Six Questions for an Economist on Employment and Generative AI

What effect might generative AI, or artificial intelligence, have on workers’ employment levels? How about on their productivity? Those are among the questions arising as the older members of Gen Z, the generation born from about 1997 to 2012, are attending college, entering the workforce or navigating the early part of their careers.

Serdar Ozkan, a senior economic policy advisor and economist in the St. Louis Fed’s Research division, answered six questions on employment and generative AI.

The unemployment rate has gradually increased for a couple years. Do you expect this trend to continue, or are there indicators that it might start declining?

While I cannot definitively predict future trends, several structural factors point toward continued pressure on the labor market. The concentration of unemployment increases among recent college graduates and workers in generative AI-exposed occupations suggests we’re experiencing fundamental labor market shifts, not merely cyclical fluctuations. Though the current unemployment rate remains historically moderate, and monetary policy easing could provide some relief, the underlying technological disruption appears likely to persist for some time—particularly for younger workers entering the labor market.

How has the COVID-19 pandemic affected unemployment trends?

The pandemic’s role has been complex. Initially, it created extraordinary demand for digital services, leading to exceptionally strong employment for computer and tech workers. What we’re seeing now may partly reflect a return to something closer to historical averages.

The older members of Gen Z are old enough to have attended or graduated from college. From your perspective, is a college degree becoming less valuable in some fields when it comes to finding or securing a job?

Historically, the unemployment gap between college graduates and noncollege workers has been cyclical—widening during economic downturns and narrowing during expansions. This reflects the greater vulnerability of less-educated workers to economic shocks. For example, the unemployment gap between young men with and without college degrees reached its peak following the Great Recession in the late 2000s but has since narrowed to historically low levels. (See the FRED chart below.) However, something different is happening now. Despite the overall unemployment rate rising since mid-2023, the unemployment gap has remained closer to zero rather than widening as we would expect. This break from historical patterns suggests the college premium for immediate employment security is indeed eroding.

What factors are driving this decline in the college premium?

Two primary forces appear to be driving this change.

First, as I said earlier, the strong post-pandemic recovery helped closing the unemployment gap between college and noncollege workers.

Second, AI and large language models are democratizing capabilities once exclusive to college graduates. These tools can now perform research, analysis and even creative tasks that traditionally justified the college premium, reducing the immediate employment advantage of a degree.

However, context matters. While the employment advantage has narrowed, college graduates still maintain lower absolute unemployment rates than non-degree holders. The degree continues to provide value, but that value increasingly depends on the specific field and how susceptible it is to technological disruption. Fields requiring interpersonal skills, complex judgment or specialized technical knowledge maintain stronger returns to education than those involving routine cognitive tasks.

How does generative AI compare with earlier booms of disruptive technologies, particularly in terms of potential productivity gains?

Gen AI is emerging to be another general-purpose technology, comparable to the internet or computers in its potential to transform virtually every sector of the economy. This parallel with previous technological developments offers both insights and cautions about what might lie ahead.

The comparison to computers is particularly instructive. In 1987, Nobel laureate economist Robert Solow famously observed, “You can see the computer age everywhere but in the productivity statistics.” This productivity paradox persisted for nearly two decades; despite massive investments in computing technology from the mid-1970s through the 1980s, aggregate productivity growth remained sluggish until the mid-1990s. Only after businesses fundamentally reorganized their processes and workers developed complementary skills did the productivity benefits finally materialize.

We may be witnessing a similar pattern with gen AI. While its capabilities are already visible everywhere—from coding assistance to content generation—productivity statistics have yet to reflect a fundamental transformation. History suggests this lag is normal: General-purpose technologies require extensive time for experimentation, process redesign and, crucially, workforce adaptation.

Does generative AI remind you of another technological development that could help indicate the future of the labor market and how things might turn out for workers?

The adoption of transformative technologies consistently involves significant learning costs, and skilled workers typically hold an advantage in this process. During the computer revolution, college-educated workers were better positioned to adapt, leveraging their stronger foundational skills in abstract reasoning and information processing.

However, gen AI presents a unique twist. While educated workers may still have advantages in learning to integrate AI systems into their work, the technology itself can perform many of the tasks that previously justified their skill premium. This creates a tension: The workers best equipped to adapt may simultaneously face the most direct competition from the technology itself.

Historical general-purpose technology adoptions reveal that while aggregate employment typically remained stable or even grew over the long run, transition periods created significant distributional challenges. The computer age ultimately created entirely new job categories—web developers, data scientists, social media managers—that were unimaginable in the 1980s. Similarly, gen AI will likely generate roles we cannot yet envision. However, the interim period may prove particularly challenging for workers caught in this transition, especially recent graduates who invested heavily in skills that AI can now replicate.

The key lesson from history is that these transitions, while ultimately beneficial for productivity and living standards, are not automatically smooth or painless. Having deliberate support systems could help workers successfully navigate the transformation.