In the current economic climate, a sense of mounting anxiety has taken hold of the American workforce, particularly among those who have spent years and significant financial resources securing college degrees. The prevailing narrative is one of disruption and obsolescence, driven by the rapid ascent of generative artificial intelligence. There is a widespread belief, bordering on panic, that AI is systematically destroying entry-level white-collar work, rendering the traditional career ladder broken and the value of higher education questionable. However, a closer examination of the data, the timing of market shifts, and the fundamental nature of professional development suggests that the reality is far more complex than a simple story of machines replacing humans.
To understand the current state of the labor market, one must look back to a pivotal moment that predates the public release of ChatGPT. Having spent a decade serving at the Federal Reserve, I have witnessed firsthand how the labor market responds when monetary policy begins to tighten. In March 2022, the Federal Open Market Committee (FOMC) began a series of aggressive interest rate hikes to combat inflation. It is a matter of record that job postings for the occupations most exposed to AI—those involving significant data processing, writing, and analytical tasks—peaked at exactly that moment and began a steady decline. Crucially, this downturn started eight months before ChatGPT became a household name. This timing suggests that the initial cooling of the high-end labor market was a reaction to the cost of capital and shifting corporate priorities rather than a direct result of automation.
The narrative of AI-driven displacement often draws support from high-profile academic research. For instance, the Stanford Digital Economy Lab has released findings indicating that workers between the ages of 22 and 25 in the most AI-exposed occupations are currently trailing their peers in less-exposed fields by approximately 19 percent in terms of employment metrics. This gap has notably widened over the past year, providing a statistical anchor for those who fear that the "canaries in the coal mine" are already falling. If young graduates in tech, finance, and professional services are struggling while those in manual or service-heavy roles are not, it seems logical to blame the technology that mimics cognitive labor.
However, a deeper dive into the evidence reveals a more nuanced picture. Economists Zanna Iscenko and Fabien Curto Millet recently conducted an extensive analysis of 238 million job postings, attempting to disentangle the effects of technology from those of broader economic policy. Their research tied the decline in job postings directly to the Federal Reserve’s tightening cycle. They pointed out that AI-exposed occupations tend to cluster in specific sectors—information technology, finance, and professional services—which are notoriously sensitive to interest rate fluctuations. When the cost of borrowing rises, these capital-intensive and growth-oriented sectors are often the first to scale back on expansion and hiring.
Further supporting this view is data from the Economic Policy Institute, which highlights a crucial inconsistency in the "AI-as-culprit" theory. Young workers without college degrees, whose occupations typically score negative on AI exposure because they involve physical or interpersonal tasks that AI cannot yet perform, have also seen their unemployment rates rise at a similar pace during this period. If AI were the primary driver of the employment decline, one would expect to see a sharp divergence between the "exposed" white-collar class and the "unexposed" blue-collar or service class. Instead, the pain appears to be distributed across the junior labor market, pointing toward a macroeconomic slowdown rather than a technological purge.
Even the researchers at Stanford, whose work is often cited as proof of AI’s destructive power, offer a necessary caveat that many media outlets overlook. Their own papers clarify that their findings represent "descriptive patterns, not causal estimates." They explicitly state that they do not yet see widespread, economy-wide job displacement directly attributable to AI. Furthermore, they have engaged with the monetary policy argument, acknowledging that while the most exposed jobs are not always the most rate-sensitive in a historical sense, the widening employment gap for young workers remains a phenomenon that persists even as interest rates have begun to stabilize or decline.
This leads us to a more profound question: If firms aren’t necessarily firing their junior employees because of AI, why is the entry-level market so stagnant? The truth is that we are living through a period of overlapping shocks, and it is nearly impossible to isolate the impact of AI from the impact of interest rates in real time. Nevertheless, businesses, educational institutions, and regulators must make high-stakes decisions today regarding hiring and training. In doing so, they are uncovering a shift in corporate behavior that may be more significant than the technology itself.
The data reveals a clear trend: firms are not engaged in mass layoffs of their junior staff; rather, they are simply hiring fewer of them. This decline in hiring is most concentrated in areas where AI is seen as a tool for automating work rather than merely complementing it. To understand why this is happening, we must reconsider the traditional role of the junior employee. For decades, organizations hired junior staff not for their immediate productivity, but as part of a long-term "formation" process.
Consider the first-year associate at a law firm or the resident doctor in a hospital. By any objective measure of short-term efficiency, these roles are often "unproductive." A junior associate’s legal brief almost always requires extensive revision by a senior partner, and a resident’s initial patient assessment is frequently repeated by an attending physician. From a pure accounting perspective, these tasks are redundant. Yet, firms have historically funded these positions because that is the only way to transform a novice into a senior professional with the judgment and expertise required to lead. The work they produced was merely a byproduct; the true goal was the development of the employee.
The introduction of AI has disrupted this delicate bargain. AI can now perform many of the "production" tasks that were previously assigned to junior analysts, such as summarizing documents, drafting basic code, or conducting initial data screens. As a result, firms see an opportunity to cut costs by investing less in these entry-level hires. However, in doing so, they are thinning the pipeline that produces the next generation of senior leaders.
This phenomenon was observed by researcher Matt Beane in operating rooms long before the current AI boom. As surgical robots were introduced, they often took over the "reps" that residents needed to build their skills. While the robots made the surgeries more precise, they quietly robbed the junior doctors of the hands-on experience necessary to become master surgeons. Today, we see a digital version of this across the white-collar world. Employers still demand experienced, high-level professionals, but many have stopped funding the very process that creates that experience.
This shift represents a fundamental misunderstanding of human capital. Employers must stop treating junior hiring as a simple operating expense that can be erased by automation. In reality, junior hiring has always been a capital investment mislabeled on the balance sheet. It is the mechanism by which a firm manufactures its own future. While AI can certainly take over the production value of a junior employee, it cannot replace the value of the training. To survive this transition, organizations must become more deliberate about how they fund and structure professional development. This means rethinking training modules, creating intentional job rotations, and prioritizing mentoring and coaching—all with the specific goal of developing the "judgment" that AI cannot replicate.
Universities face a mirror image of this problem. As a former dean and university president, I have seen how the academic apprenticeship is under threat. The Ph.D. process, for example, is traditionally funded by the productive work of researchers in training. However, as the editors of the journal Nature warned recently, early-career researchers are now in danger of having the "crucial training tasks" of science performed by machines.
The evidence from the classroom is equally stark: students learn when they encounter "friction." Learning requires effort, the possibility of failure, and the necessity of working through difficult problems. When AI provides the answers instantly, that friction disappears, and with it, the development of deep understanding. There is a saying from my former boss in Texas: "No friction, no traction." If we allow AI to remove all the struggle from the educational and professional "formation" process, we will produce a generation that knows how to get the right answer but does not understand the work behind it. We must return to the old mantra of the mathematics teacher: "Show your work."
The current trajectory suggests a future where AI handles the routine work, and an entire generation is denied the experiences that turn raw talent into seasoned judgment. This outcome is not an inevitable result of the technology itself; it is a choice made by institutions. It happens if employers view formation as a cost to be cut and if universities continue to certify work that was actually performed by a machine.
The job posting data will eventually recover as the economic cycle turns; it always does. What will not recover on its own is the "old bargain" where the production of work quietly paid for the formation of the professional. Rebuilding that bargain—deliberately, strategically, and with a clear budget—is the most pressing challenge posed by the rise of AI. The question is not whether the machine can do the work, but whether we are willing to continue investing in the humans who must eventually oversee the machine.
