The AI layoff trap

Jorge Costa Oliveira

It is worth reflecting on a recent study by Brett Hemenway Falk and Gerry Tsoukalas of Boston University, which highlights what they call “the artificial intelligence layoff trap.” The paper asks a simple question: if companies know AI-driven layoffs will hurt consumption – and their own revenues – why do they continue?

Because they have to.

The study shows the economy risks a vicious cycle of automation. When firms replace workers faster than displaced employees can find new jobs, consumer demand weakens. Those workers were also customers. Less income means less spending – and that feeds back into lower revenues.

Yet no firm can step aside. If one company cuts costs with AI, it can lower prices and gain market share. Competitors must follow or fall behind. What looks like strategy is really survival.

The authors describe this as a demand externality. Firms capture the benefits of automation but share the damage from lost demand across the market. The gain is private; the cost is collective.

This drives a race to automate beyond what makes sense overall. Every firm knows excessive automation erodes demand, but none can slow down alone. The result resembles a prisoner’s dilemma: rational individual decisions produce a bad collective outcome.

Importantly, the losses are not just a transfer from workers to shareholders. Demand destruction is a net loss of economic welfare. Value disappears – hurting workers and firms alike. That erosion compounds over time, quietly weakening the broader economic base.

The more unsettling finding is what happens as AI improves. Intuition suggests better AI should ease the problem. The model shows the opposite.

More powerful AI strengthens the incentive to replace human labor. Each firm expects a competitive edge. But when all automate, those gains cancel out, leaving deeper demand destruction. The authors call this the “Red Queen effect,” referencing the Lewis Carrol’s Red Queen – running faster just to stay in place.

More competition makes it worse. In fragmented markets, demand loss spreads wider, widening the gap between individual decisions and what would be collectively optimal.

The study also reviews public policy measures on whether they correct the competitive incentive that leads to excessive IA and automation. Universal basic income supports consumption but does not change incentives. Capital taxes redistribute profits but do not affect automation decisions. Worker ownership reduces inequality but not demand loss. Retraining helps, but cannot fully offset permanent income loss. Even voluntary limits between companies are likely to fail under competitive pressure.

The only solution that works is a Pigouvian tax on AI and automation – a levy on each automated task reflecting the broader demand loss it creates. Named after Arthur Cecil Pigou, it forces firms to bear the full social cost of AI and automation.

Properly designed, it would align private incentives with the public good and fund retraining or income support to rebuild demand. Without such intervention, the trajectory is clear: faster automation, weaker consumption, and diminishing returns even for the firms driving the shift.

The takeaway is blunt: the problem is not just technological unemployment, but the competitive system driving it. AI is not only innovation – competition leads to a market failure unfolding in real time, and one policymakers can no longer afford to ignore. Regulation, including fiscal measures, is required to solve the trap.

linkedin.com/in/jorgecostaoliveira