The doomsayers promised a bloodbath. By now, according to the breathless forecasts of 2023 and 2024, artificial intelligence was supposed to have rendered vast swaths of the white-collar workforce obsolete. Accountants, paralegals, copywriters, junior analysts—all were meant to be clearing out their desks, replaced by tireless language models that never sleep and never ask for raises.
It hasn't happened. Unemployment in advanced economies remains stubbornly low by historical standards. The mass layoffs that were supposed to accompany the AI revolution have, so far, failed to materialize at anything like the predicted scale. And yet something is clearly wrong—something that the aggregate employment statistics are failing to capture.
The entry-level evaporation
The problem is not that AI is eliminating jobs wholesale. It is that AI is eliminating the first jobs—the junior positions, the trainee roles, the unglamorous grunt work that has traditionally served as the on-ramp to professional careers. When a law firm can use AI to handle document review that once occupied a dozen first-year associates, it doesn't fire its partners. It simply stops hiring as many first-years. When a consulting firm can generate initial research reports with a few prompts, it doesn't lay off its senior consultants. It just needs fewer analysts fresh out of university.
The result is a labor market that looks healthy in the aggregate but is quietly closing its doors to the young. Graduate employment rates are softening across multiple industries. The time between degree completion and first professional role is stretching. And the traditional pathway—start at the bottom, learn the craft, work your way up—is being hollowed out from below.
The training paradox
There is a bitter irony here. The very tasks that AI handles most competently—routine research, basic drafting, data compilation, preliminary analysis—are precisely the tasks that once taught junior workers how their industries actually function. A young lawyer learned to think like a lawyer by spending years in the weeds of discovery. A junior banker learned to build models by building hundreds of them, badly at first, then better. The grunt work was never just labor extraction. It was education.
Now firms face a peculiar dilemma. They can use AI to handle the work that juniors once did, saving money in the short term. But in doing so, they may be failing to develop the next generation of senior talent. The pipeline that produces experienced professionals is being crimped at its source.
Our take
The optimists will argue that new roles will emerge, that young workers will find different paths into the economy, that this is simply another chapter in the long history of technological displacement and adaptation. They may be right, eventually. But "eventually" is cold comfort to the twenty-three-year-old with a degree and a resume that keeps disappearing into automated screening systems. The AI employment crisis we feared—sudden, dramatic, visible—would at least have been impossible to ignore. The one we got is slower, subtler, and far easier for comfortable incumbents to overlook. That makes it more dangerous, not less.
