Everyone keeps asking when AI will start taking people’s jobs.
That question assumes the disruption will arrive the way layoffs usually do. Pink slips. Headlines. CEOs talking about efficiency during earnings calls.
But the early signal may look different. AI isn’t primarily replacing workers yet. It’s replacing hiring. The first victims of AI are graduates.
Every year, millions of young professionals enter the workforce expecting to start at the bottom rung of the ladder. Junior analyst. Associate developer. Research assistant. Entry-level marketer. These roles were never glamorous, but they served a purpose. They were how people learned the trade.
Now, many of those tasks are exactly what generative AI systems do best.
Recent research from Anthropic looked at how large language models interact with real workplace tasks. The findings are striking. In several white-collar occupations, AI systems could theoretically perform a large share of the work.
Computer and math roles show roughly 96% potential task coverage. Business and finance jobs approach 94%. Office and administrative work sits in the same range. Legal and media work are not far behind.
Actual usage today is still much lower, but the capability already exists. Once technology reaches that point, adoption usually follows.
For decades, companies hired junior workers to handle the routine work that kept offices running. Compiling research. Summarizing documents. Preparing presentations. Drafting early versions of reports. Writing basic code. Analyzing spreadsheets.
These jobs were the apprenticeship system of the knowledge economy.
They also happen to be exactly the kind of work large language models now handle remarkably well.
Give one capable employee access to strong AI tools and the productivity math changes quickly. Instead of hiring five junior analysts, a company may hire one and let software handle much of the routine work.
That decision does not produce layoffs. It simply means the other four jobs never appear.
The labor market may already be showing early signs of this shift. Entry-level job postings have declined across several sectors over the past two years. Hiring for junior roles at major technology companies has fallen compared with the early 2020s.
Economists studying AI-exposed occupations have found another interesting signal. Employment among workers in their early twenties has declined in several AI-exposed fields while older workers remain relatively stable. Companies appear to be keeping experienced employees while slowing the intake of new ones.
Recent graduates are already feeling the impact. Unemployment among new college graduates has risen above the overall labor market in some periods, and a growing share of graduates are working outside the fields they studied.
This pattern should not be surprising.
Research from the Brookings Institution estimates that roughly 30% of workers could see at least half of their job tasks affected by generative AI. Analysis from Goldman Sachs suggests that about two-thirds of jobs are exposed to some degree of AI automation.
The estimates vary, but they all point in the same direction.
Large language models automate work that is text-based, analytical, repetitive and digital. That describes a large portion of modern office work.
Technology has been reshaping jobs for centuries. Industrial machinery transformed manufacturing. Spreadsheets changed finance and accounting. ATMs altered how banks staffed branches.
Most of those shifts targeted physical or routine work. AI is different because it targets routine cognitive work.
Writing, research, coding assistance, documentation and analysis are increasingly within reach of these systems. That places the modern office economy directly in the path of automation.
The real risk is not simply job loss. It is the disappearance of the career ladder.
For decades, companies hired large cohorts of junior employees who gradually learned the profession and moved up. If AI reduces the need for entry-level workers, that pipeline begins to shrink. Fewer people enter the field each year, which eventually means fewer people gain the experience needed to become senior professionals later.
Labor markets depend on those pipelines. Remove the bottom rung, and the entire structure starts to wobble. For decades, people assumed automation would hit factory workers first and office workers later.
In many ways, that timeline has reversed. Manufacturing experienced automation decades ago.
Now the office is beginning to feel it. Not through waves of layoffs. Through something quieter.
The job a new graduate is expected to start with may simply never be posted.
