I came across a paper this week called The AI Layoff Trap, by Brett Hemenway Falk (University of Pennsylvania) and Gerry Tsoukalas (Boston University). It put hard economics behind something a lot of us have sensed but struggled to say out loud.
The argument is simple. When a company replaces workers with AI, it keeps all the savings. But those workers were also customers, and the spending they stop is lost across the whole market, not just the company that let them go. Everyone bears a little of everyone else's cuts.
So even when leaders can see where this leads, nobody stops. If you hold back while rivals automate, you lose the demand anyway and miss the savings. So you cut too. They all cut. And here's the sting: the authors show this isn't just money moving from workers to shareholders. Both sides end up worse off. The pie gets smaller.
I'd add something the paper doesn't fully capture, and it's something I see in almost every business case that crosses my desk: the cost of running AI is badly underestimated.
On paper, AI looks cheap. In practice, it rarely is. The build is harder than anyone budgets for, connecting to old systems, cleaning up data, security, change management. The running costs don't sit still either; they grow as usage grows, and they shift every time a model or vendor changes. And someone still has to keep a human in the loop, to watch the machine, catch its mistakes, and answer for them. In a regulated industry, that oversight is a real cost, not a footnote.
A salaried person's cost is roughly steady. AI's cost often climbs the more you lean on it.
Put those two things together and you get a double trap. You shrink the customer base your revenue depends on, and the savings you promised may not even show up. You can end up paying more and selling less.
What's the better path? I'd put it in five words: resourceful augmentation over unsustainable automation.
The most useful idea in the paper supports exactly that. When displaced people move into better, higher-paying work, the whole effect reverses. Automation that lifts people actually grows demand. The trap becomes a flywheel.
For anyone leading AI work right now, three things matter:
Estimate the real cost first. Not just heads removed, but the full price of building, running, and overseeing the system. If the case only works when you assume AI stays cheap forever, it isn't a case.
Automate the task, keep the person. Free people for judgment, exceptions, and the work machines can't do, then let throughput multiply.
Treat reskilling as protecting your own revenue. It isn't charity. It's how you keep your customers.
The companies that win the next decade won't be the ones that cut fastest. They'll be the ones that did the math honestly and chose resourceful augmentation over unsustainable automation, keeping a human in the loop and making their people more valuable alongside AI.
That's not soft thinking. It's the smarter bet.
