The AI Layoff Trap, When Productivity Becomes a Problem

In many ways, a gig economy worker in a third-world country lives better than any 19th century monarch. The average 21st century individual lives decades longer, while enjoying better health and access to far superior medicare. Our infrastructure lets us travel easily across continents, and instantly communicate with people on the other side of the world. We can sit at home and watch an opera in Milan, while enjoying a Thai curry matched with a South African wine. All this has been enabled by technological advance. Starting with the Industrial Revolution, there have been startling gains in productivity across three centuries. New technology goes mainstream and translates into huge improvements in human development indicators. Thus, over three centuries it’s become a truism that the disruptions caused by technological advances eventually make things better: the jobs lost as an old technology is superseded are more than compensated for by new employment opportunities that arise. But what if this time is different? What if the productivity gains are effectively limitless, and the new employment opportunities are net negative? A new paper, “The AI Layoff Trap” by Brett Hemenway Falk and Gerry Tsoukalas, ends with a sentence that is stark and dystopian: “At the limit, firms automate their way to boundless productivity and zero demand.”

The Historical Pattern: Disruption Then Compensation

The arrival of the automobile and the disruptions it caused are often cited. Many jobs and professions disappeared when the horse ceased to be the primary mode of transport. Blacksmiths, stable hands, carriage makers, and horse breeders all lost their livelihoods. But far more new opportunities arose across the new value chain: assembly line workers, mechanics, road construction crews, petrol station attendants, car salespeople, and eventually a vast ecosystem of auto parts manufacturers, insurers, advertisers, and logistics companies. The positive externalities from the availability of easy personal transport created even more opportunities: suburban housing, drive-through restaurants, motels, and holiday travel. The net effect on employment was positive. The standard of living rose.

The pattern has repeated with each major technological shift. The steam engine, the telegraph, the telephone, the computer, the internet—all destroyed some jobs, but created many more. There can be issues during the transition period though. There’s a loss of consumption as old jobs disappear. It’s assumed that this is eventually offset by the new consumption generated by those who benefit from the new opportunities. But if the income lost by displaced workers isn’t quickly replaced by consumption from beneficiaries, there’s a problem. Historically, the transition has been bumpy but manageable. The Industrial Revolution caused social upheaval—the Luddites smashed machines. The Great Depression saw mass unemployment. But in each case, the economy eventually absorbed displaced workers into new roles.

The New Reality: Productivity Without Employment

What if productivity comes without commensurate growth in employment? We’ve seen “pilot” versions of this over the last 30 years, in scenarios that have played out across borders and industries. Manufacturing and information technology (IT) services have moved out of expensive high-wage locations. Much of the turmoil around immigration has its roots in that transition. Low-skilled workers in the US have been left stranded as their employment opportunities vanished. This is a primary reason why non-college graduates in America are anti-immigration. They have seen factories close and call centres move overseas. They have not been compensated by new jobs. The new jobs—software engineers, data scientists, AI trainers—require skills they do not have.

The historical pattern assumed that displaced workers could be retrained. That may have been true when the gap between old skills and new skills was narrow. A blacksmith could learn to fix a car. A stable hand could work at a petrol station. But today, the gap is widening. A warehouse worker displaced by robotics cannot easily become a machine learning engineer. An accountant replaced by AI cannot easily become an AI ethicist. The barriers to entry are higher. The cost of retraining is higher. The time required is longer. And while they are retraining, they are not consuming.

The AI Layoff Trap: A Mathematical Model

The paper, which was published in March 2026, by two academics at the Wharton School, University of Pennsylvania and Boston University, is peer reviewed and mathematically modelled. It has a simple premise. Assume an economy that can produce everything quickly and efficiently. But there is no demand for anything because nobody is employed. The outcome would be catastrophic, not disruptive. This is not a far-fetched scenario. It is a logical consequence of current trends.

Consider a typical corporation. Its goal is to maximise profit. Cost reduction is a primary lever. Labour is a major cost. If AI can replace a human worker at a fraction of the cost, the corporation will do so. This is rational. It is not evil; it is capitalism. Company A fires a large chunk of its workforce and replaces it with AI. Competitor company B does the same. Rinse and repeat. This is perfectly rational—we’re already seeing dark factories (fully automated manufacturing plants) and mass replacement of low-level IT workers and white-collar clerical workforce.

As unemployment rises, consumption drops. A worker who has been laid off stops buying. She cancels her subscriptions, eats out less, postpones purchases. The businesses that depended on her consumption—restaurants, shops, entertainment venues—see their revenues fall. They, in turn, lay off workers. This causes a negative feedback loop where companies cut costs, with more layoffs and more AI. This leads to another round of falling demand, and another round of seeking more efficiency via automation. This could spread across the macro economy as AI improves.

The paper calls this the “AI layoff trap.” Once the cycle begins, it is self-reinforcing. The only way to escape is through an external intervention. But what intervention could work?

Policy Interventions: What Works, What Doesn’t

What sort of policy interventions could break this death spiral? The paper considers concepts like universal basic income (UBI), capital income taxes, worker ownership of equity, and upskilling programmes. The writers believe none of these would work, though they may be wrong.

Universal Basic Income (UBI): The idea is simple: give everyone a regular cash payment, regardless of employment status. This would maintain demand even if employment falls. But UBI is expensive. Funding it would require massive tax increases. And it does not address the social and psychological costs of unemployment—the loss of purpose, structure, and identity that work provides.
Capital income taxes: Taxing profits from automation and redistributing to workers could slow the cycle. But capital is mobile. Companies could relocate to jurisdictions with lower taxes. And high taxes could discourage investment, slowing innovation.
Worker ownership of equity: If workers own shares in the companies that automate, they would benefit from the profits even if they lose their jobs. But most workers do not have significant equity. And equity ownership does not provide a steady income stream.
Upskilling programmes: Training workers for new jobs is the classic solution. But as noted, the gap between old and new skills is widening. And upskilling takes time. In a fast-moving economy, the jobs that exist today may be automated by the time workers are trained.

The only intervention that the researchers think would work is odd. Charge an “automation tax”. Every time a human is replaced by automation, impose a levy to compensate for the lost demand. The mechanism for translating this into demand would be complicated, and I cannot recall it being discussed seriously hitherto. The idea is to make automation more expensive than retaining human workers, or at least to capture some of the productivity gains to fund demand elsewhere. But how would the tax be calculated? Based on the number of jobs eliminated? The wages lost? The productivity gain? And how would it be enforced? Companies could restructure to avoid the tax, outsourcing work to contractors or reclassifying employees.

The paper’s authors are sceptical that any intervention would fully solve the problem. They end on a note of despair: “At the limit, firms automate their way to boundless productivity and zero demand.”

The Terrifying Implication: AI Could Destroy the Economy

There’s nothing illogical about the extrapolation of visible trends to this extreme conclusion. It challenges the historical patterns of the last three centuries but that doesn’t mean it couldn’t happen. The implication is terrifying: AI could destroy the economy simply by being very good at what it does.

This is not a prediction; it is a possibility. It is a scenario that economists must take seriously. The historical precedent is not reassuring. The Industrial Revolution caused severe dislocations. The Great Depression showed that economies can get stuck in low-demand equilibria. The difference today is the speed and scale of change. AI is not a specific technology; it is a general-purpose technology that applies to nearly every sector. Its impact will be broad and deep. And it is improving rapidly.

What Should We Do?

Even if the extreme scenario is unlikely, the risks are too great to ignore. Policymakers should consider a portfolio of responses:

Strengthen social safety nets: Unemployment insurance, food assistance, and housing support should be expanded to cushion the transition.
Invest in education and retraining: The focus should be on lifelong learning, not one-time schooling. Workers should have access to free or subsidised courses throughout their careers.
Encourage worker ownership: Companies should be incentivised to offer equity to all employees. Tax breaks could be linked to employee ownership.
Consider an automation tax as a backstop: Even if it is difficult to implement, the concept should be studied. A modest tax on automation could fund a UBI or retraining programmes.
Promote job-sharing and reduced work hours: If the total number of jobs is shrinking, the available work could be shared among more people. A four-day work week, with government subsidies, could be explored.

The goal is not to stop automation. Automation is inevitable. The goal is to manage the transition. The historical pattern of technology creating more jobs than it destroys may continue. But it may not. We must prepare for both possibilities. The AI layoff trap is a real risk. We ignore it at our peril.

Q&A: The AI Layoff Trap and the Future of Work

Q1: What is the historical pattern of technological disruption and employment, and why might this time be different?

A1: Historically, “the disruptions caused by technological advances eventually make things better: the jobs lost as an old technology is superseded are more than compensated for by new employment opportunities that arise.” The automobile eliminated blacksmith, stable hand, and carriage maker jobs but created far more new opportunities (assembly line workers, mechanics, road construction, petrol stations, etc.). However, this time may be different because the gap between old and new skills is widening. A warehouse worker displaced by robotics cannot easily become a machine learning engineer. The cost and time of retraining are higher. The paper warns that “at the limit, firms automate their way to boundless productivity and zero demand.”

Q2: What is the “AI layoff trap” described in the paper by Falk and Tsoukalas?

A2: The AI layoff trap is a self-reinforcing negative feedback loop. Company A fires workers and replaces them with AI (rational profit-maximising behaviour). Competitor company B does the same. As unemployment rises, consumption drops. Businesses dependent on that consumption see revenues fall and lay off workers. This leads to “another round of falling demand, and another round of seeking more efficiency via automation.” The cycle spreads across the macro economy as AI improves. The paper concludes that “at the limit, firms automate their way to boundless productivity and zero demand.” The economy could produce everything but have no consumers because nobody is employed.

Q3: What policy interventions does the paper consider, and which do the authors believe might work?

A3: The paper considers:

Universal Basic Income (UBI): Give everyone cash regardless of employment. Expensive; doesn’t address social costs of unemployment.
Capital income taxes: Tax automation profits and redistribute. Capital is mobile; companies could relocate.
Worker ownership of equity: Workers benefit from automation profits. Most workers have little equity; doesn’t provide steady income.
Upskilling programmes: Train workers for new jobs. Gap between old and new skills is widening; training takes time.
The only intervention the authors think might work is an “automation tax” —a levy every time a human is replaced by automation, to compensate for lost demand. The article notes the mechanism would be “complicated” and has not been seriously discussed.

Q4: What evidence does the article cite that the disruption of low-skilled workers has already begun?

A4: The article cites two examples from the last 30 years:

Manufacturing and IT services moving overseas: Low-skilled workers in the US have been “left stranded as their employment opportunities vanished.” This is a primary reason why “non-college graduates in America are anti-immigration.”
Current AI-driven displacement: We are already seeing “dark factories” (fully automated manufacturing plants) and “mass replacement of low-level IT workers and white-collar clerical workforce.” A warehouse worker displaced by robotics cannot easily become a machine learning engineer. The author notes that “the gap is widening.”

Q5: What portfolio of responses does the article recommend for policymakers?

A5: The article recommends five responses:

Strengthen social safety nets: Expand unemployment insurance, food assistance, and housing support.
Invest in education and retraining: Focus on “lifelong learning, not one-time schooling.” Workers should have access to free or subsidised courses throughout their careers.
Encourage worker ownership: Incentivise companies to offer equity to all employees; link tax breaks to employee ownership.
Consider an automation tax as a backstop: Even if difficult to implement, the concept should be studied. A modest tax could fund UBI or retraining.
Promote job-sharing and reduced work hours: If total jobs are shrinking, share available work among more people. A four-day work week with government subsidies could be explored.
The article concludes: “The goal is not to stop automation. Automation is inevitable. The goal is to manage the transition.” The AI layoff trap is a “real risk” that “we ignore at our peril.” The historical pattern may continue, but “it may not.” Policymakers must prepare for both possibilities.

The AI Layoff Trap, When Productivity Becomes a Problem