The AI Layoff Trap
Source: https://www.semanticscholar.org/paper/d3a791d037376e859f9e6e8e114cbf3c8c28ea78 ↗
Full text: open-access via OpenAlex ↗
Falk and Tsoukalas construct a task-based competitive model to show that the real problem with AI-driven displacement is not ignorance but a classic collective action failure: each firm rationally automates while the aggregate result destroys the consumer base everyone depends on.
The paper's analytical contribution is the systematic demolition of proposed remedies — wage flexibility, UBI, upskilling, Coasian bargaining, capital taxes — leaving only a Pigouvian automation tax standing.
For product directors who regularly sit inside the competitive logic the paper describes, this is a rare case where formal economics makes the strategic trap legible rather than merely naming it.
It belongs alongside Brynjolfsson on the productivity paradox and Coase on the firm as the intellectual grounding for why markets in technology do not self-correct toward socially optimal outcomes.
The paper directly fills the library's identified gap in platform-economics and behavioral-economics by addressing the structural incentive architecture of automation decisions, not their downstream consequences.
Central argument
Falk and Tsoukalas argue that competitive markets create a demand externality that traps rational, fully informed firms in an automation arms race that harms everyone—including firm owners. Each firm captures the full cost savings from replacing workers with AI but bears only a fraction of the resulting demand destruction, since displaced workers spend less across all firms. The distortion is not a coordination failure solvable by communication or agreement; it is a dominant-strategy equilibrium that only a Pigouvian automation tax can correct, while widely proposed remedies—UBI, upskilling, capital taxes, worker equity—leave the core incentive intact. Crucially, better and cheaper AI widens the gap rather than resolving it, via a Red Queen dynamic where perceived competitive gains from automating faster cancel out at equilibrium, leaving only more distortion.
Critique
The model is deliberately stripped of several real-world dynamics—most notably, it assumes zero capital-income recycling in the baseline and that displaced workers cease spending entirely on the sector's output, which likely overstates demand destruction in industries where firm owners are also significant consumers. More substantively, the paper's central policy prescription—a Pigouvian automation tax calibrated to the uninternalized demand loss per task—requires regulators to measure a counterfactual (what aggregate demand would have been without the automation) that is deeply difficult to estimate in practice, and the paper does not address how the tax rate would be set, updated, or administered across heterogeneous industries, leaving the normative conclusion somewhat detached from implementable policy design.
Why it matters for product
For a CPO, the paper reframes a common build-vs-hire tradeoff: replacing support agents, ops staff, or junior PMs with AI tooling is not just a cost decision but a systemic one—if the industry converges on the same playbook simultaneously, it erodes the addressable market itself, particularly for B2C and SMB-facing products where end users are also workers in adjacent sectors. The Red Queen finding is directly actionable in product strategy: investing in 'better' AI capabilities to outpace competitors does not produce lasting differentiation at the market level and may accelerate collective harm, which argues for sequencing automation around genuine new-task creation—new product surfaces, new user segments—rather than pure headcount substitution. It also has organizational design implications: the paper suggests that teams with P&L accountability for both cost reduction and demand generation are structurally better positioned to internalize this tradeoff than those measured only on efficiency metrics.