AI Abundance Requires a 4-Hour Workweek

Last year, walking through downtown New York, I saw a ClickUp ad on an MTA bus promising it can help me save one day per week. Then, a few months ago, I saw a newer promise that it would save 6-7 days per week. At this rate, next year's ad would promise to save 30 days per week, and the year after, 365. It seems this is the future we are being sold: tools, chatbots, and agents so good that productivity savings will bend the space-time continuum. If this ends up being even directionally true, why would we still demand people to work 40 hours per week?

The AI-maximalist pitch is that AI will transform the economy into one of abundance. It will enable teams to run leaner and the companies to do more. But the dominant narrative says far less about how those gains would benefit workers. The implicit message the executives are sharing is that the future is exciting, for them and their shareholders, and that the workers, the ones who are left employed anyway, should be grateful to be included. In their worldview, if AI leads to a 10x productivity improvement, they see a proportional more leverage and profits, not their employees working less. To that, I say no. If AI were indeed to deliver a 10x productivity boost, it should also lead to a 4-hour workweek.

71% of U.S. adults expect AI to reduce the number of U.S. jobs over the next twenty years. That expectation is understandable in light of history: they saw technological progress create enormous wealth, but the gains have been distributed unevenly. Total family wealth nearly quadrupled between 1989 and 2022, but 61% of the $147 trillion increase accrued to the top 10% of families, while 6.5% accrued to the bottom half. The sentiment is different in China, where 85% see AI offering more benefits than drawbacks, versus 38% in the United States. It's not that Americans are more pessimistic. Most living Chinese adults encountered technology as part of a four-decade modernization cycle during which almost 800 million people escaped extreme poverty. Americans, by contrast, have lived a recent history in which mobility declined and the rewards of technological progress increasingly accumulated at the top.

The construct of a 40-hour week only feels immovable because we were born after the fight that set it into law 80+ years ago. Weekly working hours fell from much higher levels over the last two centuries. But the businesses didn't become kinder on their own. People bargained and pushed to redefine what could be asked of them. The shorter week arrived through conflict.

A review of 183 industries from 1987 to 2015 found that productivity growth outpaced compensation growth in 83% of them, and that labor’s share of income declined in 77%. Productivity gains do not automatically become worker gains. Those gains can become profit, reinvestment into the business, lower prices, shorter weeks, public services, reduced hours, or some mix of those things. But by default, the majority of the dividend flows toward output and concentration; toward better tools for businesses to extract more value from the workers.

We don't know for sure if AI will lead to 10x productivity. Anthropic CEO Dario Amodei predicts AI could wipe out half of entry-level white-collar jobs within one to five years, and, within five to ten years, cure most diseases. I think this is conceivable, but perhaps not on this timeline. But we should be ready regardless, so that if it happens this fast we would ensure that the outcome is fair to all.

On the other side of today’s AI frontier duopoly, Sam Altman says that AI hasn't and won't lead to a 4-hour workweek because people are competitive and always want more. This is what the AI leaders are pitching to their Fortune 500 customers: AI spending will be offset by productivity gains and leaner operations, and the fear of displacement will force obedience into the workers.

In the past year, American resistance to AI is increasingly visible around data center build-out. 57% say they would oppose one in their community, and at least 100 localities have adopted moratoria against building them. This strategy is adding friction, but I doubt it would be an effective or lasting one. When Jensen Huang predicts that AI could expand global GDP from $100 trillion to $500 trillion, rest assured that business interests will find a way to pursue it. There does not seem to be a ceiling for how far they would go to materialize their vision, when their stated Plan B is to literally fly data centers into freaking space if those pesky townspeople wouldn't accept building them in their backyards.

A core duty of government is to secure the conditions for a safe and decent life. But the current discourse around AI regulation is centered on model safety and the negative effects of the infrastructure build-out. These are important topics, but even more pressing should be finding an answer to what the economy would look like for its people if we indeed march into the future the AI research labs are aiming for.

Talks of having public stakes in AI companies, whether it's 5% or maybe 10%, is a good start but insufficient, since even a 10% public stake would leave the overwhelming majority of AI-created wealth in the hands of the few. We need regulation that funnels the majority of the upside to the people, while still allowing them to retain a purpose in life. A world in which everyone has a house, a car and no purpose is hollow and dehumanizing. And according to Mark Zuckerberg, AI will even work 24/7 on our behalf to “improve” our hobbies, whatever that means. The government should take a for-the-people approach in designing a world post-AI-abundance: and etch into law progressively shorter standard hours without proportional pay reductions.

If AI gives executives a 10x machine and gives workers no time back, then we built an unjust engine for the same type of extraction we have seen before. This isn’t what abundance should look like. The dividend from AI should not be framed only as 10x output. It should be framed as more life.