The benefits of artificial intelligence are relatively easy to measure and can quickly be seen in reduced headcount and increased production. The costs are hard to see and can take years to become visible.
When routine work is automated, the first roles to be filled are often the most junior. While it makes sense on paper, it cuts off the supply of future managers and creates an accumulation of expensive experience and poor judgment.
This year, a new wave of AI tools from companies like Anthropic have begun taking on tasks in a variety of professions, from law to banking to accounting, raising questions about how much of this work still has to be done by entry-level staff.
In areas such as marketing, communications and customer service, executives say AI can now absorb much of the routine work that was once assigned to entry-level employees. Professional services firm PwC saw a 35% increase in applications for entry-level roles last year, even as AI took over lower-value tasks. But importantly, the company is refraining from automating some junior tasks to preserve training and judgment.
This highlights how simple the economic logic of culling young jobs may seem, and how the trade-offs are anything but.
It’s true that hiring fewer junior employees reduces salaries, reduces training and onboarding costs, and reduces supervisory time from senior staff. New employees in white-collar knowledge work are often net costs rather than net contributors because they take some time to become fully productive during the first few months.
When you reduce ranks using AI, you get parts of the three cost tiers at once. However, please be careful. This is short-sighted and overlooks how junior roles develop future talent.
Entry-level staff aren’t just there to create presentations and aggregate data. They are there to learn how higher level work is done. That’s how judgment is built, and it’s built gradually over time, with guidance from experienced staff.
Some of this can be replaced with formal training or shadowing. However, these are imperfect substitutes for learning by doing and cannot simply be automated away.
Designing junior roles around AI
Much now depends on how companies design junior AI roles. Some simply hand over tools and let the AI absorb tasks that once served as training. It replaces the learning those roles were supposed to provide.
Some staff members are more cautious and let the AI handle routine work, while junior staff members focus on judging, questioning, and refining the output. This distinction is important. One is to hollow out the pipeline. The other can accelerate the building of judgment.
The risk is that most companies don’t make this choice intentionally. As economic conditions tighten and companies are forced to cut costs, it’s easier to cut headcount than redesign roles. Losses only become visible later when repairs become more difficult and expensive.
It makes sense to use it to reduce costs, and companies have an obligation to their shareholders to increase efficiency and create long-term value. The question is how the culling of the young workforce will affect it over time. The answer lies in weak management pipelines, bottlenecks in the middle of the organization, and a gradual erosion of competitiveness.
Companies that treat junior hires as a cost to reduce rather than build capacity are likely to pay the price later in the form of higher wages or slower implementation for experienced talent. This happens because more AI-generated work needs to be checked, but fewer senior staff are responsible for checking it.
Not all AI output requires close scrutiny. But in professional work, such as legal advice, financial reporting, and auditing, where errors can have serious consequences, the limitations are significant.
And this becomes even more important as economic growth comes under pressure from rising energy costs and persistent inflation. Performance depends on good decision making, which makes capability gaps much harder to hide.
This also emphasizes judgment. AI is changing where economic value comes from in white-collar jobs. Producing output is becoming easier and cheaper. Marginal value thus shifts to identifying, interpreting, and deciding what to trust and what to question.
Gap company risks
This is a gap, a risk that arises when companies hollow out their young employees. And that shift is now changing what it means to prepare to enter the workforce.
The truth is that no graduate has ever entered the job market as prepared. Differences in degrees, educational institutions, and personal experience will always determine how well-prepared they are for the serious work. But AI has introduced a new layer to that distinction.
The question is not whether graduates will use AI, but how. Some people think of AI as a replacement, leaving the job to them. Others use it more expansively to explore complex problems, test ideas, and accelerate learning.
For employers, this means they need to change the way they evaluate job applicants. Some places are already doing this. Consulting firm McKinsey is now asking graduates to use its AI assistants in job interviews, testing not just what the AI assistants produce, but how candidates can encourage, challenge, and refine the AI assistant’s output.
And the benefits from this can be huge. When used successfully, AI can enable small teams of graduates to tackle problems that once required much more senior staff. With well-designed AI workflows, junior analysts can explore scenarios and test assumptions and surface patterns at a speed and scale that was impossible just a few years ago. However, human constraints still remain. Someone needs to ask the right questions, recognize when the output makes sense, and know when to challenge it.
Without it, companies risk cutting off the very pipeline of training future executives and paying for the savings now later.
José Parra Moyano is Professor of Digital Strategy at IMD.
