AI and Entry Level Jobs: How Will the Next Generation of Experts Learn?

[Harry64]

In our thread on whether AI can beat a human expert at a real job, one reply raised a question that deserves its own discussion. If AI takes over the routine tasks that junior staff used to learn on, where will the next generation of experts come from? Every senior lawyer, accountant, engineer and journalist started out doing the basic work that taught them the trade. If that ladder loses its bottom rungs, how does anyone climb it? It is a serious question, but I think there are good reasons to be hopeful about the answers

There is some early evidence that the effect is real. A 2025 working paper from Stanford's Digital Economy Lab, titled Canaries in the Coal Mine, used payroll data to look at employment in jobs most exposed to AI. It found that employment for workers aged 22 to 25 in the most exposed occupations had fallen by about 13% relative to less exposed jobs since late 2022. Employment for more experienced workers in the same occupations held steady or grew. That pattern suggests AI is affecting entry level hiring first, rather than replacing experienced professionals

The same research found an important difference depending on how AI is used. Where AI mainly automates tasks, younger workers saw declines. Where AI mainly augments people, helping them do their work better rather than doing it for them, the effects were much smaller or even positive. That is a hopeful finding. It suggests the outcome is not fixed, and that how organisations choose to use AI makes a real difference to young workers' prospects

Other evidence is more mixed. The Harvard study on AI coding agents published this month found no significant effect on overall or engineering employment, even as the amount of code written rose sharply. Many companies are still hiring graduates, and some are hiring more people to review and manage AI generated work. Labour markets are complicated, and AI is only one of many forces at play. It is too early to say how large or lasting the effect on young workers will be

The deeper question is about learning rather than jobs. Junior work has always served two purposes. It gets useful tasks done, and it teaches people the judgement they need for more senior roles. A trainee accountant learns what matters by checking hundreds of invoices, and a junior developer learns good design by fixing countless small bugs. If AI handles those tasks, organisations need new ways to build that judgement

There are plenty of promising ideas. Apprenticeships and structured training programmes can give young people deliberate practice rather than relying on whatever work happens to come along. Junior staff can review and correct AI output, which teaches them to spot mistakes and understand why things work. Simulations and case studies can recreate the experience of handling real problems. Some firms are already redesigning graduate roles around supervising AI tools rather than doing everything by hand

AI itself could also become a powerful teacher. A patient, always available tutor that explains concepts, gives feedback and adapts to each learner is something past generations could only dream of. Used well, AI could help young professionals learn faster than ever before. The risk is using it as a shortcut that skips the learning altogether. The difference lies in whether people use AI to understand the work or simply to avoid doing it

New kinds of entry level roles are also appearing. Someone has to prompt, check, test and integrate AI systems, and many of those tasks suit people early in their careers. Roles in AI evaluation, data quality, safety testing and human review barely existed a few years ago. History suggests that technology which removes some jobs tends to create others, even if the transition can be uncomfortable. The printing press, the spreadsheet and the internet all reshaped work without ending it

Employers have a big part to play here. Cutting graduate hiring might save money this year, but it creates a shortage of experienced staff a decade from now. Smart organisations will see junior roles as an investment in their future leadership, not just a cost. Governments and universities can help by supporting apprenticeships, updating courses and encouraging lifelong learning. Young people themselves can focus on skills AI struggles with, such as judgement, communication and understanding real customers

My own view is that the bottom rungs of the career ladder are changing rather than disappearing. The routine tasks that once filled a junior's day will shrink, but the need to learn judgement will not. If employers, educators and young workers adapt thoughtfully, the next generation could become experts faster and with better support than any before them. That outcome is not guaranteed, but it is well within reach

Mentoring will matter more than ever. When routine tasks shrink, the informal learning that happens by watching experienced colleagues becomes even more valuable. Firms that pair junior staff with mentors, involve them in real decisions and explain the reasoning behind choices will develop experts faster. Remote and hybrid working has already made this harder in some places. Deliberate effort to keep young people connected with experienced colleagues will pay off

Schools and universities are starting to respond too. Many courses now teach students how to use AI tools well, check their outputs and understand their limits. Some are putting more weight on projects, presentations and practical work that show real understanding. These changes take time, but they point in a good direction. The aim is to produce graduates who can think for themselves and use AI as a powerful assistant

Are you seeing changes in entry level hiring where you work? How did you learn the judgement your job requires, and could that be taught differently now? And what advice would you give someone starting their career today?

Canaries in the Coal Mine? Six Facts about the Recent Employment Effects of Artificial Intelligencesiepr.stanford.edu