Recently, the phrase “quiet at first, then louder” has become increasingly popular in economic circles. “Cortés Moment.” It alludes to Spanish conquistador Hernán Cortés’ choice in 1519 to burn his own ships after arriving in Mexico, leaving his men with no choice but to advance.
No turning back. No second thought. Just dedication to whatever comes next. The more you sit with the comparison, the more difficult it is to disagree with Moody’s Analytics Chief Economist Mark Zandi’s description of the current state of America’s relationship with artificial intelligence.
| Information Category | Details |
|---|---|
| Name | Mark Zandi |
| Title | Chief Economist, Moody’s Analytics |
| Institution | Moody’s Analytics |
| Field | Macroeconomics, Labor Markets, Financial Risk |
| Key Warning | U.S. businesses are approaching an irreversible “Cortés Moment” in AI adoption |
| Associated Researcher | Antonio Coppola, Assistant Professor of Finance, Stanford GSB |
| Stanford Affiliation | Stanford Institute for Economic Policy Research (SIEPR) |
| Reference Paper | AI-driven macroprudential regulation study (co-authored with Christopher Clayton, Yale SOM) |
| Training Data Used | 14 years of financial holdings data; model tested against 2020 COVID market crash |
| Reference Website | Stanford Graduate School of Business |
Zandi’s worry isn’t theoretical. Businesses have already relocated. They have reallocated funds, reorganized departments, retrained employees, and in certain instances—such as the fintech company Block, which laid off 40% of its workforce—made decisions that are difficult to reverse.
In other words, the ships are already ablaze. The remarkable thing is that no quantifiable productivity gains from AI have yet to appear in the economic data. Before the cards were even on the table, the bets were made.
That discrepancy between dedication and outcome is precisely what unnerves Zandi. The rate of layoffs has increased to its highest level since 2009. The number of technical job openings is declining. For longer than most forecasters anticipated, overall hiring has remained slow. Whether this is a transient adjustment or the first signs of something more structural is still unknown.
However, Zandi has outlined four possible outcomes for this, and he believes that the most likely one is that AI will eventually produce a wave of productivity growth that propels steady economic expansion. What is concerning is what will occur in the years that lie ahead.
He points out that the healthcare industry is currently serving as the final significant buffer in the labor market, taking in displaced workers who have nowhere else to go. That arrangement is delicate. It’s difficult to ignore how much emphasis is being placed on a single industry that is under tremendous technological pressure.
Meanwhile, researchers at Stanford’s Graduate School of Business are addressing a different but related set of questions approximately forty miles south of the location where Zandi delivers these warnings. Antonio Coppola, a faculty fellow at the Stanford Institute for Economic Policy Research and assistant professor of finance, has been investigating whether artificial intelligence (AI) can do something that would have seemed unthinkable ten years ago: anticipate the next financial crisis before it occurs.
Coppola and his Yale colleague Christopher Clayton developed a deep learning model known as a “graph transformer” that was trained on fourteen years’ worth of financial holdings data. The outcomes were truly remarkable. The model correctly predicted trading behavior during the chaos of the 2020 COVID market crash, despite the fact that its training window ended in 2019. This kind of tool must be a revelation for regulators who have long wished they could see the entire financial system in real time.
Coppola, however, takes care to avoid overselling it. Since Robert Lucas brought up the issue in the 1970s, economists have recognized a problem inherent in predictive models: the behavior a model was trained to predict may begin to alter as soon as it is used to influence policy.
Knowing that the model is observing, investors may covertly move risk into areas of the financial system that the model is unable to see. After 2008, traditional banks were subject to stricter regulations, which contributed to the growth of shadow banking, which includes hedge funds, ETFs, and pension plans. Oversight by AI might just hasten that shift into more dangerous areas.
According to Coppola, this could be a “Faustian bargain” for regulators: you might lose causal understanding while gaining predictive accuracy. Knowing that a fire is beginning is only helpful if you know why it started and whether your actions will make things better or worse. He suggests using AI-driven models in conjunction with conventional economic theory rather than in its place. The two strategies only need to be open about what each can and cannot see; they don’t have to be antagonistic.
Irreversibility is the unifying factor between Coppola’s financial risk research and Zandi’s labor market warnings. In both situations, the claim is that AI has surpassed a certain threshold—possibly the institutional commitment threshold rather than the capability threshold. The money has been invested. The structural adjustments are in progress.
The ships have caught fire. Reading both sets of work gives the impression that economists are no longer arguing over whether or not this change will occur. They are attempting to comprehend what happens to the systems and individuals caught in its current. That’s a more pressing query. And the truth is that no one really knows, at least not yet.
