Report from the McKinsey Technical Institute

Automation’s challenge for policy makers, business leaders, and workers everywhere is a formidable one: how to capture the positive effect on the global economy, at the same time as navigating what is likely to be a complicated period ahead, one with potentially epochal social, economic, and employment repercussions.

At a time of sluggish GDP growth and weak productivity gains—and when demographic trends are starting to work against growth in a broad range of countries—automation could serve as an unforeseen boon to the world economy. Yet anxieties about lost jobs and reduced incomes are already creating a backlash against globalizing and modernizing trends, especially in advanced economies, influencing election outcomes in several countries. Ever since the Industrial Revolution, evolving technologies have aroused fears as well as excitement. The risk that automation could become a scapegoat is real.

As already noted, the public debate over automation takes place against a backdrop of a growing gap in incomes and employment prospects of high-skill and low- and middle-skill workers. 102 The share of national income that is paid to workers, the so-called wage share, has been declining in many advanced economies even as productivity has risen, suggesting a disconnect between productivity and incomes, which automation could potentially exacerbate further. The wage-share decline is due in part to the growth of corporate profits as a share of national income, as a result of rising capital returns to technology investments, lower returns to labor from increased trade, rising rent incomes from homeownership, and increased depreciation on capital. 103

Moreover, there is already a significant mismatch of skills in the global workforce, with high levels of youth unemployment and, at the same time, a shortage of job seekers with critical skills. Overcoming this mismatch is a complex undertaking that requires close cooperation among education providers, governments, and businesses. 104

Uncertainty about the timing of automation adoption and its potentially variable impact from sector to sector, from country to country, and from workplace to workplace, make the challenge of preparing for it even more complex. Yet preparation is both possible and necessary, within the business world, at a policy level, and for individuals. Automation technologies are advancing rapidly, and those who harness them effectively and take the lead in their sectors will gain a competitive advantage. 105 It is never too early to think through strategic options and appropriate responses.

102. See the discussion of technical change and skills in Chapter 4.
103. While overall spending on capital goods has been weak, there has been considerable investment in information technology, whose prices have declined. See Loukas Karabarbounis and Brent Neiman, The global decline of the labor share, NBER working paper number 19136, June 2013; Loukas Karabarbounis and Brent Neiman, Declining labor shares and the global rise of corporate saving, NBER working paper number 18154, June 2012; and How CBO projects income, Congressional Budget Office (CBO), July 2013. See also, Poorer than their parents? Flat or falling incomes in advanced economies, McKinsey Global Institute, July 2016.
104. For a detailed discussion of the education to employment transition, see Education to employment: Designing a system that works, McKinsey Center for Government, January 2013. See also, A labor market that works: Connecting talent with opportunity in the digital age, McKinsey Global Institute, June 2015.
105. For details of how companies that are leaders in digitization increase profit margins and raise productivity, see Digital America: A tale of the haves and the have-mores, McKinsey Global Institute, December 2015.


Capturing the full opportunities offered by automation including both the labor substitution and other performance benefits is likely to give companies a competitive advantage, but doing so will likely require them to conduct a thorough review of corporate activities and potential overhaul of business processes and workflows.

Automation will enable new forms of competitive advantage, but it will also require companies to raise their game just to keep up

There are many opportunities for companies to take advantage of the potential of automation to seek competitive advantage. Automation of various activities can improve the performance of almost any business process, as we noted in our case studies in Chapter 3. Certainly, automation can be used to transform the costs of a process by reducing labor costs, for example when end-to-end digitization is used to create straight-through processing of a transactional process. As we have also documented, automation can not only enable a reduction in labor costs, it can also bring a range of other benefits related to performance improvements, such as greater throughput, improved reliability, raised quality, better safety, and other gains. Straight-through processing of financial transactions, for instance, is usually faster than the manual process it replaces, and reduces the number of errors introduced into the process.

Thus, any of the benefits that automation can unlock could become a basis of competition. An industrial company that is able to ensure significantly lower downtime than its competitors by automated monitoring and predictive maintenance of its equipment can compete on this basis of better reliability. A consumer company that can provide faster delivery and 24×7 customer service, through automation in its supply chain and contact centers, can compete on the basis of being more responsive to its customers. Automation can also enable companies to create new products, services and/or business models, for example, a professional services company that is able to provide customized advisory services to small and medium-sized businesses, or even consumers, through an automated conversational interface.

Some of the strategic capabilities that automation can unlock are more subtle and wide ranging than improving the performance of a particular process or offering. Some forms of automation, for example those that are based on machine learning techniques such as deep learning, improve their performance over time when they have access to more data. Companies that are able to create platforms with increasing returns to scale can create network effects that result in winner-take most dynamic. Machine learning on user data allows the platform to become more compelling to users, which in turn generates more data. 106 We have seen these types of dynamics in online search platforms, social media platforms, media delivery platforms, and increasingly platforms that support activities in the physical world, such as platforms for Internet of Things data, healthcare, and travel and transportation.

Automation could also unlock the otherwise unlikely combination of scale and agility, with the ability to instantly propagate changes across an entire organization. When an organization’s activities are controlled by automatic systems, modifying the behavior of the enterprise can be accomplished by software download, rather than an extensive change management program. When automation is used to augment human management, traditional organizational orthodoxies, such as about spans of control, can be challenged. For example, Uber takes advantage of automation for coordination, and has only about one human manager per 1,000 drivers compared with a typical limousine company that has about one manager per 20 to 30 drivers. However, greater scale and speed also means that the effects of negative changes are amplified. And these are not the only challenges that automation could bring to businesses. From a technology standpoint, as software and data underlie nearly all new automation technologies, cybersecurity issues become more critical than ever.

Strategically, automation can also heighten competition, enabling startups or firms from other sectors to encroach on new turf and exacerbating a growing divide between technological leaders and laggards in every sector. 107 At the same time automation can support increasing scale, it also can provide the leverage that enables smaller companies to compete with much larger firms. When the basis of competition using automation is algorithms and data, the raw materials of competition are more readily accessible, through the cloud and open data. For example, we have seen the advent of new competitors in the media and game sectors deploying automation tools to quickly gain the reach of much more established players.

The fact that automation displaces human work activities creates a set of real challenges for companies that deploy it. Worker displacement creates the potential for labor unrest, and the size and scope of retraining/placement programs, whether put in place by companies or others, will have to mirror the size and scope of automation programs. The effective use of automation requires the transformation of processes, changing what people do, even those that are not made redundant by automation. In general, workflows will change, and new roles will emerge, such as that of robot trainer or exception handler.

Companies embracing automation could experiment and map areas of likely impact, even as they focus on workplace changes and new skills for workers

Companies who recognize both the opportunities and threats of automation to competitiveness will engage and embrace the potential that these technologies represent, prioritizing a set of active experiments to start climbing the learning curves earlier rather than later. To help diagnose where automation could most profitably be applied to improve performance, business leaders may want to conduct a thorough inventory of their organization’s activities and create a heat map of where automation potential is high. Business processes shown to have activities with high automation potential could be reimagined under scenarios where they take full advantage of automation technologies (rather than mechanically attempting to automate individual activities using current processes). The benefits and feasibility of these automation-enabled process transformations could then be used to prioritize which processes to transform using automation technologies.

Business leaders and their organizations will also need to become more knowledgeable about the evolution of the technologies themselves, understanding the art of the possible, and the potential for the future, in order to best position their enterprises to take advantage of automation. This is not just “book knowledge” that comes from reading about technologies, or visiting global centers of innovation, but practical knowledge that comes from devoting some resources to continually and purposefully experimenting with technologies on real problems, and then scaling those that demonstrate promise.

Perhaps the most vital component to being successful at deploying automation is the hard work that has to be done to prepare and adapt human capital to work in complementary ways with technology. As our activity-based analysis has shown, almost every role will change, and every workflow eventually will be transformed. Workers will have to be continually retrained as the work activities that they do, to work alongside machines, continue to evolve. Others will have to be redeployed, potentially to other positions in the economy, and businesses have a role to play in aiding these transitions. This will require not only changes in skills, but also changes in mindsets and culture, in a world where work activities continue to change, and “co-workers” include not only other people, but also machines.


Policy makers must recognize the pressing need for productivity acceleration to compensate for demographic aging shifts in order to enable GDP per capita growth. Automation technologies can provide a major contribution to accelerating productivity growth. Thus there are two broad categories of issues for policy-makers to consider. First, how can we accelerate the development and deployment of automation to generate greater growth in productivity? Second, how can we support the redeployment to other productive activities of workers whose activities are automated?

Policy makers can accelerate early development and adoption of automation technologies

Early adoption of automation could benefit from policy support, both in regard to the technology development and for its deployment. This support could include investments in developing the technologies themselves, including funding basic research and support for commercialization, as well as supporting investments in digitally enabled infrastructure for automation.

Investment in enabling infrastructure for automation adoption could be an early priority, especially for emerging economies that may not be as digitally enabled as some advanced economies. In general, large scale automation will require substantial investment, and the tax and other treatment of this investment could enable—or hinder—the adoption of automation technologies. For regulators, automation can pose challenging issues for safety and liability; for example, in the case of self-driving vehicles, who could be held liable for accidents—the automaker, the owner, or the algorithm creator? Thoughtful regulatory dialog and policy making will be important to ensure that the benefits of automation are achieved while protecting other societal concerns.

Engaging a broader societal dialog about automation, the need for productivity growth, and shifts in labor markets is another role that policy-makers can play. Deployment of some technologies could face concerted opposition from unions or other labor organizations over concerns about the employment impact. Governments will need to find cogent answers and coherent policies to engage in these debates.

Exposing and stimulating the work that needs doing

Governments are often not particularly able by themselves to anticipate the types of jobs that could be created, or new industries that will develop (and they are not alone in this limitation). However, they are well positioned to catalyze dialogues about what work needs doing, and the grand societal challenges that require more attention and human effort. 108 The 1966 US Report of the National Commission on Technology, Automation and Economic Progress devoted multiple chapters to “Unmet Human and Community Needs,” including sections on education, healthcare, urban transportation, air pollution, water resources, housing, and international development, all of which seem as relevant in this era as they were 50 years ago. 109 Perhaps a similar report today would add caring for the elderly to the list, as we have documented the demographic effects of aging.

While policy-makers might not be able to predict the new activities and occupations that can be created, they can help create the conditions under which innovation in the use of human labor becomes more likely. Governments could also encourage new forms of technology enabled entrepreneurship. Digital technology itself can enable new forms of entrepreneurial activity. Workers in small businesses and self employed occupations can benefit from higher income earning opportunities. A new category of knowledge-enabled jobs will become possible as machines embed intelligence and knowledge that low skill workers can access with a little training. In India, for example, Google is rolling out the internet Saathi (friends of the internet) program in which rural women are trained to use the internet and then become local agents who provide services in their villages through internet-enabled devices. The services include working as local distributors for telecom products (phones, SIM cards, and data packs), field data collectors for research agencies, financial-service agents, and para-technicians who help local people access government schemes and benefits through an internet-based device. There is a need for these services in rural India where broad internet reach and digital literacy are still low, but where an increasing number of services are being provided online. Google’s program aims to create more than 50,000 internet Saathis who will provide services to more than 50 million households in rural India. 110

106. Michael Chui and James Manyika, “Competing at the digital edge: Hyperscale businesses,” McKinsey Quarterly, March 2015.
107. This is already happening with digital technologies. See ibid.
108. Tim O’Reilly, “Don’t replace people. Augment them,”, July 17, 2016.
109. Technology and the American economy: Report of the National Commission on Technology, Automation, and Economic Progress, US Department of Health, Education, and Welfare, 1966.
110. India’s technology opportunity: Transforming lives, empowering people, McKinsey Global Institute, December 2014.

Addressing wages, skills gaps, and labor market mismatches

One of the challenges of the new era will be to ensure that wages are high enough for the new types of employment that will be created, to prevent continuing erosion of the wage share of GDP, which has dropped sharply since the 1970s. 111 While some governments may be tempted to look for ways to slow automation adoption, out of concern for possible employment effects, such moves could prove counterproductive, holding back productivity without protecting jobs durably.

Automation could exacerbate a skills gap, even as it touches all occupations. There is already a growing divide in income advancement and employment opportunities between high-skill workers and those who are low- and medium-skill. In the past two decades, there has been a clear pattern of consistent job growth for high-skill workers and little or no growth for low- and middle-skill ones. For example, in 1981, college-educated workers in the United States earned a 48 percent wage premium over high school graduates. By 2005, that premium had risen to 97 percent—in other words, an American college graduate earns almost twice as much as a high school graduate. 112 The growing gap between productivity and wages is not new, but automation could accelerate the process. In its 2016 report on automation, the White House noted that the trend toward skill-biased change brought about by computerization and communications innovations is likely to continue in the decade ahead as a result of artificial intelligence’s effects on the labor market. 113

To address this gap, policy makers could work with education providers to improve basic skills through the schools system and put a new emphasis on capabilities that are among the most difficult to automate, including creativity, understanding human emotions, and managing and coaching others. For people who are already in the workforce, they could intervene to help workers develop skills best suited for the automation age. For example, many economies are already facing a shortage of data scientists and business translators. 114 Governments working with the private sector could take steps to ensure that such gaps are filled, with new education and training possibilities established rapidly and prioritized. They could also foster the growth of technology-enabled solutions for the labor market that improve matching and access to jobs, such as online talent platforms. 115 As automation reshapes the workplace, independent work could become increasingly important, and policy makers will want to address issues such as benefits and variability that these platforms can raise.

Furthermore, while important work that needs doing might be identified, there is a possibility of market failures in the wages that might be paid for that work, a situation for which public policy interventions might be appropriate.

111. Poorer than their parents? Flat or falling incomes in advanced economies, McKinsey Global Institute, July 2016.
112. David Autor, “Skills, education, and the rise of earnings inequality among the ‘other 99 percent,’” Science, volume 344, issue 6186, May 2014. See also Poorer than their parents? Flat or falling incomes in advanced economies, McKinsey Global Institute, July 2016.
113. Artificial intelligence, automation, and the economy, Executive Office of the President, December 2016.
114. The age of analytics: Competing in a data-driven world, McKinsey Global Institute, December 2016
115. A labor market that works: Connecting talent with opportunity in the digital age, McKinsey Global Institute, June 2015.

Rethinking social support

Full or partial automation will result in labor displacement, and it will be important to support workers as they transition from one set of activities to another. As work evolves at higher rates of change between sectors, locations, activities, and skill requirements, many workers may need assistance in adjusting to the new age. This could involve providing support during transitional periods, for example retraining or income support. While our modeling suggests a higher likelihood of labor shortages than labor surpluses, there might be people whose skills and capabilities are mismatched to the work that needs doing, or where wages are put under pressure by specific increases in labor supply (for example, within a geography, for workers with particular skills, in specific industries). In these cases, adapted social safety nets could help provide support. Various ideas have been considered, including work sharing, negative income taxes, and universal basic income (see Box 8, “When some old policy ideas are new again”).


Regardless of the longer-term implications, in the short to medium term, men and women in the workplace will need to engage more comprehensively with machines as part of their everyday activities. Tighter integration with technology will free up time for human workers including managers to focus more fully on activities to which they bring skills that machines have yet to master. This could make work more complex, and harder to organize, with managers spending more time on coaching.

As young people in particular make education and career choices, it will be important for them to be made aware of the factors driving automation in particular sectors, to help them identify the skills that could be useful for them to acquire from a labor-market perspective, and what activities will be complements of activities that are likely to be automated. 116

High-skill workers who work closely with technology will likely be in strong demand. Those involved in developing and deploying automation technologies will have many opportunities. In addition, workers who are paid to do activities that are complements of automation will also find themselves in an advantageous position, as Brynjolfsson and McAfee have described it, racing with the machines rather than racing against the machines. 117 These and other workers may be also able to take advantage of new opportunities for independent work as the corporate landscape shifts and more project work is outsourced by big companies. Low-skill workers working with technology will be able to achieve more in terms of output and productivity but may experience wage pressure given the potentially larger supply of similarly low-skill workers.

Education systems will need to evolve for a changed workplace, with policy makers working with education providers to improve basic skills, with a new emphasis on topics such as creativity, emotional intelligence, and leading and coaching others. For all, developing agility, resilience, and flexibility will be important at a time when everybody’s job is likely to change to some degree.

Finally, automation will create an opportunity for those in work to make use of the innate human skills that machines have the hardest time replicating: social and emotional capabilities, providing expertise, coaching and developing others, and creativity. For now, the world of work still expects men and women to undertake rote tasks that do not stretch these innate capabilities as far as they could. As machines take on ever more of the predictable activities of the workday, these skills will be at a premium. Automation could make us all more human.

116. Erik Brynjolfsson and Andrew McAfee, The second machine age: Work, progress, and prosperity in a time of brilliant technologies, W. W. Norton & Company, 2014.
117. Ibid.

Box 8. When some old policy ideas are new again

Many of the potential policy measures that could be adopted to help the labor force adjust to the impact of automation are not entirely new. The 1966 US Commission on Technology, Automation, and Economic Progress recommended taking actions that included improving education and training, facilitating better matching between workers and work (including greater transparency for workers), creating portable benefits that follow workers across different jobs, and increasing work-hour flexibility. These are all ideas that find echoes in today’s discussions around the world. 1

Another idea that has returned is providing a universal basic income, in other words, providing all citizens with an unconditional sum of money. Automation has given it a new lease of life among policy makers, some academics and a number of business leaders in Silicon Valley, although it remains controversial. 2 In a June 2016 referendum, Swiss voters overwhelmingly rejected a proposal to establish a universal basic income. 3

A full basic income program has never been enacted and properly studied. However, in Finland, an experiment that started on January 1, 2017, will pay an unconditional basic income of 560 euros per month for two years to a random sample of 2,000 individuals drawn from current working-age beneficiaries of unemployment benefits. The experiment is aimed at comparing the employment rate of beneficiaries of the basic income with those who receive traditional unemployment benefits. 4

Others have suggested that if we need human labor working alongside automation to achieve economic growth, social assistance programs should incentivize work, such as negative income taxes. The history of a negative income tax for low-paid workers spans back to the 1940s, when it was proposed by British politician Juliet Rhys-Williams, and it was advocated by Milton Friedman in the 1960s. In 1975, the United States introduced a negative income tax, the earned income tax credit, which provides income subsidies to the working poor. The program has survived for 40 years and today annual payments range from $500 for an individual with no children earning less than $14,820, to $6,242 for a family with three or more children and household income of less than $53,267.5 Other countries have similar programs.

1. See, A labor market that works: Connecting talent with opportunity in the digital age, McKinsey Global Institute, October 2016; Independent work: Choice, necessity and the gig economy, McKinsey Global Institute, October 2016; and The world at work: Jobs, pay and skills for 3.5 billion people, McKinsey Global Institute, October June 2012.
2. See for example, Charles Murray, “A guaranteed income for every American,” Wall Street Journal, June 3, 2016. Among business leaders, Elon Musk has spoken out in favor of such a program.
3. “Swiss voters reject proposal to give basic income to every adult and child,” Guardian, June 5, 2016.
4. Preparations for the basic income experiment continue, Kela, December 14, 2016. In the 1970s, Canada launched a five-year experiment in guaranteed basic income, known as “Mincome,” in Dauphin, Manitoba. The poverty level decreased, hospitalization rates fell, and high school completion rates rose. The drawback was that non-primary income earners (often mothers of small children) dropped out of the labor force. See Evelyn L. Forget, The town with no poverty: A history of the North American guaranteed annual income social experiments, University of Manitoba, May 2008.
5. US Internal Revenue Service.

Considerable uncertainties surround the advent of the automation era—and have done so for years. A half century ago, a US commission on technology, automation, and economic progress wrestled with some of the same questions about the future of work and employment that we do in this report.118 The speed with which automation will be adopted into the workplace will vary, and the effects on employment, on national economies, and on businesses and workers globally will play out in myriad ways. At its core, however, automation represents a considerable opportunity for the global economy at a time of weak productivity and a declining share of the working-age population. For corporate leaders, too, automation will reshape the business landscape and create considerable future value. How to capture the opportunities and prepare for the possible consequences will be a key political, economic, corporate, and social question going forward. This is not something that can be watched from the sidelines. Automation is already here, and the technological advances continue. It is never too early to prepare.