Researchers tied to Anthropic modeled what happens if AI takes over half of cognitive work by 2030. In the worst case, national unemployment hits 11.9%, a number Miami's own labor data has been circling for months.
The working paper comes from five economists, including Anton Korinek and Charles I. Jones, affiliated with Anthropic's economic institute. Released in September, it's explicit about what it isn't. Not a forecast, and not a prediction.
Three Scenarios, One Widening Gap
The model splits workers into two groups. Cognitive occupations, management, professional, sales and office jobs, made up 62.4% of U.S. employment in 2025. The model assumes AI doesn't directly touch the other group.
In the modest scenario, the economy barely moves. GDP lands 1.6% above baseline by 2030, and unemployment holds near 3.9%.
The substantial scenario is where things start to shift. GDP rises 8.3% above baseline. Cognitive employment falls 3.9%. Labor's share of national income drops about four percentage points. The authors compare that to the full four decades after 1980.
What the Extreme Scenario Actually Shows
The extreme scenario is the one making headlines. AI performs almost half of today's cognitive work by 2030, more than doubling productivity on the tasks it handles.
GDP reaches 32.4% above baseline. Overall unemployment hits 11.9%, driven by joblessness near 18% among workers who started in cognitive roles.
Labor's share of national income falls from 60% to about 45%. Total labor income stays close to its no-AI baseline. That means the additional output flows almost entirely to capital owners, whose income rises 81%.
The authors ran the numbers on what it would take to fix that gap. Transfers equal to about 9% of GDP, roughly the combined size of Social Security and Medicare, could keep cognitive workers whole. That would still leave the rest of the economy more than 20% better off.
They're upfront about the catch. No technological shift in history has come with compensation at that scale.
What Miami's Own Numbers Already Show
MAIN has reported that an MIT labor exposure map estimates AI models could perform work equivalent to 360,000 Miami-area jobs. That's roughly 14% of the local workforce.
That's not the same model as this paper, and the two shouldn't be read as confirming each other directly. But the exposed categories line up closely. Customer service, administrative and entry-level cognitive roles dominate both the national paper's at-risk group and Miami's own local exposure estimate.
Miami Dade College didn't wait for a consensus scenario before acting. Its Applied AI degree track was built specifically for workers moving into roles this exact kind of modeling flags as exposed.
What This Means for Miami's Workforce
The paper's authors are careful to say the real path forward could look like any of their three scenarios. It could also be something the model doesn't capture at all. Robotics, political dynamics, and financial shocks all sit outside its scope entirely.
For Miami, that uncertainty cuts both ways. The city's own labor exposure could track closer to the modest scenario or the extreme one. Nothing in either the national paper or the local map settles which. What both agree on is which workers are holding the exposure if the answer turns out to be the harder one.
Miami's Family Offices Eye the Biggest IPO Ever