computing platforms, demand for on-premise computer servers has decreased. And because companies aren’t as likely to need a local server, they don’t need server administrators, which has contributed to a 15 percent decline in the number of employed computer hardware technicians since 2001.37 Cloud and edge computing are creating entirely new jobs and transforming traditional ones.
Radiologists and a variety of other physicians are also predicting a major skills shift.38 One of the things a computer can do reliably better than a human is recognize patterns, which is a core function of a radiologist. Today aspiring radiologists in the United States must attend four years of college and four years of medical school before they’re qualified to scrutinize scans for abnormalities and diagnose conditions such as cancer and heart disease. But machine learning could change this. Leaders in the tech field anticipate that giving a computer millions of images to analyze could eventually help it learn what different diseases look like, resulting in an algorithm that could offer quicker, cheaper, and more accurate diagnoses.39 This doesn’t mean we won’t need radiologists, but their jobs will certainly change. Some will learn how to monitor their machines’ results, while others will pursue specialty fields.
Automation will bring with it plenty of opportunities, and workers will need access to educational opportunities to take advantage of all that technology has to offer.
THE SHARK TANK
To encourage AI-infused thinking in our own business, Guardian holds periodic “Shark Tank” competitions among colleagues and partners. As in the television show, the Sharks—experts at evaluating the promise of ideas—listen to brief entrepreneurial ideas designed to innovate. The winning idea must create loyal customers, increase revenue, reduce costs, manage risk, and take into consideration the effort required to implement it. From a product perspective, how can you help us operate at scale using cognitive computing? From a technology perspective, how can you help us build a modern hosting strategy? And from a customer perspective, how can you help us secure the future of our distribution?
These are critical questions, especially now. According to Deloitte, the number of households holding life insurance is at a fifty-year low. “Digital disruption has changed demand and how consumers interact with agents. To adapt, carriers must use more predictive analytics, increase digital offerings, and activate agents to give advice in a digitally enabled ecosystem.”40
During one such Guardian Shark Tank session, the late summer sun blazed through the windows of a conference room on the twenty-first floor of our headquarters in the Hudson Yards neighborhood of New York City. Below, walkers and joggers meandered along the elevated High Line Park. Ambulances blared and horns honked. More than thirty partners submitted 103 ideas, and just a handful were selected for presentation. About seventy Guardian voters joined this particular Shark Tank either in the room or remotely via Skype. They would hear five presentations:
• AI-based claims pre-adjudication—Use AI to improve data collection and triage.
• Cognitive medical intake accelerator—Reimagine the disability claims customer experience by applying the latest digitization, AI natural language understanding, and machine learning.
• AI-enabled intelligent search—Imagine your customer contact center equipped with AI capabilities that assist contact center agents to respond faster and more consistently.
• Customer/Agents self-service bot—An AI agent will respond to claims and ID card inquiries with lower cost.
• Mainframe modernization—Transform legacy applications from the mainframe to the cloud.
For the first presentation, a Guardian leader is paired with an external partner with the capacity and expertise to help. We’re told that more than 75,000 short-term disability claims are filed with us each year. Nearly three-quarters are processed by eighteen intake agents during extended but still limited business hours. The opportunity is to use AI to streamline that process and make it a 24/7 service—to improve quality, customer experience, and manage costs.
“Innovation is about creating customer value,” one presenter reminds us.
Using a prototype, we meet “Gia,” the Guardian Intelligent Agent that can process a claim in five minutes or have you speak with a live agent if you prefer. Gia uses natural-language processing to understand the customer and to produce a transcription so that input automatically flows into the proper fields of the required form. Gia’s machine-learning core draws on data for prediction and inference. “How long will it take for me to get paid?” Gia is asked, and the correct answer and process are fed back instantly. The customer can speak in his or her native language.
At the end of the presentation, voters cast ballots from their computers and smartphones. The results are shown live on a screen in front of the room. Gia receives high marks for reducing costs, requiring little effort to implement and for creating loyal customers.
Next up is another AI product: CMIA, the cognitive medical intake accelerator. This technology can review thousands of pages of clinical documents, digitize them, and extract findings and summarize all of them them for the case manager. Its predictive ability helps Guardian know where to help. To demonstrate, we meet a fictitious patient, John Smith, who is a 54-year-old truck driver with a herniated disk in the L4 vertebra. He’s been out of work for several months. The CMIA product examines medications, social history, latest vital signs, family history, and other factors. Digitizing these data and feeding them into a machine-learning program will reduce costs, help to manage risks, and improve customer service.
The friendly competition is engaging all parts of our organization. Meanwhile, we’ve also assembled an internal corporate venture fund. The goal is to fuel both growth and learning by investing in early-stage technology companies. We’re looking to leapfrog our legacy systems and practices while investing in high-impact businesses that can accelerate learning.
After I became CEO at Guardian, we created a venture fund to invest in early-stage companies aligned with our tech priorities and in areas where we wanted to learn. This embryonic idea was supercharged when Andrew McMahon joined the company. Andrew grew up on Long Island, the son of a State Farm executive, so you could say insurance ran in his veins. A math whiz and computer science grad, he had served tours of duty with GE, McKinsey, and AXA insurance and had brushed shoulders with Jerry Yang at Yahoo. Andrew joined Guardian convinced that data and technology were being underutilized in insurance. Today Andrew is president and CEO-elect of Guardian.
The idea with the venture fund was to take 5–10 percent stakes in promising startups. Though the portfolio is not large, the potential impact on the business is. To advance this we recruited Mike Kryza, a Kellogg School of Management grad with a long résumé of experience in financial and insurance technology. Ever since he joined Guardian, I’ve reviewed weekly reports covering the myriad investment recommendations, which fall within four investment categories.
The first is operational innovation. These are companies that are applying next-generation tech to make our back office more efficient and productive. Human API, for example, is an electronic health records firm that digitizes enormous amounts of information to help our underwriters. Underwriters, as Investopedia defines them, evaluate and analyze the risks involved in insuring people and assets. They establish pricing for accepted insurable risks. To accomplish this they must review what’s known as an “attending physician statement,” which is typically a 120-page medical file emanating from one or multiple sources. It’s not easy to assemble all of that information—it arrives in analog form—and it’s even more time-consuming to receive, often requiring thirty to sixty days. Human API has built a database that provides these data digitally, which are sorted and in a consistent format. Underwriters can get the information they need in minutes, not months. And the database puts us on the path to an AI risk assessment system that will make it even smarter and faster.
The second is the future of distribution. Increasingly, people want to shop and buy online. We invested in Aktibo for its sophisticated data analytics. Just as your profile might suggest you like hiking or traveling, Aktibo can help us understand your propensity to need life insurance.
Data analytics and technologies that help manage our company benefits fill out the investment “sleeves.” We have invested in several companies in these categories. Tuition.io is a tech platform for student loan solutions. Student loans amount to a more than $1 trillion liability nationally. People are entering the workforce in debt. One of the most important benefits for new hires is student loan management. It can be even more important than a 401k benefit early in someone’s career. Tuition.io helps those workers build a strategy for reducing student loan debt. In chapter 1, I mentioned another investment, Jobble, which brings together gig workers with various contractors who need their time and skills. Jobble fills a gig worker’s day—drive for Uber in the morning, go to a convention center to set up chairs and table at mid-day, go to a bakery to unload flour, and then drive for Lyft. We invested to learn about gig economy workers but also to offer dental and vision insurance.
We’ve all been on calls with sales representatives for airline tickets or other consumer products. Call centers handle massive numbers of transactions with varying degrees of customer satisfaction. The same is true for insurance products. It can be frustrating. To help, we invested in Cogito, an AI-based platform for real-time sentiment analysis for call center representatives. The human capacity for empathy is limited, especially after hearing the same story over and over again. Cogito analyzes a customer’s spoken cadence and volume to provide instant insights: “go slower, the customer is not understanding,“ or “go faster, the customer is becoming impatient,” or “repeat that.” We are seeing promising results, including reduced call times and increased consumer satisfaction.
Finally, one of our partners in this work is Vestigo, an assembly of smart investors working to discover the companies that will shape the future of financial services. Mark Casady is a founding partner who, along with his colleagues, have grown Vestigo—based near the Massachusetts Institute of Technology campus in Cambridge—into New England’s largest venture capital firm focused on early-stage financial technology. His view is that consumers are increasingly engaged in reactive purchases—for instance, one click to buy a book on Amazon or a song on Apple Music. Consumers buy or try, and if they like it, they continue to engage. But financial services are different. Rather than being a reactive purchase, choosing a financial product is a highly considered decision. Decisions affecting your money and your life are not easily decided on. Trust and the regulatory environment factor into the decision. “It’s the same revolution as publishing and music, but it’s on a different channel,” Mark says. “Incumbent firms like Guardian who get it will change their activities and adopt new ways to interact with customers. It’s an exciting time.”
Guardian and Vestigo are looking for two types of investments: those that drive down costs and those that delight clients. We’ve long had algorithms that would help with these objectives, but today we have the data and computer power to accomplish them.
Putting on his hat as futurist, Mark envisions two provocative scenarios for the insurance industry. The first requires us to remember what we did before Netflix. If you wanted to watch a film at home over the weekend, you bundled up against the cold and drove or walked to your nearest Blockbuster video store. You rented a film or two and drove home. Today many pay a monthly subscription fee to Netflix for access to countless films streamed digitally to any device. What if the same were true for life insurance? Rather than paying a price determined at one place and one time in your life, what if you paid a subscription that could adjust as your life needs changed? What if it were real-time? Technology would know you’re going on a trip to, say, Spain, and adjust your insurance accordingly. Or say you are ride sharing on Uber or Lyft. The subscription model will adjust your insurance to cover you during that ride.
The second idea conjures up images of playing with Legos. What if you could construct and deconstruct your insurance policies—put them together and take them apart for different scenarios? It would help to create a safety net. For example, what if you needed insurance to help you to pay to retrain for a new job?
As leader among businesses, Guardian has to be flexible and imaginative. That is also true for the workforce of the future.
JOB SITES
Few follow labor trends more closely than today’s job sites. Jed Kolko, chief economist for the employment search engine Indeed.com, sees these types of highly specialized innovations as the foundation for jobs of the future. In an interview for this book, he noted that there a lot of attention is being paid to the growth of data scientist jobs, but the real growth will come from people who know how to use what he describes as approachable versions of AI tools such as those we saw in our Shark Tank. We don’t need someone to invent Excel, he notes; we need millions who know how to use it. The same will be true with AI tools.
“Tools will be so widespread you won’t even think of them as sophisticated,” Kolko said. “It’s hard to predict what skills will be in demand in thirty years. But basic quantitative aptitude and communications skills, the ability to translate complicated ideas for different audiences—these are the skills that will remain important.”
Health care and energy will be industries with growing demand and wages, but roles in those industries will increasingly require familiarity with advanced technologies.
LinkedIn, the global professional network, boasts that 165 million workers in the United States have profiles on the site, and members showcase more than 35,000 skills. The September 2019 workforce report showed more than 3 million job listings on LinkedIn. Economists at LinkedIn study skills gaps, which are defined as the gap between supply and demand for a specific skill, in a specific market, at a specific point in time. New York City, San Francisco, Los Angeles, Boston, and Seattle had the largest skills gaps. Cities with the highest skills surpluses were Philadelphia, Chicago, New York City, Detroit, and Minneapolis. New York is on both lists because of a mismatch of skills.41
“Skills gaps can be narrowed in a variety of ways,” according to LinkedIn. “By people moving to cities where their skills are in demand; by businesses opening up shop in cities where there’s an abundance of the skills they need; by training people to learn the skills that are in demand from employers; by employers offering higher pay for in-demand skills.”42
We are in the middle of a technological transformation that feels, at times, like science fiction. We have voice recognition technology that can translate multiple languages in real time; chatbots with natural-language processing that can answer customer service questions; online ads targeted to our precise demographic, geographic, and psychographic profiles; image recognition software that can tag our friends in photos; and algorithms that can predict when factory equipment will need to be serviced or inventory restocked.
There is no question that automation has improved our daily lives as consumers in ways big and small. Already many of us can’t imagine what life would be like without so-called modern-day conveniences—a shocking number of which were invented in just the past fifteen years. It has become second nature to call cars with the tap of a button, watch movies at 10,000 feet, dictate texts to our friends, and order groceries without setting foot outside.
In this era of rapid innovation, in only a few years, industries that have remained largely unchanged for a century are being disrupted. The transportation industry offers an apt example. For almost one hundred years, New York City’s yellow cabs had a near-monopoly on the taxi business. But in 2011, Uber’s ride-sharing app went live, and in less than six years, its ridership exceeded that of the iconic yellow cabs.43 In this case technology disruption didn’t claim jobs. The skills needed to drive a taxi are the same as those needed to drive an Uber. In fact, some taxi drivers also began driving for the ride-sharing app.
However, the transportation industry is far from done with disruption. With the advent of self-driving cars on the horizon, bigger changes are ahead. Experts are predicting that as soon as 2025, autonomous vehicles could replace 300,000 driving jobs a year—with long-haul truckers being the first to be displaced.44
But even though people frequently point to self-driving cars as evidence of mass-scale disruption, what’s often overlooked is the breadth of jobs that will be created as autonomous vehicles become advanced enough for widespread adoption. ZipRecruiter, an online job board, has already reported a 27 percent uptick in jobs related to autonomous driving.45 Although most of the need right now is for engineers, when autonomous vehicles hit the streets and highways, an entirely new sector of skilled workers will be required to support them: people to manage the fleets, technicians to fix bugs and breakdowns, and remote operators to control traffic and monitor safety.46
Autonomous vehicles will open up opportunities in service-related industries, too. Without drivers and the typical safety concerns that plague automobile travel, self-driving vehicles present greater opportunities for entertainment systems, consumer experiences, and midride conveniences, ranging from WiFi-connected and fully equipped workspaces to manicures and massages. A study from Intel has declared this the “passenger economy,” predicting that it will grow into a $7 trillion industry.47
Changes like the ones happening in transportation are rippling across industries. New techniques in agriculture promise to revolutionize farming and increase the world’s food supply.48 Checkout kiosks in retail shops are already automating point-of-sale interactions.49 And along with self-driving cars, drones and delivery robots are demonstrating new ways to transport goods.50
These technologies and others have the potential to catalyze economic growth, but they also pose urgent problems for our workforce—the very workforce on which our economy depends.
TECHNOLOGY’S EFFECT ON SOCIETY
As economists dig deeper into the collateral effects of automation, they have carefully studied which jobs, tasks, and skills are at the greatest risk of disruption. And they’ve identified a disturbing trend for middle-skill, middle-income jobs.
These jobs require more than a high school education but less than a college degree. Although the roles are diverse—from administrative positions in accounting and law to assembly-line work—their common thread is repetitiveness. With increased efficiency due to automation, demand for middle-skills jobs is plummeting,51 and the poles of our workforce are drawing farther apart.
According to MIT economist David Autor, our labor market is beginning to take on a U shape, with the highest-demand jobs at either end of the pay scale. On one side we have lucrative knowledge sector professions in law, medicine, and data science. On the other we have low-wage jobs in agriculture, caretaking, and maintenance, with median wages that hover around $10 an hour.52
There’s a reason for this. AI is not yet advanced enough to take over the functions of high-skill jobs that require critical thinking and unstructured problem solving—at least not yet. And on the opposite end of the spectrum, AI has difficulty performing work that requires mobility and social interaction. Machines can’t handle cleaning hotels or taking care of the elderly, for example—and these jobs traditionally don’t pay well enough to be profitable for companies to automate.
This hollowing out of middle-skill jobs can best be described by Moravec’s paradox. In the 1980s, AI experts discovered that robots had a critical shortcoming. Though they could perform complex calculations and high-level cognitive tasks with ease, they struggled with activities that you or I would find intuitive. Hans Moravec, a researcher who helped to articulate the paradox, said, “It is comparatively easy to make computers exhibit adult-level performance on intelligence tests or playing checkers, and difficult or impossible to give them the skills of a one-year-old when it comes to perception and mobility.”53
Partly as a result of this U-shaped shift, the working and upper classes have grown over the past decade while the middle class has shrunk.54 Even though technology has become a major economic driver of this shift, the benefits have not been realized equally. As AI advances, the problem could get worse.
This poses a significant challenge to our economy and society. Economic inequality takes a toll on economic growth by contributing to unequal access to education, jobs, and basic services such as health care, as well as driving down participation in the economy.55 From 1990 to 2010, according to the OECD, growing inequality in the United States resulted in a loss of five GDP points per capita.56
Society, and particularly, American society, thrives on the promise that each generation will be better off than the one that came before. That’s a promise I believe we can keep.
First, we need a plan.