Executives are blaming layoffs on AI, but research shows AI is ‘not the main driver’ of US labor slowdown
Source: Yahoo Finance
Published: 2026-05-16
Entity Analyzed: General Knowledge Worker Category
URL SCAN
AI continues to be blamed for a spate of job cuts, with Cisco the latest tech firm to jump on the trend this week. But a new analysis from researchers at the New York Fed suggests that may not be due to the technology itself. The Fed researchers scrutinized vacancies in fields considered vulnerable to automation and found that while there had been a relative decline in postings for occupations with higher AI exposure, the trend had pre-dated ChatGPT. The divergence between high- and low-exposure occupations began before 2022 and does not show a clear additional break in trajectory after 2022.
The Triage
This is the rare article that actually reads the data instead of the press release. The New York Fed — not a random think tank, the actual central bank of the world’s financial capital — analyzed job postings in the most AI-exposed occupations and found that the hiring decline started before ChatGPT existed. The divergence between high- and low-exposure occupations began pre-2022 and stabilized after 2023. If AI were gradually displacing workers, the gap would keep widening. It isn’t. Oxford Economics notes AI adoption is “mainstream in leading sectors” but “usage still appears relatively low,” explaining muted aggregate impacts. Goldman Sachs adds that labor market mismatch has actually declined in AI-exposed fields. The triage: the panic is running ahead of the physics. But panic has its own physics.
The Autopsy (with DT-LAG)
Mechanical Collapse Point
The mechanical reality is that labor markets do not move at the speed of headlines. Cisco can cut 4,000 jobs on Wednesday and blame AI on Thursday — that is a capital decision, not a labor market measurement. The Fed’s job posting data measures what employers actually do, not what they say at earnings calls. What it shows: hiring slowed in early 2022 (pre-ChatGPT), picked up in March 2026 to the best level in two years, and layoff rates remain between 0.9% and 1.2% — a range they have occupied since 2021. The mechanical collapse point is not here yet. The visible destruction is concentrated in press releases, not payroll data.
Lag-Weighted Social Timeline
Immediate (0-6 months): The “AI is killing jobs” narrative accelerates regardless of the data. Cisco, Meta, Amazon, Block, Snap — every layoff announcement now carries the AI justification as standard boilerplate. This creates a self-fulfilling panic: workers respond to the narrative, not the BLS tables. Resignations rise. Career pivots accelerate. The labor market is being reshaped by fear before it is reshaped by automation.
Short-term (6-18 months): The Fed’s finding that “the gap stabilizes after 2023” becomes impossible to ignore. If the displacement thesis were correct, the gap would widen as AI capabilities compound. The stabilization suggests something else: the first-stage deployment coincided with labor shortages in AI-exposed occupations (as Goldman Sachs notes), meaning AI filled gaps rather than eliminated roles. But Goldman Sachs also warns: “the next stage of deployment will likely require more adaptation by the workforce.” The lag is not absence. It is staging.
Medium-term (1-3 years): The narrative-data divergence becomes politically potent. If aggregate job losses remain modest while executives continue blaming AI, the credibility gap widens. Workers will correctly identify that the blame-shifting is a pretext for capital reallocation, not technological displacement. This could accelerate unionization, regulatory pushback, or talent flight from AI-boasting employers. Alternatively, if the next deployment stage does produce measurable displacement, the current research will be memory-holed and the panic will be retroactively validated.
Long-term (3-7 years): The Hemenway Falk/Tsoukalas automation arms race model meets its test. If competitive pressure drives displacement beyond what is individually rational, the current gap between narrative and data closes violently. The stabilization the Fed observed could be the calm before the compound-capability storm. Or it could be evidence that AI is genuinely labor-augmenting at current deployment levels, and the obsolescence narrative was premature.
Lag Factors
Narrative Velocity vs. Data Velocity: Headlines move in hours. BLS data moves in months. Fed research moves in years. By the time the data contradicts the panic, the panic has already restructured behavior.
Executive Blame-Shifting: Cisco, Block, Snap, Meta — the AI justification is now a standard line item in layoff communications. It costs nothing to say and deflects shareholder scrutiny from capital reallocation to technological inevitability. The lag factor is that workers internalize the blame while the data says the blame is misplaced.
First-Stage Fortuity: Goldman Sachs’ observation that “the first stage of AI deployment has been fortuitously timed because it coincided with a labor shortage in the most AI-exposed occupations” is crucial. AI did not displace workers because there were no workers to displace. The lag is not technological; it is demographic. When that shortage ends, the math changes.
Information Sector Churn: The one sector where AI adoption is high and labor churn is spiking is the information sector. Oxford Economics notes that “the net change in jobs — hires minus layoffs — remains little changed,” but the velocity of turnover is accelerating. If this churn pattern bleeds into multiple sectors simultaneously, unemployment rises not from displacement but from skills mismatch at scale.
Defensive Moats
Data Literacy: The Fed paper is a moat for anyone who reads it. The evidence that AI is not yet the main driver is a shield against premature career abandonment. But it is a thin shield — the next deployment stage may invalidate it.
Labor Shortage Cushion: The fortuitous timing Goldman Sachs identified is a moat. AI-exposed occupations started with shortages, which blunted displacement. As those shortages resolve, the cushion deflates.
Skills Mismatch Decline: The unexpected finding that labor market mismatch has declined in AI-exposed fields suggests workers are adapting faster than the narrative assumes. This is a moat — but one that requires continuous adaptation to maintain.
Physical World Inertia: Real-world hiring processes, credential requirements, and organizational structure still create friction. The Fed data shows this friction is substantial enough that even when AI capability exists, adoption lag is significant.
Future-Proofing Scorecard
| Timeline | Score | Commentary |
|———-|——-|————|
| 1 year | 4/10 | The narrative-data divergence persists. Some workers benefit from the calm while others panic-exit. The information sector churn pattern is the warning signal. |
| 2 years | 3/10 | If the next deployment stage hits during resolved labor shortages, the gap closes. The Fed’s stabilization finding becomes a historical footnote. |
| 5 years | 2/10 | Either AI proves labor-augmenting (scored too low) or the compound-capability storm arrives (scored too high). Uncertainty is the only certainty. |
| 10 years | 1/10 | The current research will be cited by both sides: optimists saying “we warned you it was early” and pessimists saying “we warned you it was coming.” Both will be right about different things. |
The Verdict
This article is the first credible data-driven challenge to the “AI is destroying jobs” consensus, and it comes from the New York Fed — an institution with no incentive to sugarcoat labor market dysfunction. The verdict is uncomfortable for both sides: the panic is premature, but the staging is real. AI is not yet the main driver of labor slowdown. But executives are treating it as if it were, and their behavior is reshaping the labor market through narrative force before technological force arrives. The discontinuity is not in the data. It is in the discourse. The gap between what executives say and what the Fed measures is the actual battlefield. Workers who read the data instead of the press releases gain a tactical advantage — but only if the next deployment stage does not close the gap faster than they can adapt. The verdict: believe the numbers, but watch the next stage. The numbers are about the past. The next stage is about everything.