Receive complimentary notifications to stay abreast
Just register for the Artificial intelligence myFT Digest, which will be sent straight to your email.
The author serves as director of Stanford University’s Digital Economy Lab and co-founder of Workhelix
Economists have struggled for more than ten years with a contemporary version of the Solow Paradox: artificial intelligence has been observed everywhere except within efficiency figures. Doubters contend this is because recent advancements in machine learning systems and, more recently, generative AI appear dim when compared to the significant discoveries of former times. Nevertheless, the most recent benchmark adjustments from the Bureau of Labor Statistics indicate that the statistical ambiguity may finally be dissipating.
Information published this week provides a notable counterpoint to the assertion that AI has yet to influence the US economy overall. Although preliminary reports indicated a year of consistent workforce growth in the US, the fresh data show that overall employment rise was adjusted lower by around 403,000 jobs. Significantly, this adjustment lower happened while real GDP remained strong, encompassing a 3.7 percent growth rate in the fourth quarter. This separation—sustaining high output with substantially reduced workforce contribution—is the defining characteristic of efficiency improvement.
My personal, revised assessment indicates a US efficiency enhancement of approximately 2.7 percent for 2025. This represents an almost doubling from the slow 1.4 percent yearly mean that typified the previous ten years.
This change corresponds with the efficiency “J-curve” that my colleagues and I examined in prior studies. Broad-application technologies, such as the steam engine and computers, do not provide instant benefits. Rather, they demand a period of substantial, frequently unquantified investment in non-physical assets—restructuring business operations, upskilling employees, and creating fresh business paradigms. During this stage, assessed efficiency is restrained as funds are redirected towards investments. The revised 2025 US data indicates we are now moving out of this investment stage into a harvest period where initial efforts start to appear as quantifiable results.
Detailed evidence further bolsters this fundamental change. In our research on the impact on jobs from AI last year, Bharat Chandar, Ruyu Chen, and I noted a slowdown in initial-tier recruitment within AI-exposed sectors; staffing for entry-level positions fell by approximately 16 percent, while those who used AI to augment skills experienced increased job opportunities. This implies firms are starting to use AI for some specified, foundational duties.
Although the patterns are indicative, a measure of prudence is advisable. Efficiency measurements are notoriously unstable, and numerous additional periods of continuous expansion will be required to establish an enduring long-term direction. Moreover, potent macroeconomic obstacles, spanning from geopolitical trade wars to financial or currency misgovernance, could offset these improvements in efficiency.
However, there is reason for greater hope when we differentiate between prospective and achieved benefits. Numerous enterprises are employing generative AI for only a modest portion of assignments. Certain entities simply utilize AI for interpretation or condensation—what could be termed ‘advanced lexicon’ use.
In contrast, my firm discovered a select group of advanced users utilizing engaging dialogues with intelligent agents to streamline entire workflows, such as producing comprehensive marketing plans, thereby condensing weeks of effort into mere hours. The difficulty for enterprises is not merely obtaining the technology, but rather elevating the capability of the typical worker through its use. This will enhance not only their individual earnings but also drive efficiency advancements throughout the entire economic system.
We are moving from a period of AI exploration to one of fundamental usefulness. We now need to concentrate on comprehending its exact workings. The efficiency resurgence is not merely a sign of AI’s potency; it serves as an urgent summons to pay attention to the impending economic metamorphosis.