Let tech give us back hours and days to do things that matter more.
Many new technologies inspire anxiety and fears, but AI suffers from a catastrophically poor public image. According to a Pew poll released this August, 55% of young Americans are more concerned than excited by the increased use of AI in daily life, compared to just 11% who are more excited. And 71% of all U.S. adults think AI use will lead to fewer jobs in the next couple of decades. Opposition to local data centers is just the most obvious sign that public unease may derail AI deployment more broadly, as calls grow for further pauses on AI development and restrictions on its use.
These are real concerns that stem from the discourse produced by AI companies themselves. Firms like OpenAI and Anthropic have spent considerable amounts of time and energy emphasizing the potentially existential risks from AI as well as near-term job losses that may result. As a counterbalance, the leading AI companies have also focused on fundamental scientific progress, with promises, for instance, to expand drug development. However, recent advances in mathematics, to name just one example, have been mitigated in public impact by the outcry from mathematicians worried about the implications of AI advances for the broader culture of mathematics.
The only hope for sustainable AI development is that the benefits it generates must be broadly felt — and I don’t mean by making it easier to manipulate photos, but through more profound transformations. One way to achieve that is to ensure that we receive an AI dividend in the form of time.
More time to be free
Taking the benefit of technological advances in the form of leisure has a long history in America. From the Industrial Revolution through the 20th century, advances in economic growth reached everyday Americans in the form of steadily lower working hours, which declined from an average of about 2,900 a year to 1,800 in 2023 (roughly a 35-hour week). These advances were spurred by a mix of private-sector action, labor victories and labor regulation. Henry Ford lowered working hours in his factories from 48 hours to a five-day, 40-hour workweek in 1926 without reducing pay, bolstered by the productive advances enabled by mass automation. The Fair Labor Standards Act of 1938 set a 44-hour week, eventually falling to 40. And Congress over the decades legislated a number of new federal holidays designed to create three-day weekends. Americans not only enjoyed a higher standard of living, but they also increasingly had the time to enjoy it even as full-time workers.
However, this pattern started to shift in the 1970s. A paper by Alberto Alesina, Edward Glaeser and Bruce Sacerdote documents how working hours started to stagnate in America, despite improvements in working conditions. Notably, European countries continued to register increases in leisure hours in this time period (they also saw larger increases in life expectancy, so total leisure enjoyed in retirement years expanded as well), in the form of longer maternity and paternity leave and vacations guaranteed by law. Germans, for instance, work just 1,340 hours a year — over 400 hours less than the typical American. Yes, it’s true that European pay and the European standard of living are lower in some notable respects, but many Europeans and Americans argue that the trade is nevertheless a good one.
What happened in Europe that passed the U.S. by? Alesina and co-authors emphasize a “social multiplier” through which institutions such as unions push for leisure hours with society-wide implications, because it shifts cultural norms and people generally prefer to take vacations together. By contrast, many Americans don’t even take all of the vacation time they are entitled to, reflecting social differences in how people approach leisure.
Part of this leisure gap with Europe can also be attributed to what Claudia Goldin has referred to as “greedy jobs” — white-collar employment in which client services demand that workers be always available, and which are rewarded through outsize compensation. Think of the tournament-style competition to become a law partner or investment bank managing director. To some extent, being always available is simply a precondition of offering client services in these domains. However, some of it is also surely an inefficient race between struggling workers to compete for the top job.
New York City, of course, is filled with these strivers. Cities are productive because they put ambitious people close to each other, who push and drive each other to succeed. These very same professions, notably, are also now in the midst of a complicated AI disruption. This means that cities are ground zero for managing the impact of AI, and it is critical to get the right balance here of taking advantage of AI’s productive advances in broadly socially beneficial ways.
How we should react to AI
Given this history, how should we, as a society, react to the advances of AI technology? Just four years into the revolution sparked by the release of ChatGPT, AI seems able to solve open mathematical problems that have eluded the best mathematicians. More prosaically, it can handle a range of common business tasks like drafting legal briefs, conducting financial analyses and working through complicated software engineering tasks. There are early signs that professions with the most AI use are producing productivity gains. One response to these dramatic shifts may be to simply take advantage of this technology and do more with it, working ever harder in an attempt to grow ever more productive. Indeed, we’ve seen some firms and employees report burnout as they spend all day querying large language models and activating legions of agents, all hoping to avoid getting behind.
We cannot stop some people from grinding this way if they so choose, and surely some of this work will give rise to great new firms and products. But we should also be conscious about what this vision of the world offers its citizens, who are going to be in a position to either advance or halt AI advances through their actions at the ballot box. If the best that AI can offer workers is a choice between endless work or permanent unemployment, it will be no surprise to see them vote for halting technological advancement altogether.
The solution, according to some such as Daron Acemoglu and Simon Johnson, is to try to steer AI in ways that avoid the technological replacement of human labor. This view has some appeal, by seeking to avoid the type of labor-market disruption which seeds political agitation. However, it is questionable how feasible this actually is. AI labs are moving ahead, and firms producing open-weight models are close behind. Diverting AI to ensure make-work jobs remain threatens to leave the economy filled with positions characterized by drudgery, so we collectively enjoy worse living standards. It is little better than stopping AI progress altogether.
There is another alternative, however, which involves fixing our institutions rather than unrealistically trying to tame the development of AI.
If the best that AI can offer workers is a choice between endless work or permanent unemployment, it will be no surprise to see them vote for halting technological advancement altogether.
The time dividend
As a concrete proposal, we should seek to make sure that any AI-induced productivity advances are directed toward increases in leisure experienced by individuals. The idea is to help produce broadly experienced benefits of AI use.
How would this work in practice? An initial step would involve making sure every month has one three-day weekend. We are already part of the way there, with Monday holidays in January, February, May, September and October, and fixed-date holidays in June (with the recently added Juneteenth), July, November and December. We should fill in the remaining months of March, April and August with three new Monday holidays to enjoy three-day weekends the full year around. This shift is likely not costless, as it would entail some loss in productive time for workers. However, three working days constitute just 1.2% of the work year, while many analyses of the impact of AI on the economy forecast greater economic value generated from the deployment of AI tools. Moreover, this shift would likely drive increases in travel and consumption, which would further help spread the economy-wide benefits to a broader class of workers.
Another step could be expanding paid parental time for both mothers and fathers. This could be achieved either through federally paid time, or mandates on employers (who should be better able to pay for it, with advances in productivity), perhaps following existing proposed bipartisan legislation. While many employers already provide some paid leave options, the U.S. is an international outlier in providing no federally guaranteed paid leave to new mothers and fathers. Adding more leave here advances a family-friendly agenda, by making sure the gains of AI growth accrue to parents and their children. Expanding these benefits to new workers also benefits even existing working parents who technically have these benefits, but face pressure to come back to the office soon after the birth of a child. Making these benefits more broadly available helps to generate a social norm that they should be used.
Finally, we should seek to lower aggregate working hours in ways benchmarked by improvements in productivity. The European Union guarantees four weeks of paid leave a year, while the U.S. currently offers none. We can start with a two-week minimum, and steadily increase the number of days depending on growth in measured productivity. The exact amount of this increase can be debated, but the broader point is to ensure a precise standard by which productivity improvements flow through to employees.
To help support these efforts, we should think about taxing firms that have a high degree of “greedy” jobs; cities like New York are places where these jobs tend to concentrate. This would help to solve the collective action problem. Workers individually may prefer to engage in less wasteful competition through face time at the office to get the next promotion, but any individual worker who does so simply loses out on that job. Taxing firms as a whole helps ensure the firm weighs the costs and benefits of high-time jobs, and only offers them when the economics truly justify such positions.
Federal law currently requires time and a half after 40 hours for hourly workers, which goes to the worker and can make long hours more attractive. Instead, we could charge a progressive surcharge on employers who have a critical mass of employees with workweeks above 40 hours, and especially for those with weekly hours above 50, along with a changed salary threshold so that salaried professionals are not covered by overtime rules. The idea would be to not stop anyone from working hard, but to instead stop rewarding firms for creating jobs that require a 55-hour workweek.
Conclusion
In 1930, as Britain was sliding into a recession, John Maynard Keynes penned a famous essay, “Economic Possibilities for our Grandchildren.” Even back in that era, something resembling AI anxiety was in vogue. He named a “new disease” which was spreading: technological unemployment, “due to our discovery of means of economising the use of labour outrunning the pace at which we can find new uses for labour.”
Keynes, however, diagnosed this as only a temporary phenomenon. He made two forecasts for the longer time horizon. One was that the standard of living in rich countries would be “between four and eight times as high” in a century. Despite the turmoil over the next century, he was proved correct, or even slightly pessimistic, about this forecast.
But Keynes was badly wrong in his second forecast: that we would take the resulting abundance largely in the form of leisure. “Three-hour shifts or a fifteen-hour week,” he wrote, “may put off the problem for a great while.” In this, he was badly off.
The question is now put to us, the grandchildren whom Keynes imagined, and the lives of our grandchildren yet to come. If AI-driven technological advances continue apace, future working Americans may be able to lead lives far surpassing ours in material abundance. If we focus on our institutions, they may also lead lives of leisure abundance as well.






