AI could mark the end of young people learning on the job – with terrible results

For a long time, the deal for a wide range of careers has been simple enough. Entry-level workers carried out routine tasks in return for mentorship, skill development and a clear path towards expertise.

The arrangement meant that employers had affordable labour, while employees received training and a clear career path. Both sides benefited.

But now that bargain is breaking down^[1]. AI is automating the grunt work – the repetitive, boring but essential tasks that juniors used to do and learn from.

And the consequences are hitting both ends of the workforce. Young workers cannot get a foothold. Older workers are watching the talent pipeline run dry.

For example, one study suggests^[2] that between late 2022 and July 2025, entry-level employment in the US in AI-exposed fields like software development and customer service declined by roughly 20%. Employment for older workers in the same sectors grew.

And that pattern makes sense. AI currently excels at administrative tasks^[3] – things like data entry or filing. But it struggles with nuance, judgment and plenty of other skills which are hard to codify.

So experience and the accumulation of those skills become a buffer against AI displacement. Yet if entry-level workers never get the chance to build that experience, the buffer never forms.

This matters for organisations too. Researchers using a huge amount of data about work in the US described the way that professional skills develop^[4] over time, by likening career paths to the structure of a tree.

General skills (communication, critical thinking, problem solving) form the trunk, and then specialised skills branch out from there.

Their key finding was that wage premiums for specialised skills depend almost entirely on having those strong general foundational skills underneath. Communication and critical thinking capabilities are not optional extras – they are what make advanced skills valuable.

The researchers also found that workers who lack access to foundational skills can become trapped in career paths with limited upward mobility: what they call “skill entrapment”. This structure has become more pronounced over the past two decades, creating what the researchers described as “barriers to upward job mobility”.

But if AI is eliminating the entry-level positions where those foundations were built, who develops the next generation of experts? If AI can do the junior work better than the actual juniors, senior workers may stop delegating altogether.

Researchers call this a “training deficit”^[5]. The junior never learns, and the pipeline breaks down.

But the disruption will not hit everyone equally. It has been claimed^[6], for example, that women face nearly three times the risk of their jobs being replaced with AI compared to men.

This is because women are generally more likely^[7] to be in clerical and administrative roles, which are among the most exposed to AI-driven transformation. And if AI closes off traditional routes into skilled work, the effects are unlikely to be evenly distributed.

So what can be done? Well, just because the old pathway deal between junior and senior human workers is broken, does not mean that a new one cannot be built.

Young workers now need to learn what AI cannot replace in terms of knowledge, judgment and relationships. They need to seek (and be provided with) roles which involve human interaction, rather than just screen-based tasks. And if traditional entry-level jobs are disappearing, they need to look for structured programmes that still offer genuine skill development.

Older workers meanwhile, can learn a lot from younger workers about AI and technology. The idea of mentorship can be flipped, with juniors teaching about new tools, while seniors provide guidance and teaching on nuance and judgment.

And employers need to resist the urge to cut out junior staff. They should keep delegating to those staff – even when AI can do the job more quickly. Entry level roles can be redesigned rather than eliminated. For ultimately, if juniors are not getting trained, there will be no one to hand over to.

Protecting the pipeline of skilled and valuable employees is in everyone’s interest. Yes, some forms of expertise will matter less in the age of AI, which is disorienting for people who may have invested years in developing them.

But expertise is not necessarily about storing information. It is also about refined judgment being applied to complex situations. And that remains valuable.

References

1. ^{^} is breaking down (theconversation.com)
2. ^{^} one study suggests (digitaleconomy.stanford.edu)
3. ^{^} at administrative tasks (digitaleconomy.stanford.edu)
4. ^{^} professional skills develop (www.nature.com)
5. ^{^} “training deficit” (www.hbs.edu)
6. ^{^} It has been claimed (news.un.org)
7. ^{^} generally more likely (news.un.org)
8. ^{^} Jelena Zelen/Shutterstock (www.shutterstock.com)