The Other Dividend

When machines do more, firms have two ways to pass on the gains: as money or as time. For decades, they've chosen money, and only for themselves. It's time to choose differently.

We've been here before and we made the wrong call

In the 1870s, textile machinery doubled output per worker almost overnight. Factory owners had a choice: reduce the workday, or pocket the difference. They pocketed it. Child labor expanded. Shifts lengthened. It took a century of strikes, legislation, and economic crisis before the 40-hour week became standard.

We are standing at an identical fork. Artificial intelligence, from code-writing assistants to autonomous logistics agents, is compressing what used to take a week into an afternoon. Goldman Sachs estimates that roughly 300 million full-time equivalents could be automated by 2030. And yet the default corporate instinct is the same as it was in 1875: take the gain, shrink the headcount, reward shareholders.

This is not just unfair; it is economically self-defeating. Firms that hoard productivity gains as profit while eliminating jobs are, quite literally, destroying the customers they need to survive.

Productivity is a river with two mouths

Every efficiency gain generated by technology must flow somewhere. The gains can go to capital, higher profit margins, share buybacks, executive compensation, and cheaper products sometimes. This is where virtually all AI gains are going today. Or they can go to labor, higher wages, more jobs, or the same work done in fewer hours. This is where almost none of the AI gains are going today.

The capital-only route has a structural ceiling: it eventually hollows out demand. If the same output is produced by fewer workers earning less, who buys the product? Henry Ford understood this in 1914 when he doubled his workers' wages, not from altruism, but because he wanted them to be able to buy a Ford. The logic is unchanged. Destroy the wage base and you destroy the market.

"An economy that makes things no one can afford to buy is not a productive economy. It is a very efficient engine of collapse."

The argument for distributing gains as time, reduced hours at unchanged pay, sidesteps the wage negotiation entirely. You don't need to transfer a single dollar. You simply ask, if the same work now takes 30 hours instead of 40, why are we still working 40? The 10 hours are freed. Workers spend more time in communities, families, and creative pursuits. Local economies, restaurants, childcare, music, and repair shops, benefit from the attention. Demand shifts rather than contracts.

The four-day week is not a theory anymore

Between 2020 and 2024, four-day week trials ran across Iceland, Japan, the UK, Portugal, Germany, Australia, and the United States. The results were striking. Revenue and output stayed flat or rose in 92% of UK trial firms. Burnout fell 71% and resignations dropped 57%. Job applications at participating firms rose 88%. Unpaid care work redistributed more evenly between partners. Commuting emissions dropped roughly 20% per participant.

Iceland's national trial, covering 2,500 workers across the public sector, was independently judged an overwhelming success. The government subsequently shifted the majority of the country's workforce to shorter hours by right. Japan's Panasonic, Hitachi, and Recruit Holdings all moved to optional four-day policies and reported improved output-per-hour metrics.

The pattern is consistent: when hours fall, workers strip out waste, needless meetings, redundant reporting, and performative busyness and protect actual productive time. AI accelerates this. When a language model writes the first draft, compresses meeting notes, and automates the status report, the case for a 40-hour week becomes genuinely hard to make.

Why capital-only distribution is a coordination failure

Here is the trap: each individual firm is rational to retain AI productivity gains as profit rather than redistribute them as time. If your competitor does not reduce hours, you lose by reducing yours first. This is a textbook prisoner's dilemma: individually rational, collectively catastrophic.

When every firm follows the same logic, the aggregate result is more unemployment, weaker demand, higher inequality, and eventually a consumer base too financially stressed to purchase the products those firms sell. The short-term quarterly win produces the long-term strategic loss.

This is not a moral argument. It is a systems argument. The economy is a closed loop. Firms are not external to it; they are nodes within it. A node that extracts without recirculating eventually starves the network it depends on.

The escape from this trap has historically required one of three things: coordinated sectoral bargaining, regulation, or culture change within firms themselves. In the AI era, we may need all three, and quickly, because automation is not arriving gradually. It arrives in capability jumps.

The four counterarguments, honestly addressed

"Workers in AI-adjacent roles are more productive, not replaceable." 

True for a subset, software engineers, analysts, and lawyers who use AI tools well. But the gains at the top of the skill distribution do not offset displacement at the bottom. Warehouse pickers, truck drivers, paralegals, call center agents, these are not complementary to AI. They are targets. The distribution of gains must account for the full labor market, not just its most educated decile.

"Firms need profit to invest in AI R&D." Fair. No one is arguing for zero profit retention. The argument is against only retaining profit. A firm could keep 60% of productivity gains as capital and distribute 40% as reduced hours. The present distribution is approximately 100/0.

"People want more money, not more time." Survey after survey says otherwise. After a salary threshold of roughly $80,000, additional income ranks below autonomy, purpose, and time with family in employee preference. The firms that will win the talent war of the 2030s are the ones that offer something money cannot buy back: hours.

"This is a government problem, not a corporate one." Governments are too slow. AI capabilities are outrunning regulatory cycles by years. Firms that wait for legislation before acting responsibly are betting that the political fallout of mass displacement lands on someone else. It will not. Populist backlash against technology firms, already building, will accelerate exactly as unemployment does.

What firms should actually do

The argument is not utopian. It requires firms to make one deliberate accounting choice: when measuring AI-generated productivity gains, ring-fence a portion to reduce working hours proportionally.

In practice: measure actual output per hour, not just revenue. Set a time-reduction trigger. When output per person per week rises 10%, reduce the standard week by one hour. Negotiate industry-wide, not just firm-wide, so that a shorter week becomes a sector norm rather than a competitive disadvantage. And lobby for enabling legislation that makes hour reduction the path of least resistance, instead of tax codes that reward capital gains and penalise payroll.

The real question isn't whether AI creates abundance. It already does. The question is whether that abundance becomes leisure for everyone or profit for a few. That is not a technology question. It is a political economy question, and firms are answering it right now, by default, every quarter they report earnings.