In this 16 page analysis we look at two of the most ground breaking research papers released in early March'26. The first one titled "The AI Layoff Trap" is a working paper written by Brett Hemenway Falk (University of Pennsylvania) and Gerry Tsoukalas (Boston University) which showcases a formal economic model to explain why competitive firms will collectively over-automate using AI even when every firm can see the harm it causes and why most proposed policy remedies fail to stop it. The paper begins with a paradox. Rational, forward-looking firms should recognize that displacing workers also destroys their own customer base. Displaced workers are consumers, and when their income is not replaced, aggregate demand falls hurting all firms. Yet firms are racing to automate anyway. The paper asks: if the cliff is visible to everyone, why do firms race toward it? We then look at another research paper released by Anthropic “Labor market impacts of AI: A new measure and early evidence” written by Maxim Massenkoff and Peter McCrory. The paper acknowledges that AI's rapid spread has sparked a wave of labor market research but cautions that past forecasting attempts have often been wrong. The authors argue that studying AI's effects before major disruptions occur — rather than in hindsight — gives researchers the best chance of accurately identifying any future economic harm. Most existing research measures theoretical AI exposure — what an LLM could do. This paper introduces a new metric called Observed Exposure. The key insight is that theoretical capability and actual deployment are very different things. A task might be theoretically automatable but still not show up in real usage due to legal constraints, software requirements, human verification needs, or simple inertia. The paper finds no current evidence of AI-driven unemployment, but identifies a tentative early signal that hiring of young workers into exposed occupations has slowed. Our own Take We believe employment loss due to AI is real, but extent can be debated ad infinitum. Automation has already eliminated or reduced roles in manufacturing, data entry, customer service, and routine white-collar work. AI accelerates this. Past technological revolutions (steam, electricity, computers) played out over generations. AI may move faster than workers can retrain. Unlike earlier automation, AI threatens lawyers, radiologists, coders, and writers — not just manual workers. Even when jobs aren't eliminated, AI can reduce bargaining power and suppress wages. Workers in routine, repetitive roles such as call centers, data processing, some legal/financial tasks face real disruption now. The biggest risk isn't permanent mass unemployment- it's a painful transition period where displaced workers don't have the skills or support to move into new roles quickly enough. So as a policy response retraining, educational budget allocations towards AI training as well as wider social safety nets become essential to reduce the employment loss impact on economic activity.
