Goldman Sachs quantifies AI's speed of impact on employment: affected occupations see average job losses of about 25% over five years

This spring, panic over AI replacing human jobs reached a peak. But as the U.S. economy showed greater-than-expected resilience, that anxiety has clearly cooled—the number of AI-related layoff announcements has fallen sharply from its high, and media coverage of AI-induced job losses has receded to levels seen at the start of the year. However, Goldman Sachs reminds the market: calm does not equal safety; AI's shock to the labor market is far from over—it is merely waiting for a more opportune moment to erupt.

According to the Wind Trading Desk, Goldman Sachs Global Investment Research concluded in an October 22 report: if AI ultimately delivers a 15% productivity gain, historical experience shows that during the transition period up to about 9% of employees may temporarily leave their current jobs; in occupations where technology renders labor roles redundant, employment falls by an average of about 25% over five years.

Current effects are still visible, but narrow in scope. Goldman Sachs says employment growth in the industries most exposed to AI over the past three months was about 10,000 below its long-term trend, though this gap was largely offset by increased hiring in AI-related construction jobs. Cross-sectional services data are beginning to show strain: the correlation between AI adoption and unemployment has turned positive and is statistically significant, while the correlation between AI adoption and employment growth is slightly negative but lacks statistical significance.

Market focus has also shifted to return on investment. Goldman Sachs says current capital expenditure expectations imply that AI users need to spend about $1 trillion a year on AI applications for hyperscale cloud providers and application vendors to achieve the profits the market appears to have priced in. For investors, the key variables are the speed at which AI saves labor costs and whether those savings manifest as layoffs, attrition and job losses.

1% productivity shock: unemployment rises 0.3 to 0.4 percentage points in the short term

Goldman Sachs uses three empirical approaches to measure the historical impact of technology on labor. The first is an aggregate-level technology shock model. Results show that for every 1% increase in labor productivity driven by technological innovation, the peak unemployment rate temporarily rises by 0.3 to 0.4 percentage points, with the effect fading after about two years.

Changes in the unemployment rate understate the actual job impact. Some displaced workers may find new jobs quickly, and some may exit the labor force, which dampens how much the unemployment rate reflects the true shock. Breaking down job flows further, Goldman Sachs finds that each 1% technology-driven productivity gain brings a cumulative increase of 0.5 to 0.6 percentage points in the total job separation rate, while dragging down the total hiring rate by about 0.2 percentage points.

Industry-level results are similar. Using industry-level utilization-adjusted total factor productivity shock data, Goldman Sachs finds that each 1% productivity gain likewise corresponds to a 0.3 to 0.4 percentage point rise in unemployment, a cumulative increase of about 0.6 percentage points in the total separation rate, and a decline of about 0.2 percentage points in the job finding rate.

In Goldman Sachs' baseline scenario, AI may ultimately deliver a 15% economy-wide productivity gain. Based on the historical relationships above, up to about 9% of employees may be displaced from their current jobs during the AI transition, and slower hiring could impose an additional drag of about 3% on overall employment.

Sample of 25 occupations: employment falls an average of 25% five years after a technology shock

Goldman Sachs selected 25 occupations most vulnerable to AI technology shocks, including photo lab technicians, proofreaders and file clerks—a sample representing the cases where technology has the greatest impact on labor demand. Goldman Sachs says this sample is closer to an upper-bound scenario for the jobs AI may affect most deeply.

Results show that employment in these 25 occupations fell by an average of about 25% five years after the technology shock. Photo lab technicians saw the largest decline, with jobs falling nearly 60% over five years, as digital photography and related technologies directly compressed demand for traditional photo processing.

Goldman Sachs' sample of 25 technology-shocked occupations includes:

The common thread among these occupations is that technology directly compressed existing manual processes. Photo processing was hit by digital photography, proofreading by word processing and spelling and grammar checking tools, file clerks by electronic document systems, meter readers by smart remote meters, travel agents by online booking channels, and bank tellers by ATMs and mobile banking.

Job losses mainly come from higher separation rates, with slower hiring a smaller contributor

One of Goldman Sachs' most important findings is that employment contraction under technology shocks comes more from job separations. Job separations include workers leaving their positions through resignation or dismissal; reduced hiring also matters, but to a lesser extent.

In the sample of 25 occupations, the annual total hiring rate fell in 13 occupations, by an average of 0.7 percentage points. By contrast, the annual total separation rate rose in 19 occupations, by an average of 3.8 percentage points.

Examining hiring and separation rates jointly, Goldman Sachs finds that in 17 of the 25 occupations, headcount declines were more driven by rising separation rates. This is consistent with the aggregate and industry models: technology shocks typically show up first in actual job losses, then in slower hiring.

This has direct implications for tracking the AI shock in markets. Goldman Sachs believes AI-related layoff announcements should serve as a high-frequency indicator for observing AI-driven job displacement. If future AI productivity gains materialize, watching hiring plans alone may react too slowly; layoff announcements, job separations and occupation-level headcount changes can provide more real-time signals.

Goldman Sachs also stresses that AI's impact on employment is still at an early stage. Only 23% of U.S. institutions have incorporated AI into some routine workflows, and even this adoption rate overstates the current degree of automation, because most institutions have not yet embedded AI into all applicable processes. Goldman Sachs says most AI-driven productivity gains still lie in the future.

Historical data show that automation shocks are markedly cyclical. When the economy is healthy, automation-related labor disruption is usually limited; when the economy weakens, firms are more inclined to cut labor costs through automation, and job losses accelerate.

Industry models show that when the overall unemployment rate rises, each 1% technology-driven productivity gain has an impact on unemployment 0.1 to 0.2 percentage points higher than the baseline estimate. Occupation-level samples also show that within five years after a technology shock, each additional 1 percentage point rise in the overall unemployment rate is associated with an additional decline of about 4% in employment in the affected occupations.

Goldman Sachs' conclusion is that if economic activity remains healthy, the AI labor transition is likely manageable, with effects emerging gradually; if the economy enters a cyclical slowdown, AI-related automation and job losses may accelerate and amplify business cycle fluctuations. For investors, the future requires tracking AI capital expenditure, corporate labor cost savings, AI-related layoff announcements and the macroeconomic employment cycle simultaneously.

~~~~~~~~~~~~~~~~~~~~~~~~

The above highlights come from the Wind Trading Desk.

For more detailed analysis, including real-time commentary and frontline research, please join the [Wind Trading Desk ▪ Annual Membership] Risk Disclosure and Disclaimer

Markets carry risk; invest with caution. This article does not constitute personal investment advice and does not take into account the specific investment objectives, financial situation or needs of individual users. Users should consider whether any opinions, views or conclusions in this article are suitable for their particular circumstances. Invest at your own risk.