Goldman Sachs: The U.S. labor market is healthier now than when ChatGPT launched. Yes, really

Goldman Sachs says its occupation-level mismatch index has fallen from the 2022 peak and is now slightly below pre-pandemic levels, with a one-standard-deviation rise in AI substitution exposure linked to 12% fewer job openings versus 2019. The improvement is attributed to AI reducing openings in already-short occupations, partially offsetting post-pandemic labor distortions. Goldman warns the next stage of AI deployment may be more disruptive as it reaches roles that are not already in surplus, especially entry-level white-collar jobs.

Analysis

The key market implication is not that AI is benign for labor, but that it is first acting as a valve on an already-overheated white-collar labor market. That means the initial macro data can look improved even while the underlying reallocation pressure is intensifying; in other words, the first-order effect is disinflationary and efficiency-enhancing, while the second-order effect is a slow erosion of entry-level knowledge work. The setup is especially important because the jobs absorbing displaced labor are more credential- and location-constrained, which raises friction costs and lengthens unemployment spells for younger cohorts.

For equities, this is a relative-value story more than an outright market call. Software, staffing, recruiting, and office-services adjacent businesses face a delayed but meaningful demand impulse as headcount growth becomes more selective and employers substitute software for junior support layers. Meanwhile, AI infrastructure beneficiaries still have room, but the labor-market angle argues that capex adoption may broaden beyond the obvious hyperscalers into enterprise workflow automation, where ROI is clearest if companies can eliminate expensive coordination labor.

The contrarian risk is that the current improvement in mismatch is a temporary “good timing” artifact. If AI diffusion expands from substitution into broader task automation before displaced workers can retrain, the labor market can worsen without a recession in aggregate employment first—an important regime shift because it would pressure consumer confidence before payrolls break. That would be bearish for cyclicals with exposure to white-collar hiring, but supportive for firms selling credentialed training, workflow software, and physical labor substitution.

Near term, the signal is too early to short the labor market broadly; the cleaner trade is to fade labor-intermediation winners and buy AI productivity enablers. Over 3-12 months, the market is likely underestimating the hit to entry-level hiring and the resulting compression in white-collar wage growth, which should show up first in job postings, then in wage inflation, and only later in headline unemployment.

Key Decisions for Investors

• Initiate a relative short in staffing/recruiting exposure vs AI workflow winners: short MAN or RHI against long MSFT/ORCL on a 3-6 month horizon; thesis is slower white-collar hiring plus software substitution compresses placement volumes before it hits unemployment data.
• Add to quality AI productivity beneficiaries on weakness: long GS/QQQ names with enterprise automation exposure for 6-12 months; benefit from higher adoption as firms try to preserve margins with fewer junior hires.
• Use put spreads on office-adjacent labor intermediaries over the next 1-2 quarters if job postings continue to roll over; risk/reward favors defined downside because the market is likely still pricing a soft landing in white-collar demand.
• Overweight physical-presence, credentialed labor themes via selected health-care and maintenance services names; these are slower to automate and should see steadier demand if office-entry jobs keep shrinking.
• Avoid broad bearish macro trades off this report alone; wait for confirmation in postings and wage data. The better catalyst is a second leg down in AI-exposed job ads, which would support a trade from a labor-friction rather than recession angle.