How many jobs does AI really replace? Goldman Sachs has three takeaways looking under the hood and to ease the hype

In the second quarter of 2025, Hatzius’ team discovered “notable advances” in AI adoption, with 9.2% of US companies using AI to produce goods or services compared to 7.4% in the first quarter. The report also provides subtle photos and finds that while generative AI and related technologies are rapidly reshaping corporate investment and productivity, the impact on employment is evolving at a slower, more subtle pace.

Below are three takeaways from Goldman AI Adoption Tracker:

1. Limited labor market disruption (so far)

Despite the surge in AI adoption across US companies, the research notes found that overall labor market outcomes have been largely unaffected so far. Simply put, “The impact of AI on the labor market remains limited and there are no signs that will have a major impact on most labor market outcomes,” in contrast to that sign. The high-tech sector is cutting down jobs exposed to AIand when some well-known CEOs warn AI, AI can drive away more than 50% of the white-collar workforce.

Specifically, Goldman states that key indicators such as employment growth, wage growth, unemployment and layoff rates in AI-exposed industries show little statistically significant deviations from the less exposed sector. AI-related job postings currently account for 24% of all IT openings, but still account for 1.5% of total job postings. This indicates that the broader workforce changes are slower while the role of technology is adapted.

In particular, unemployment rates for jobs exposed to AI are now reconciling with the wider economy, rebutting the early fears of mass displacement. There have been no recent layoff announcements explicitly cited as the cause of AI, further highlighting the current containment of disruptions in certain functions rather than in the industry as a whole.

Meanwhile, analysts said payroll growth continues to be insufficient in occupations where AI is anecdotal impact, similar to the prominent examples of telephone call centres. This suggests that something is happening just whispering. Still, it’s early on.

2. Productivity is concentrated, but it is important

Goldman says the impact of AI on already deployed productivity is significant. Hatzius’ team cited academic research and company anecdotes showing that generative AI adoption increases labor productivity by an average of 23% to 29%. Estimates vary, with academic studies generating a median of 16% and an average of 23%, while company anecdotes generating a median of 30% and an average of 29%. Nevertheless, this suggests a concrete improvement in efficiency for early adopters.

Sectors that utilize the most active generation of AI – information, finance and professional services will see the biggest increase in productivity as companies move from experiments to integrating into AI core workflows.

Business leaders and economists hope that as adoption deepens and more organizations build AI on their infrastructure, the impact of gross productivity will become more prominent in macroeconomic data.

3. AI Employment Story: Still in the early chapters

The recurring theme of the Goldman Sachs analysis is that the full employment effectiveness of AI is still developing. AI-related openings are growing, but in IT, there is a growing demand for roles such as machine learning engineers and AI researchers. The survey reflects that a significant share of companies is planning employment for these skill sets.

Improved productivity could ultimately expand to more industries, with “AI strength” (which plays a large role using AI) being the highest in the information technology and professional services sector, indicating that future employment changes could be first realized.

The report said that while the current impact of AI on the labor market is limited, the species of transformation is owned. Increased corporate AI adoption, particularly among large and medium-sized companies, points to changes in future productivity and roles. But for now, the fear of widespread job losses from AI appears to be exaggerated, at least until deeper integration with technology and business processes occurs.

As businesses continue to expand AI and mature support infrastructure, both opportunities and challenges will be amplified, ensuring close observation by policymakers, business leaders and workers.

Goldman Sachs declined to comment further.

For this story, luck Generated AI was used to assist with initial drafts. The editors checked the accuracy of the information prior to publication.