Alex Imas on Why Economists Might Be Getting AI Wrong

The article discusses how AI could be uniquely disruptive to the labor market, with particular focus on which jobs are most at risk and whether the speed of AI development makes this cycle different from prior general-purpose technologies. It is a conceptual interview with no specific company, policy, or economic data point. Market impact is limited, though the discussion reinforces the strategic importance of AI adoption and labor displacement risks.

Analysis

The market is still pricing AI mostly as a software-margin story, but the bigger second-order effect is labor supply reallocation and wage compression in white-collar service sectors. That matters because the earliest winners may not be the model labs themselves, but the firms that can turn headcount replacement into operating leverage fastest: enterprise software, workflow automation, and outsourced business services with high labor intensity. The key asymmetry is that AI adoption can be scaled almost instantly, while labor markets adjust slowly, so earnings upgrades can arrive well before the macro data reflects the disruption.

The more interesting risk is not broad unemployment; it is a short-cycle disinflation impulse in specific services categories that feed into margins and policy expectations. If productivity gains start showing up in legal, customer support, back office, and coding roles over the next 2-6 quarters, that creates pressure on wage growth in subcomponents of CPI and PCE, potentially extending the window for lower rates. That would support duration-sensitive growth assets, but it also means the market could underappreciate the earnings headwind for staffing firms, BPOs, and IT services companies that depend on human labor arbitrage.

A contrarian point: consensus likely overestimates how evenly AI benefits will spread. The first wave should concentrate returns in firms with proprietary data, workflow integration, and distribution, while smaller competitors face margin erosion because they cannot afford the same capex or talent spend. This creates a barbell outcome: a handful of platform winners and a long tail of at-risk incumbents, rather than a broad productivity uplift. The speed dimension is the real catalyst — not whether AI is transformative, but whether adoption compresses the usual multi-year diffusion curve into a 12-18 month earnings re-rating cycle.

Key Decisions for Investors

• Long MSFT vs short WDAY on a 6-12 month horizon: MSFT should monetize AI through distribution and bundling faster, while WDAY faces a tougher sell against embedded workflow automation; target 15-20% relative outperformance if enterprise adoption accelerates.
• Short staffing/services exposure via KELYA or BAH over 3-9 months: these names are vulnerable to early margin compression if clients begin substituting AI for lower-value labor; use tight risk controls because macro softness can mask the thesis.
• Long QQQ or XLK vs short XLI on a 6-12 month horizon: AI-driven productivity is most likely to show up first in tech-heavy earnings revisions, while industrials benefit less from intangible labor substitution; seek a 200-300 bps relative move if rates drift lower.
• Buy deferred call spreads in ADBE or CRM for a 9-15 month window: these are leveraged to workflow automation adoption, but position via spreads to reduce valuation risk; thesis works best if management teams start quantifying AI-driven seat expansion and retention benefits.