Automation, augmentation, and the future of work

June 5, 2019

This is an extract from “AI and human capital“, the third report of the four-part series, “Asia’s AI agenda”, by MIT Technology Review Insights.

Much has been written about technology’s disruptive effect on employment markets, with the world’s research houses making wildly varying predictions about the numbers of jobs that AI could destroy or create. The World Economic Forum, in a study of 15 countries, estimates that AI will destroy 75m jobs and create 133m new ones by 2020. Management consulting firm McKinsey & Company predicts that by 2030, 400-800m jobs will be eliminated. One University of Oxford study estimates that there will be 67.7m job losses within the US alone over the next 20 years.

The trouble with the varying estimates, says Michael Priddis, chief executive officer at Faethm, is that “divergence masks the truth, which prevents action.” Faethm is a software-as-a-service analytics platform that combines proprietary and public data sets with its customers’ workforce databases, which delivers detailed scenario modeling about the talent- and finance-related implications of a wide variety of technology deployments.

The company’s customers include governments, companies, and universities seeking to understand the impact of emerging technology on industries and job roles, develop informed strategies, and make technology investment decisions.

The outlook for Asia

Faethm’s platform can analyze any workforce data set to create a deconstructed, task-level view of specific roles. It uses information about the technological, sociopolitical, and demographic environment to predict how that job will be affected once specific AI capabilities are introduced. The platform enables policymakers and technology executives to understand the degree to which the job will be completely automated (making the human redundant), enhanced or “augmented” with AI (the human is supported with AI), and how much capacity is created for full-time-equivalent tasks that free up people to pursue other activities.

Analysis of 11 Asia Pacific markets shows that AI will impact one in five jobs—removing some but enhancing others. On average, 12% of current jobs in Asia will be removed by automation within five years’ time. Eight percent of Asia’s current jobs will benefit from, and be augmented by, AI capabilities. In all but one of the economies, the number of jobs AI will eliminate will exceed the number of jobs it enhances. In countries that have large process driven sectors such as manufacturing, twice as many roles are eliminated as enhanced. These include emerging markets such as Vietnam and Indonesia, but also Japan, which already has an advanced and automation-driven manufacturing industry. In Australia, AI will augment as many jobs as it will eliminate (11%). This shows that the economy is both mature and diversified with large employment pools in knowledge-intensive and labor-intensive sectors.

Developed, middle income, and emerging markets

Priddis is sanguine about the impact of AI on developed Asia. “I’m broadly optimistic about the way that developed countries will emerge from the fourth industrial revolution,” he says, “because most share a half-dozen supporting attributes: benevolent governments that collaborate with industry to promote economic growth, capital to invest in new growth, an education culture, and social contracts that allow for [citizens’] provision of care.” Notwithstanding the stress and inconvenience of finding new employment for those displaced, he argues there will be new jobs: “You have skills, and because of that there’ll be more work in other areas for which you can retrain.” Automation will also be an opportunity to manage the growing shortage of manpower. South Korea and Japan (the latter has seen its population shrink by an average of 200,000 people each year since 20102) will see more than 20% of job roles in industries such as manufacturing and logistics automated within five years.

AI will automate a higher percentage of roles in developed Asia (14%) than in the region’s emerging markets (10%). However, many more jobs (11% of the total) in mature Asian economies will also benefit from augmentation than their less-developed peers (6%). In Priddis’ view, it is this combination of automation and augmentation that will drive AI to become a net positive for workers in higher income countries, where knowledge-intensive sectors make a greater economic contribution and generally constitute a higher percentage of overall employment than in developing markets. Knowledge jobs, creativity, problem-solving, and interaction with other people are not skills easily replaced by AI, but they can be well-supported and enhanced with technology.

In contrast, policymakers in Asia’s emerging and middle-income economies will be facing a more worrying reality. Having benefited from Western companies’ past offshoring and outsourcing strategies, these jobs will be readily replaced wholesale by AI and other forms of automation. “If you work in manufacturing in Bangladesh, if you work in the call centers in Manila, if you work in a BPO in India—you’re in real trouble,” says Priddis. The speed and scale at which these transitions are happening is also likely to take governments by surprise, potentially causing economic and social shocks leading to new patterns of migration.

The Philippines, India, and Vietnam, for instance, each will experience a workforce automation rate of between 8% and 9% considering their large, automation-susceptible labor pools. All three countries will see 14% of their manufacturing jobs disappear, but manufacturing is a much greater employer proportionally in Vietnam and India than in the Philippines. Yet the Philippines has a large population of workers in the business process outsourcing (BPO) industry, with equally as many jobs at risk there.

Is this different from previous waves of automation? Yes, says Priddis, in that “the speed of change is greater, the extent of globalization is greater. The geographic implications are different.” In previous waves of automation, the relative capital costs were high. “What we’re talking about now,” he says, “is very, very capital-light—anybody can be involved in RPA [robotic process automation] work. The accessibility, the speed, the low cost mean that it is different this time around.” In preparing for the next wave, many Asian governments remain “systematically unprepared.”