Worker Types, AI Exposure and the Recent Decline in Job-Finding Rates | Richmond Fed

TEXT START: The recent decline in unemployment outflow rates is concentrated among specific groups of workers, rather than spread evenly across the labor force.

The Dissection

This brief converts an aggregate unemployment problem into a diagnostic instrument. It isolates worker attachment and occupational AI exposure, then identifies a sharp post-2023 pattern: job-finding rates deteriorated most for strongly attached workers and for those in highly AI-exposed occupations. The 13-point decline for primary workers versus 2 points for secondary workers is the important finding.

The article is building a cautious bridge from correlation to technological explanation. Its language — consistent with AI affecting the labor market — is deliberately narrower than its evidence’s implication. It also defers the decisive question to a future decomposition of aggregate versus technological effects.

The Core Fallacy

The central error is scale blindness. The brief treats declining job-finding as a labor-market allocation problem to be decomposed, with displaced workers potentially reallocated toward less-exposed occupations. Under the Discontinuity Thesis, that is only the first visible symptom of a capital regime change. Once AI achieves durable cost and performance superiority across cognitive work, less-exposed sectors cannot absorb the displaced majority at equivalent scale, wages, or necessity.

The brief’s AI measure is also an exposure proxy based on patent-task overlap, not proof of deployment, causality, or durable superiority. It detects the wound. It does not yet establish the full killing mechanism.

Hidden Assumptions

• Less-exposed occupations have enough capacity to absorb workers displaced from exposed occupations.
• Reallocation preserves economically meaningful employment rather than merely producing lower-paid residual work.
• Occupational exposure measured through AI patents tracks actual workplace substitution.
• The post-2023 divergence is primarily technological rather than a mixture of untested sectoral, cyclical, demographic, or institutional effects.
• Aggregate and technological components can be cleanly separated.
• Job-finding rates remain an adequate measure of labor-market health after AI changes the need for human labor.
• Strong labor-market attachment makes primary workers resilient, rather than making them the next high-value target for automation.
• Historical recession patterns remain a valid baseline for a discontinuity that may be structurally unlike previous downturns.
• Human institutions can preserve stable, human-only economic domains at sufficient scale.

Social Function

Classification: partial truth with a transition-management function and a prestige-signaling wrapper.

The empirical result is real and materially important. But the institutional framing narrows a systemic break into measurable labor-flow categories, allowing the system to continue administering the decline as a policy variable. It identifies who is being locked out without confronting whether the economy still requires most people’s productive participation. This is not simple copium or propaganda; it is a technically serious early-warning memo whose language remains safer than its implications.

The Verdict

This is a credible early-warning instrument, not a complete autopsy. Its own findings show the hiring gate closing first on workers who are most attached to employment and most exposed to AI — precisely where a cognitive labor shock should appear. It supports the early signature of P1, but it does not establish P2 or P3 by itself.

The brief still assumes a labor market capable of reallocating the casualties. Under the Discontinuity Thesis, that assumption is the corpse being measured. The standard recession story is failing: the gate is closing before the factory visibly empties.