While hand-wringing about AI’s implications for work and labor markets is understandable, economic history and current data make clear that the most common fears are largely overblown. Current applications are nowhere near levels that can threaten jobs or even have a noticeable effect on productivity statistics.
LONDON – Artificial-intelligence technologies are advancing both quickly (in terms of widespread adoption) and in big steps. Hardly a day passes without some commentary in the media about the latest AI breakthroughs and their connections with robotics, the metaverse, and the world of work. Some are optimistic, focusing on the potential to raise productivity, eliminate boring tasks, and free up time for more enjoyable activities. Others are pessimistic, focusing on the potential to destroy jobs and steer societies in unknown directions. Both views have elements of truth. But the most important point is often forgotten: AI was invented and developed by humans, and humans can control its development and uses.
Workers’ biggest fear about AI is that it will take their jobs by doing what they do better. After all, employers always want to maximize productivity, and if capital can perform tasks faster and better than labor, they will deploy it for that purpose. But this fear is overblown. It has been around ever since industry replaced agriculture as some countries’ main production activity, and it has sometimes received support from famous economists; yet the supply of jobs has never shrunk because of mechanization.
The usual result of mechanization is a rise in productivity and incomes, and a rise in the economy-wide demand for labor, which is offset by higher wages. Although AI, and especially robotics, is currently being employed at increasing rates in industrialized countries, unemployment is at record lows. The reason, it seems, is that the mechanization-driven increase in productivity and economy-wide demand causes a structural transformation in the economy. For example, we might see a fall in labor demand in the sector that introduces the new technology, followed by a rise in labor demand in many other sectors. Historically, adjustment to such structural change has been relatively easy, without resulting in long-term unemployment.
That brings us to workers’ second common fear: That they will have to find a new home in another sector of the economy, about which they currently know nothing. Those employed writing country reports for senior management – which ChatGPT can now do – may have to pursue jobs in the health sector, where AI has yet to gain as much traction.
This fear is also overblown, though admittedly not as much as the first one. It helps to remember that transitions driven by different rates of technological progress in different sectors already happen all the time. The most sweeping shifts have been the decline of agriculture, the rise (and fall) of manufacturing, and the rapid growth of services. Most of the affected workers naturally do not like such transitions, but society is made better off by the process. Moreover, the changes tend to be gradual; and in most cases, the affected workers move to related activities that do not actually require big jumps in skills.
Of course, the worry now is that AI is an exception. Though it is difficult to measure if AI is indeed causing bigger shifts than earlier technologies did, there might be something to the claim. The movement from agricultural laborer to factory worker might sound like a big deal, but it tends to involve one type of unskilled task being swapped for another. Generative AI and related technologies, the argument goes, could replace one skilled job with another requiring altogether different skills.
Nonetheless, in a recent study based on millions of online job postings in the United Kingdom, the Institute for the Future of Work found that while there are some changes in skill requirements, mainly in the information-technology sector, most of the skills that employers want are not changing. These include building good customer relations, communicating clearly with others, and some basic IT competence (word processing, spreadsheets, and so forth).
SMOOTHING THE TRANSITION
Governments have an important role to play in managing technological shifts and labor-market adjustments, and in mitigating workers’ fears. If workers leave one sector for another, they may need some training and social support. To achieve a good match, such policies must be generous but also well-targeted to avoid creating work disincentives. Some northern European countries, such as Sweden and Denmark, have developed successful frameworks for managing these shifts.
One lesson from their experience is that policies should support workers generously for a short time during their job search – say, six months; but if they fail to find a job, additional support should be conditional on attending a sponsored training program in the new skills required by the growing employment sectors. The short duration of the unconditional support acts as an incentive to look for a job, whereas the training helps unsuccessful workers become more employable. Many other countries, however, will need to upgrade their support systems. Much of the fear generated by AI can be minimized if workers know that the government will support them in adapting.
But the more common risk confronting workers concerns changes within their current companies. Even if one’s job is secure, the adoption of AI technologies introduces a new source of uncertainty. In advanced post-industrial economies that have completed the transitions from agriculture to manufacturing and then to services, “role transitions” are the most frequent effect of technological progress and diffusion.
In our work with British data, we found that jobs involving a larger share of tasks that can be automated tend to offer workers less satisfaction, and that job satisfaction improves when companies use digital technologies that workers understand (such as email or smartphone apps). These findings suggest that workers’ fears are not so much about digital technology as about uncertainty.
This is a problem that employers can do something to ameliorate, even though most do not. The key is communication. Companies need to keep their workers informed of plans and proposals that come out of the boardroom, including through frequent informal interactions between workers and line managers. Managers who pay attention to suggestions from their workers about role adjustments are more successful both in employee retention and in adoption of new technology. In job-satisfaction surveys, workers consistently rank good social relations at work very high.
Looking ahead, successful companies will be those that set aside some time every week for training – a process that workers should own by pursuing their interests based on a menu of options from the company. A sense of autonomy can go a long way toward mitigating the fear that workers have from new technologies. Most workers know more about their abilities and preferences than anyone else, and if they are well informed about their company’s plans, they will be more likely to see technological change as an opportunity for career advancement, rather than as a threat.
That brings us to a final unjustified fear about AI: that it will be adopted at work before anyone even has time to react. Yes, AI is spreading fast, perhaps faster than any other recent technology. But to paraphrase the late Robert Solow, you can see it everywhere except in the unemployment statistics. Although companies are talking about it and making plans for it – and though many claim to be using it – the current applications are nowhere near levels that can threaten jobs or have a noticeable effect on productivity statistics. Even in the financial sector, fintech accounts for only a small fraction of business, most of which is still carried out by traditional banks using traditional methods.
The biggest justified fear, then, is not that current workers will lose their jobs, but that young people (either in school or just starting a career) are failing to equip themselves with the right tools for the future. On one hand, perhaps the focus should be less on STEM (science, technology, engineering, and math) – which involves the kinds of tasks that AI excels at – and more on preparing for the kind of work that requires empathy. On the other hand, without bright minds in STEM, who will develop and adapt future AI to our work needs? Are those now specializing in STEM modern-day Dr. Frankensteins, sowing the seeds of their own destruction?
These are the difficult questions that societies need to confront. AI applications could complement our skills, rather than destroy them – but only if we take pains to bring about that outcome.