Tech Frontiers™

The AI glass floor

Artificial intelligence is taking over the grunt work that once was a rite of passage for junior staff learning the ropes. That shift may reshape white-collar careers more than job losses ever could.

12 Jun 2025

A handshake between two individuals set against a blue background

Tech evangelists have long had a tendency to overstate the economic and societal impact of their inventions. Artificial intelligence (AI) may prove no exception. But concerns about its effects have spread beyond Silicon Valley, with many white-collar workers now echoing alarmist forecasts about being displaced by the brain-like tech. 

Some of that anxiety may be premature. AI is altering professional life—but not yet by outright displacing white-collar professionals. Rather, it is taking over the routine tasks—from drafting emails and scheduling meetings, to extracting data—that once offered junior employees informal training by exposing them to the job. If firms fail to rethink how early-career staff gain practical knowledge, AI may not eliminate jobs, but it could choke off entry into them.

Changing landscapes

In a recent interview, Jason Clinton, chief information officer of Anthropic, a leading AI model maker, underscored that the prospect of an AI “coworker” is closer than most imagine. A study by Microsoft—the software giant has invested heavily in workplace AI tools—tries to estimate how close: 82% of leading firms, it says, expect to deploy digital labour to scale their workforce within the next 12 to 18 months. The suggestion is that companies are quickly moving from using AI co-pilots to fully autonomous AI agents. 

These agents promise not just to support knowledge workers in everyday tasks, but to perform multi-step tasks independently, shifting AI from an assistant to a teammate. Some firms are already experimenting. UBS, a Swiss bank, has introduced AI clones of its analysts to respond to client queries without human oversight, freeing its overburdened human staff. IBM, which offers AI agents to clients, has used them internally to automate 94% of routine HR tasks, such as vacation requests and pay statements; in the software giant’s IT division, AI has helped reduce support tickets by 70%.

Despite the growing real-world capabilities of AI agents, there is scant evidence that this wave of automation is displacing white-collar workers in a significant way. A new report by PwC, which analysed nearly 1bn global job postings, finds that job numbers are rising across nearly every occupation the consultancy considers “AI-exposed”, such as customer support agents and data-entry clerks. Growth in these roles has been slower than in less-AI-exposed ones—38% over the past five years, compared with 65%—but the direction is clear: there is no outright displacement.

More telling is what has happened to pay. In 2025, the report finds, workers with AI skills command a 56% wage premium over those who lack them, up from 25% just a year earlier. Put simply, employers are paying more for workers who can delegate the routine to AI and concentrate on the judgement and problem-solving tasks of the job.

Entry, restricted

AI may not be eliminating white-collar jobs, but it is narrowing the path into them. The role of AI in this shift is difficult to measure, but signs are emerging. A report by Oxford Economics finds that in America, the unemployment rate for recent college graduates has climbed to 6% in the first part of 2025 —a historically high level, 1.6 percentage points above its 2019 average, and, unusually, higher than the national rate.

Some of this may be due to structural changes: colleges, some argue, offer less of a labour-market advantage than they once did, and the job prospects for young people never fully recovered after the covid-19 pandemic. But much of the pressure appears to be coming from a more recent shift. Since mid-2023, 85% of the increase in graduate unemployment has come from new entrants to the labour market—those who have not been exposed to any corporate job before. And the sharpest losses have come in business and professional services, where AI adoption has been relatively fast, with firms automating precisely the tasks that graduates used to perform.

These tasks, though routine, once served a purpose. They gave junior staff exposure to how firms operate, what good work looks like and how decisions are made. Now much of that foundational work is delegated to machines. Matt Beane, a researcher at the University of California, argues that this has quietly broken the traditional bond between expert and novice. In The Skill Code: How to Save Human Ability in an Age of Intelligent Machines, he describes a workplace in which senior staff work alongside AI agents, while juniors, increasingly, are left on the outside, looking for where to develop corporate skills. 

The path forward

Firms face a twofold challenge. First, to use increasingly capable AI in a way that enhances—rather than replaces—what humans do. Second, to reconsider corporate entry routes so that juniors can have a way to learn on the job. One promising practice, highlighted by the Harvard Business Review, is to move from static job descriptions to a dynamic classification of tasks and skills. This would allow firms to work out where machines outperform people and where a human remains indispensable. In turn, firms will be able to deploy AI where it adds greatest marginal value and to set priorities for the areas humans should emphasise. But companies also need to ensure that employees—experienced staff and juniors alike—are exposed to different functions within their organisations as AI takes over more routine work. “Micro-rotation” schemes could be one approach. They would allow newcomers to build skills through breadth rather than repetition. These programmes would also help longer-serving employees pivot diagonally, preserving career flexibility as AI reshapes parts of their job. Microsoft, for example, runs the Aspire programme, a two-year development track for graduates and other early-career hires that helps them build networks, understand the company’s operations and explore roles across different teams. For now, AI need not be the object of workplace envy. As firms adjust, a more deliberate approach to training—focused on higher-value tasks and interpersonal skills—can help junior employees adapt to working alongside AI. The route into white-collar work is changing, but with guidance from experienced managers, newcomers can still find their way.