Despite its rapid proliferation over the last couple years, AI has had no discernible effects on major labor market metrics, according to Goldman Sachs. 

“Aggregate labor market impacts are still negligible,” wrote Goldman Sachs economists Joseph Briggs and Sarah Dong. “Although AI exposure is highly correlated with adoption, there is no economically or statistically significant correlation between AI exposure and job growth, unemployment, job finding rates, layoff rates, weekly hours, or average hourly earnings.”

At least not yet.

Companies across the U.S. have yet to fully incorporate AI into their operations, Briggs told Fortune. 

“I expect that will change over the next several years,” he said. “We’re just still a little bit in the early days of the AI transition.”

Most firms took a gradual approach to integrating AI because the technology was so new and expensive that it required extensive overhauls to existing processes. AI also posed entirely new concerns over data privacy and security that companies had to contend with before rolling it out across their organizations. Some companies may also be playing a waiting game to figure out which AI tools will end up being the best, according to Briggs. 

“The technology is a little bit early, and so you don’t want to be locked into what, in the long run, is an inferior platform,” he added. 

Goldman expects AI will eventually boost the economy overall starting in 2027 with increases to labor productivity and GDP. That uplift will continue through to the late-2030s, according to Briggs and Dong’s note.

Previous Goldman estimates forecasted the full adoption of AI could lead to a 15% increase in U.S. labor productivity and a 7% increase in global GDP. 

“We have a still fairly positive outlook and bullish outlook on the longer-run impacts of AI, and we expect that it will unlock a lot of economic value,” Briggs said. 

Any overarching changes to the economy stemming from AI would have first shown up in the labor market because employment data is regularly tracked and the technology’s ability to automate tasks would, in theory, boost productivity, Briggs and Dong wrote. But since there haven’t been any major changes to ongoing trends in the unemployment rate and productivity, Goldman reasons, AI’s effects haven’t yet kicked in. 

Most labor market indicators have remained somewhat steady over the last year or so. The unemployment has hovered around 4% since December 2023, never dipping below 3.8% or rising above 4.2% over that time. The latest labor productivity measure for the fourth quarter of 2024 was 2% growth, which was a slight slowdown from the 2.5% in the third quarter. 

That’s not to say there haven’t been some changes to the overall labor market. Certain professions, like computer programming, customer service and professional services, that are highly susceptible to generative AI have started seeing some minor changes to hiring practices. 

Specifically, those industries where AI is more heavily adopted have seen slight declines in labor demand, with fewer job postings, according to Goldman’s analysis of data from the jobs website Indeed.com. 

“Labor demand trends over this period show a common decline across nearly all service subsectors that is correlated with exposure to AI automation, suggesting that the onset of generative AI tools may have led some companies to reevaluate hiring plans,” Briggs and Dong wrote in their note. 

Briggs said the next metric his team at Goldman will be watching is “actual job growth” across the economy to see if hiring slowdowns in certain industries are offset by new jobs. 

Big changes to the labor market are expected in the coming years. Goldman’s own 2023 forecast that 300 million jobs in the U.S. and Europe were susceptible to some level of change due to AI captured the scale of the potential impact.

However, Brigg’s report said that it will be at least two years until those changes really manifest themselves. The shakeup to the labor market will ultimately come down to how quickly AI is adopted across broad swaths of the economy. That will only start to happen once costs come down, for both AI companies and their eventual customers, and developers continue building out more applications.  

The AI industry has already started to make strides toward cheaper solutions. The release of the Chinese company DeepSeek’s latest chatbot sent shockwaves throughout the industry when it was reported to have outperformed some of OpenAI’s best models for a fraction of the development costs. Major U.S. developers have also taken strides to find new, cheaper sources of energy in a bid to lower their own costs. 

There has been some progress in business-to-business applications for specific industries like financial services and design. However, the market for these tools has yet to translate to an overhaul of the traditional software market or, as Goldman points out, lead to a substantial, measurable increase in worker productivity outside of a few examples. Though that will change as companies hire more AI talent and cobble the money to invest in new tech; both of which many still lack. 

“There’s just a lot of companies that don’t have the expertise or the internal capital to figure out how to develop applications and restructure workflows so that they can incorporate AI and reap the benefits to productivity that it promises,” Briggs said.