Goldman Sachs: If AI Delivers 15% Productivity Gains, Up to 9% of Workers Could Be Displaced During the Transition
nashnova research
Goldman Sachs estimated on October 22 that if AI ultimately delivers a 15% productivity boost, up to 9% of workers could be displaced during the transition — and an economic slowdown would amplify the hit.
What does history say about tech-driven job displacement?
Goldman used three empirical models and found: for every 1% rise in tech-driven labor productivity, peak unemployment temporarily climbs 0.3 to 0.4 percentage points, then fades after roughly two years.
Breaking it down by job flows, each 1% productivity gain produces a 0.5–0.6 pp cumulative rise in the separation rate and drags the hiring rate down by about 0.2 pp.
This means → scaling that historical ratio to AI's projected 15% productivity uplift, up to 9% of workers could be displaced from their current roles, with hiring slowdowns creating an additional 3% drag on total employment.
Which jobs have already been "killed" by technology?
Goldman sampled 25 occupations most vulnerable to automation — photo-lab technicians, proofreaders, file clerks, meter readers, travel agents, bank tellers, among others.
Over the five years following each tech shock, employment in these 25 roles fell by an average of about 25%. Photo-lab technicians suffered the steepest drop, losing nearly 60% of jobs.
In plain terms = digital cameras wiped out film processing, word processors wiped out proofreading, electronic filing wiped out clerks, online booking wiped out travel agents — each wave of technology surgically compressed a specific manual workflow.
How do jobs disappear — layoffs or hiring freezes?
One of Goldman's most important findings: job losses from tech shocks come more from separations (layoffs and departures) than from reduced hiring.
Of the 25 occupations, 19 saw their annual separation rate rise, by an average of 3.8 pp. Only 13 saw hiring rates fall, with an average decline of just 0.7 pp.
This means → watching hiring plans alone to gauge AI's labor impact will lag badly. Goldman recommends tracking AI-related layoff announcements as a high-frequency leading indicator — layoffs arrive earlier and hit harder than hiring freezes.
Where does AI's employment impact stand right now?
Only 23% of U.S. organizations have integrated AI into part of their regular workflow. Most AI-driven productivity gains are still ahead.
Early signals are visible: the most AI-exposed industries added roughly 10,000 fewer jobs over the past three months versus trend, though this gap was largely offset by hiring in AI-related construction roles.
This reflects a phase Goldman calls "calm but not safe" — spring panic over AI replacing jobs has cooled, yet cross-sectional service-sector data show the correlation between AI adoption and unemployment has turned positive and is statistically significant.
How would an economic slowdown change the picture?
Historical data show automation shocks are procyclical: when overall unemployment rises, each 1% of tech-driven productivity growth lifts unemployment by an extra 0.1–0.2 pp beyond the baseline estimate.
At the occupation level, for every additional 1 pp rise in aggregate unemployment during the five years after a tech shock, employment in affected roles drops an extra ~4%.
In plain terms = in a strong economy, displaced workers can find new jobs; once the economy weakens, tech-displaced and cyclically unemployed workers stack on top of each other, amplifying the shock.
What should investors watch?
Goldman's bottom line: if the economy stays healthy, AI labor-force transition is most likely manageable. If the economy enters a cyclical downturn, AI-related automation and job losses could accelerate and amplify business-cycle swings.
Four threads to track simultaneously: AI capex, corporate labor-cost savings, AI-related layoff announcements, and the macro employment cycle.
This means → of those four, layoff announcements are the most real-time leading signal — they reflect AI's displacement effect earlier than any shift in hiring plans.
市场有风险，内容仅供研究参考，不构成投资建议。