Goldman Sachs published two separate findings about AI and jobs in the same week in April 2026, and the conversation almost entirely focused on the first one. The first finding: AI is eliminating roughly 25,000 US jobs per month and creating back about 9,000, for a net loss of approximately 16,000 positions monthly. That number was picked up widely. The second finding received considerably less attention. Goldman’s economists drew on 40 years of individual-level earnings data to quantify what happens to workers after technology displaces them, and the results challenge one of the most repeated assumptions in the current debate about AI and employment: that Gen Z is the generation most at risk.
The two findings together tell a more complete story than either one does alone, and the complete story is more useful and considerably more troubling for a specific group of workers that most of the current coverage is not focused on.
The 16,000 Number in Context
Goldman economists Daan Struyven and Ronnie Walker, whose work was reported by Fortune and Axios, found that AI substitution is eliminating approximately 25,000 US jobs per month, while AI augmentation, the effect of AI making workers more productive and creating demand for adjacent roles, adds back roughly 9,000. The net monthly loss of 16,000 is the headline number.
That figure sits in a broader context the site has been tracking. As we covered in our analysis of the Stanford 2026 AI Index, entry-level software developer employment for workers aged 22 to 25 has fallen nearly 20% since 2022. The Goldman number adds a monthly displacement rate to what the Stanford data described as a structural shift. Together they describe not a projection but a documented, ongoing trend with a measurable pace.
Goldman’s economists were careful to note that their estimate likely understates the offsetting effects. AI infrastructure investment, specifically the data centers, power systems, and chip manufacturing driven by the current build-out, is creating jobs that their methodology does not fully capture. The 16,000 figure is probably directionally correct but is not a precise accounting of total net employment change from AI. It is an estimate of the displacement effect in occupations directly affected by automation.
The Scarring Data and Why It Changes the Picture
The second Goldman report, written by economists Pierfrancesco Mei and Jessica Rindels, draws on four decades of individual-level earnings data to measure what they call the “scarring” effect of technology-driven job displacement. Scarring in labour economics refers to the long-term wage and employment consequences that persist after a worker loses a job to technological change, as distinct from the immediate displacement itself.
The findings, reported in detail by Fortune, show that real earnings for technology-displaced workers grow nearly 10 percentage points less than for workers who were never displaced, measured over the ten years following job loss. Compared to workers displaced for other reasons (economic downturns, company closures), technology-displaced workers suffer an additional 5 percentage points of cumulative earnings underperformance. They take approximately one month longer to find new employment and take earnings hits more than 3% larger when they are re-employed.
The mechanism Goldman identifies is occupational downgrading. Workers displaced by technology tend to slide into roles that are more routine and require fewer analytical and interpersonal skills, because the same technological forces that eliminated their original job also eroded the market value of the specific skills that job required. Their knowledge does not transfer cleanly. The experience they accumulated over years in their previous occupation becomes partially obsolete. They land somewhere lower in the labour market and tend to stay there.
Why Gen Z Is Not the Most Vulnerable Group
This is where Goldman’s scarring data diverges from the prevailing narrative. Most coverage of AI and employment concentrates on Gen Z: young workers entering a disrupted labour market, entry-level roles disappearing, the generation most exposed to automation at the start of their careers. That concern is real, as the 20% decline in junior software developer employment makes clear. But scarring data tells a different story about who bears the worst long-term economic damage.
Goldman’s economists found that younger, college-educated, and urban workers who are displaced by technology experience cumulative earnings losses roughly half as large as other technology-displaced workers over the decade following job loss. The reason is mobility. Younger workers have shorter tenure in their current occupations, which means they have less occupation-specific knowledge to lose. They are earlier in the process of building skills, which means the disruption, while real, arrives before they have invested deeply in a path that is now foreclosed. They can retrain, relocate, and reorient with lower relative cost than workers who have spent 15 or 20 years developing expertise in a specific domain.
The workers Goldman’s data identifies as most vulnerable are those with deep, domain-specific skills in occupations that AI is directly automating: mid-career accountants, experienced paralegals, seasoned financial analysts, senior customer service managers with years of company-specific process knowledge. These workers are not concentrated in the 22-to-25 cohort. They are in the 38-to-55 range, at the point in their careers where their accumulated expertise should be their most valuable asset, and where the cost of retraining, financially and in terms of career trajectory, is highest.
The Retraining Finding That Most Employers Are Ignoring
Goldman’s scarring analysis includes a data point that organisations with any interest in workforce planning should be tracking. Workers who retraining after a technology-driven job loss saw an average 2 percentage point increase in cumulative real wage growth over the following ten years, and their probability of unemployment over that period declined by approximately 10 percentage points.
The effect is meaningful. The ten-year earnings gap between a displaced worker who retrains and one who does not is significant enough to represent a materially different economic outcome. Retraining does not fully close the gap with never-displaced workers, but it substantially narrows the scarring effect.
The practical challenge is that retraining support in the United States remains fragmented, underfunded, and structurally misaligned with the speed of AI displacement. Federal workforce development programs operate on timescales and funding levels designed for the pace of previous technological transitions. The AI transition is moving faster. A mid-career worker whose role is automated in 2026 cannot wait for a two-year retraining program to begin in 2028.
What the Combined Data Actually Shows
The 16,000 monthly displacement number and the scarring research are not two separate stories. They describe the same process at different time horizons. The first tells you the pace at which workers are being pushed out of their current roles. The second tells you what happens to those workers over the next decade if they do not retrain, and who among them is most likely to be permanently worse off.
The workers most at risk are not the ones who graduated into a disrupted market with no prior investment to lose. They are the ones who invested heavily in occupational skills that AI is now devaluing, at a stage of their career where the window for recovery is narrowing. A 24-year-old displaced from a junior analyst role has 40 years of working life ahead to adapt. A 48-year-old displaced from a senior role in the same occupation has a fundamentally different calculus.
The narrative that frames AI job displacement primarily as a Gen Z problem is not wrong about the short-term disruption to entry-level hiring. It is incomplete about where the long-term economic damage will settle. Goldman’s data suggests it will settle on older, more experienced workers in occupations that are being automated from the top of the skill range, not just the bottom. That is a group with significant political and economic weight, and a scarring outcome with consequences well beyond individual career trajectories.
This article draws on Goldman Sachs research as reported by Fortune and Axios. Analysis and interpretation reflect the author’s reading of publicly available information and should not be treated as financial advice.