Every Business Is Automating. Economists Just Proved That’s the Problem.
Your competitor just deployed an AI tool that handles in minutes what used to take someone hours. Even if you privately think this wave ends badly, for workers, for consumer spending, for the economy, you’re probably still going to do it. Because if you don’t, they win.
That pressure you’re feeling? Two economists just proved it’s not paranoia. It’s math.
What the research actually says
Brett Hemenway Falk (UPenn) and Gerry Tsoukalas (Wharton) published a paper in June called “The AI Layoff Trap”. The argument is that competitive markets push businesses to automate beyond the point that’s good for anyone, including the businesses doing it.
When a firm cuts a worker and replaces them with automation, it captures 100% of the wage savings. But the demand destruction, those wages no longer circulating as consumer spending, gets spread across the whole economy. The firm that fired bears only a small slice of that hit. Its rivals absorb the rest. So every business makes the same rational calculation, everyone automates, and collectively they hollow out the customer base they all depend on.
Tsoukalas put it plainly: “Waiting for the firms to figure it out for themselves, I think, is the worst possible thing we can do.”
The model also shows that better AI makes this worse, not better. More competition makes it worse. The paper frames this as a prisoner’s dilemma, the individually rational move produces a collectively terrible outcome, and no single player can escape by choosing differently.
Why standard fixes don’t actually fix it
UBI, upskilling, profit-sharing, worker equity, the paper works through all of them. None correct the underlying incentive. They address symptoms after the externality has already run. The only intervention that works mathematically is a Pigouvian automation tax: a charge applied at the point of the automation decision, calibrated to the social cost of that demand destruction.
That’s what’s now before the US Congress. H.R. 10044, introduced August 12 by Rep. Greg Casar (D-TX), would levy charges on AI companies’ token sales and product revenue, with escalating rates when unemployment crosses 5%. Whether it passes or not, the political appetite is there.
The Australian angle
Australia has no automation tax on the table, but it has a version of the same anxiety playing out in political geography.
University of Sydney researchers Clinton Free and Lachlan Harris warned in a July SMH piece that a “wired belt” of white-collar Labor seats faces AI-driven displacement that could reshape Australian electoral politics the way deindustrialisation reshaped the American rust belt. Not mass unemployment, just enough economic anxiety to shift votes. Albanese gave a USyd speech on July 14 pledging a “muscular approach to AI” and job protection. Specifics remain vague.
No Australian outlet has yet connected this local political risk to the Falk/Tsoukalas paper. The academic framing, the problem isn’t AI itself but the competitive incentive structure around it, gives the “wired belt” concern a sharper edge. If the trap is structural, Albanese’s muscular speech needs to name the structure.
What this means for SMB owners
You’re running a business where your customers know you by name. That’s a different competitive position to Block or Salesforce, and the paper’s model partially reflects this, large firms in highly competitive markets, with many players and thin margins, face the worst version of these dynamics. Small businesses competing on relationships and local reputation have a buffer the aggregate model doesn’t capture.
That said, the “55% of leaders regret AI layoffs” signal came from mid-market companies that moved fast and found the operational complexity nobody warned them about. Workflow redesign, staff retraining, exception handling, customer trust, all expensive to get wrong, and small businesses are less able to absorb those costs.
The pressure is real. The paper explains why it feels inescapable, because at the market level, it mostly is. The practical question isn’t whether to automate. It’s what.
Automate the workflow, not the headcount
The businesses coming out of this period ahead aren’t the ones that cut staff to hit a margin target. They’re the ones that mapped their processes, found the repetitive steps, automated those, and kept the humans on judgment calls and customer relationships. Forbes reported in August that companies winning with AI are measurably not replacing workers, they’re deploying AI alongside staff to handle volume and reduce errors, keeping people on the work that actually differentiates them.
Your competitor automating creates pressure. But matching them move for move on headcount isn’t the play. Automate what makes you slower or more expensive. Keep what makes you better. That’s competing on value, not price, and it’s the only contest a small business can actually win.
Related reading
- The AI Jobs Reckoning: What Australian Workers and Business Owners Need to Know
- Australia’s Biggest Bank Admitted It Moved Too Fast on AI
- Companies That Replaced Staff With AI Are Already Rehiring
- Which Australian Jobs Are Safe From AI, And Which Ones Aren’t
📖 More deep dives: News Deep Dives
📊 Compare AI tools side by side | 💼 Free resources & AI prompt packs
Sources
More hot takes: Hot Takes →
If this interests you
- Is Your AI Subscription Funding Silicon Valley? Australia’s 80-Cent Problem
- Australians Are Using AI Every Day. Their Businesses Aren’t. That’s the Problem.
- Australia’s First AI Royal Commission Has No Mandate for Small Business
More hot takes: Hot Takes →
