A close friend called me over the weekend. He has been quietly anxious for months. Several of his friends in middle management are circling the same worry, asking the same question in coffee break chats and group threads: am I next?

It is not paranoia. It is pattern recognition.

When I saw this paper pop up on my X timeline, I thought, let me actually sit with it and read it properly. The more I read, the more I felt that something important was being said, and most people were walking right past it.

We have all seen the headlines. From tech giants to retail staples, the “AI layoff” is now standard boardroom vocabulary. The logic looks clean on a slide deck: swap expensive human workers for tireless AI agents, cut the payroll, watch the margins grow. Shareholders cheer. The stock pops.

But underneath that strategy sits a massive, mostly invisible trap. The paper is called The AI Layoff Trap by Brett Hemenway Falk (University of Pennsylvania) and Gerry Tsoukalas (Boston University), and it lays out exactly how we walk into it.

The Part Everyone Is Ignoring

We love talking about the supply side. How much more we can produce. How fast. How cheap. AI fits perfectly into that conversation. What we are not talking about, at least not loudly enough, is the demand side.

Here is the uncomfortable truth: displaced workers are not just numbers on a balance sheet. They are also your customers.

When a firm replaces thousands of workers with AI, those workers lose their paychecks. They stop buying things. The individual company saves on wages, but the broader economy quietly loses purchasing power. If every company runs the same playbook, the math gets scary fast: endless productivity, and almost no one left with money to buy the products.

The researchers call this the “demand cliff.” We are not there yet. But we are slowly drifting toward it. And that slow drift is exactly what makes it so easy to ignore until it is too late.

Why Smart CEOs Cannot Stop Even If They Want To

This is the part that should genuinely unsettle you.

You might assume that a smart executive, someone who reads the big picture, would slow down. Surely they can see the cliff?

Maybe. But seeing it does not mean they can stop.

The research frames this as a classic Prisoner’s Dilemma. When a firm automates, it keeps 100% of the cost savings for itself. The hit to consumer demand, however, gets spread across the entire market. Every competitor absorbs a tiny piece of the damage, while the firm that automated pockets all the gains.

In a competitive market, that math means automation becomes the obvious move every time. Companies will keep laying people off even if they privately understand it could eventually hollow out their own customer base. The system makes it nearly impossible to do otherwise without giving up market share.

This is not villainy. This is a broken incentive system doing exactly what broken incentive systems do.

The Red Queen Problem: Better AI Makes Things Worse

Here is where the research takes a genuinely surprising turn.

Most people assume that as AI gets more capable, the problem sorts itself out. Productivity rises, costs fall, and the economy adjusts. The paper suggests the opposite may be true: better AI actually deepens the trap.

The authors borrow the Red Queen effect from Lewis Carroll’s Through the Looking-Glass, where you have to run faster and faster just to stay in the same place. Each firm races to automate to gain a slight edge over competitors. But because every firm is doing it at the same time, the advantages cancel out. Everyone ends up with smaller margins and a shrinking customer base, having gained almost nothing relative to each other while collectively doing real damage to the economy.

The race speeds up. The cliff gets closer. Nobody wins.

Why the Usual Fixes Fall Short

At this point, most policy conversations reach for familiar tools. The paper tests several of them, and the findings are worth paying attention to.

Universal Basic Income provides a real floor for people’s lives, which matters a lot for basic dignity. Elon Musk has actually argued for this on X multiple times, saying it will become necessary as AI displaces more jobs.

“As a reminder, I’m in *favor* of universal basic income”

Elon Musk (@elonmusk), Jul 24, 2020

“There will be universal high income (not merely basic income). Everyone will have the best medical care, food, home, transport and everything else. Sustainable abundance.”

Elon Musk (@elonmusk), May 2025

“Universal HIGH INCOME via checks issued by the Federal government is the best way to deal with unemployment caused by AI. AI/robotics will produce goods & services far in excess of the increase in the money supply, so there will not be inflation.”

Elon Musk (@elonmusk), April 2026

But here is the catch the paper points out: UBI does not change the incentive for a firm to automate. Companies will still race toward the cliff. They will just do it while workers receive a government-funded safety net on the way down.

Retraining programs are useful, but history shows that displaced workers rarely recover their previous income quickly enough to offset the loss in spending power in real time.

Taxes on profits, applied after the fact, do not touch the core decision in the moment, which is the choice to replace a worker with an AI system. By the time the tax arrives, the damage is already done.

The One Fix That Actually Targets the Root Problem

The researchers argue that the only tool capable of breaking this cycle is what economists call an automation tax. The idea is simple: tax the act of automating a job, at a rate that reflects the economic damage that gets passed on to the rest of society.

By building that cost into the decision upfront, rather than afterward, you change the calculation for the individual firm. Replacing a worker is no longer a purely private choice. The wider damage gets factored in. The cliff becomes visible on the company’s own books.

The money raised could then go toward retraining programs that actually move displaced workers into better-paying roles, which over time would reduce the very damage the tax was designed to address. Done right, it is a tool that could shrink its own necessity.

It is not a perfect solution. No policy is. But it is the only one in this analysis that actually targets the incentive instead of just managing the consequences.

The Bottom Line

We are watching a race where the winners may eventually find themselves standing alone in an empty stadium.

Every firm that automates is making a sensible choice inside a system whose rules are producing a harmful outcome for everyone. That is the trap. Blaming corporate greed misses the deeper problem: even well-meaning companies, operating in competitive markets, have almost no reason to behave differently.

My friend’s anxiety is not unfounded. Neither is yours, if you feel it too.

The real question is whether we can build policies that change these incentives before the demand cliff stops being a theory and starts being something we all live through.

Because by then, it will be a lot harder to fix.

Read the original research paper here: The AI Layoff Trap by Brett Hemenway Falk and Gerry Tsoukalas (arXiv, March 2026).