Artificial intelligence is reshaping labor markets in real time. Goldman Sachs economists and MIT researchers agree the disruption ahead is real but manageable, with one crucial caveat: how companies choose to deploy AI will determine whether this era is remembered as a rising tide or a crashing wave.
The Headline and the Reality
The word "apocalypse" is doing a lot of work in the current AI labor debate. Rapid advances in model capability, accelerating corporate adoption, and a wave of high-profile layoffs attributed explicitly to AI have stoked genuine concern that this technological shift is categorically different from those that came before. Goldman Sachs Research's Joseph Briggs, MIT's Daron Acemoglu, and MIT's Neil Thompson have each examined that claim carefully. None believes an AI job apocalypse is imminent. But their reasons differ, and those differences carry real implications for how advisors and investors should think about labor exposure, productivity, and the AI trade.
Briggs, GS Senior Global Economist, estimates that the AI transition could displace over 9% of the US labor force, roughly 15 million workers. That number is striking. But Briggs frames it clearly in context: displacement would be spread across a 10-year adoption cycle, and the US labor market churns through 25 to 35 million gross job gains and losses every year under normal conditions. "Despite our expectation that AI-related job losses will lead to a meaningful amount of labor displacement," Briggs writes, "we continue to expect that labor market headwinds will be temporary. Key to this view is our expectation that over the long run AI will create many new jobs even as it destroys existing ones."
Acemoglu, MIT Institute Professor and 2024 Nobel laureate in Economic Sciences, agrees the scale of near-term losses will be modest, concentrated in white-collar workers performing "cognitive and routine" tasks "under predictable conditions rather than typically confronting new and unusual things or constant innovation, creation, and intense social interaction or judgment." Customer service representatives and back-office workers, he argues, are most exposed, numbering roughly 8 to 9 million in the US. His estimate for net job losses over the next five years: less than 2 to 4% of the labor force. "The scale of job losses won't be anywhere close to the very large layoffs some are predicting."
Thompson takes the most structurally grounded view. Capability, he emphasizes, is only one step in the chain from technology to labor market change. AI must also perform reliably, access the right information, and be cost-effective. "The impact AI ultimately has on the labor market may not be nearly as large as its impressive capabilities suggest." His preferred metaphor is precise: a rising tide, not a crashing wave. Real change, real churn, but not sudden widespread displacement.
Substitution vs. Augmentation: The Numbers So Far
GS US Economist Elsie Peng brings the most granular read on current conditions. She finds that AI is already doing both things at once: substituting for labor in some occupations, augmenting it in others. The net aggregate effect is small but negative. AI substitution has reduced monthly payroll growth by roughly 25,000 jobs and raised the unemployment rate by 0.16 percentage points over the past year. AI augmentation has added back approximately 9,000 jobs monthly and reduced unemployment by 0.06 points. The result: a net drag of 16,000 jobs per month and a 0.1 percentage point rise in the unemployment rate.
The industries with the highest substitution risk, telephone operators, billing clerks, insurance claims processors, are seeing employment contract. Industries with high augmentation potential, construction managers, education administrators, physicians and surgeons, are experiencing stronger hiring. Crucially, Peng finds that early-career workers are bearing a disproportionate share of the substitution burden, with wage and unemployment rate gaps between entry-level and experienced workers widening by 1.3% and 0.6 percentage points respectively from pre-pandemic averages.
The Longer Horizon and the Investment Read
Acemoglu's longer-term view is the most cautionary. "No general law of economics says that job creation must match job destruction," he notes directly. If investment continues to prioritize replacement over complementarity, if agentic AI matures, if AI-robotics integration advances, losses in the 10 to 15 year window could be substantially larger. The direction of investment matters enormously. A shift toward what Acemoglu calls "machine usefulness" rather than "machine intelligence" could change the outcome materially.
For markets, GS Senior US Equity Strategist Ryan Hammond finds the signal still faint. Only 11% of S&P 500 companies quantified AI labor productivity gains in specific use cases in Q1 2026 earnings calls, and just 2% tied such gains to earnings. Investors are voting accordingly: a basket of stocks with plans to implement AI into internal workflows has traded roughly in line with the equal-weight S&P 500. The AI infrastructure complex, by contrast, has risen 74% year-to-date, outperforming by 65 percentage points. "Uncertainty remains high about how AI will ultimately impact companies' labor force needs and, in turn, corporate profits," Hammond writes. GS's baseline forecasts an eventual 15% lift to economy-wide productivity and GDP from full AI adoption, but Hammond is direct that markets are not yet in a position to reward that forecast.
Five Key Takeaways for Advisors and Investors
1 AI-driven labor disruption is real but gradual. Goldman Sachs estimates up to 9% labor force displacement over a decade, not a sudden shock. The labor market has time, though not unlimited time, to adapt.
2 Augmentation is happening alongside substitution. The net effect today is modest: a 16,000 job monthly drag and 0.1 percentage point unemployment increase. Watch how this balance shifts as enterprise adoption accelerates.
3 The occupational divide matters. Routine cognitive roles face the highest substitution risk. Roles requiring judgment, creativity, and physical presence face less. Portfolio exposure to companies concentrated in the former warrants scrutiny.
4 The earnings signal is not yet there. Only 2% of S&P 500 companies connected AI productivity gains to earnings in Q1 2026. The AI infrastructure trade remains the most tangible beneficiary; the labor productivity trade awaits evidence.
5 The long-run outcome depends on investment choices, not technology alone. Whether AI complements or replaces workers is not predetermined. Policy, corporate strategy, and the direction of AI investment will shape the outcome more than model capability alone.
Footnote:
Nathan, Allison, et al. "An AI Job Apocalypse?" Top of Mind, Issue 149, Goldman Sachs Global Investment Research, 25 June 2026.
