LLMs and the Entry-Level Job Market

Sketching out the solution space (rather than just raise the issue)

LLMs will have huge impact on labor markets. However, nobody really knows what will happen even in the most basic of terms, which makes writing & thinking about these matters so exciting: Will this be good for us, or terribly bad? Take your pick along the spectrum from super optimistic outlooks (we only work if we want to, productivity goes through the roof, we have another Industrial Revolution and cure all diseases and cancer) all the way to very bad outlooks (mass unemployment, social unrest because democratic institutions can no longer deliver prosperity).

Within this broader uncertainty, Dario Amodei, CEO of Anthropic (Claude), recently raised a particular sub-issue that is surely worth examining: the immediate impact of LLMs and AI on entry-level white-collar positions.

“AI could wipe out half of all entry-level white-collar jobs — and spike unemployment to 10-20% in the next one to five years”

Amodei does well to put more focus on young people. They are probably one of the most vulnerable social groups, standing at the beginning of their working life (and often have significant debt with student loans). A wrong start into your professional career may be difficult to recover from over the next ~40 years. At the same time, entry-level staff may adapt more readily to LLMs or AI reshaping work practices, so they are a rather different animal from the 55 year old expert that does not really like using any new technology. Also, young people likely will shape public opinion of AI for the next years and decades, so it's crucial that these structural changes work in their favor.

Of course, it’s worth keeping in mind that entry-level jobs have always been special, including in my recent memory and certainly when I was looking for my first job some 20 years ago:

• Hiring any staff, but especially young people is a long-term bet into the future with an initial investment by the employer and a payback period of a few years, even without AI or LLMs around. Any company hiring staff fresh off university or school needs to take that into account

• Salaries for the most junior staff are comparably high compared to the value that they can create for their employer, at least for a few years

• Also, companies need to invest significant amount of time and resources into training entry-level staff and supervising them by more experienced colleagues, often for months if not years

• Still, companies generally see the long-term benefits and hire young people - as they plan for long-term, want to fill their talent pipeline, perhaps also accept their social responsibility

Very good from Amodei to raise the issue of LLMs impacting entry-level jobs, therefore. However, it would have been even better for him to propose a framework to address these issues, or even bring along some more specific solutions. ^_😉 Let’s take his discussion a bit further and look at a range of potential, really preliminary options for decision-makers in business and politics.

I am shooting from the hip - ideas sorted from good to really bad for job seekers. For the sake of good order, some of these solutions could be implemented simultaneously rather than as mutually exclusive options.

The good category:

• Companies shift entry-level hiring toward roles that are inherently human-centric: relationship management, complex problem-solving, or roles requiring physical presence. So, in future, all roles may start in customer service

• Alternatively, companies create hybrid roles where entry-level staff work alongside AI tools as "human-in-the-loop" operators, focusing on judgment, creativity, and client interaction while AI handles routine tasks

• Companies still hire entry-level staff, but look increasingly for tech skills, perhaps prompt engineering or basic programming skills. Students need to adjust for that, as do schools and universities

The "middle" range (around government incentives/training programs):

• Companies still hire entry-level staff, incentivized financially by governments, e.g., with tax credits, lump sum payments, etc. This is costly and may work in more affluent, social-democratic societies, but not necessarily in the Anglosphere?

• Companies set up dedicated apprenticeship or training programs, and hire young people into their programs, probably at lower salaries, with clear progression paths. Staff that do well get taken over into full positions

• Universities and companies create "bridge programs" - extended internships or co-op arrangements that gradually transition students into full employment while providing real-world AI integration training

• Within an industry or regional cluster, companies pool resources to create shared training facilities, spreading the cost of developing AI-literate junior talent across multiple employers

Toward the "worse" end:

• Companies still hire entry-level staff, but at even lower salaries than today, or in the worst case paying no salaries (like unpaid interns). After 2-3 years, staff move into fully paid positions

• Entry-level positions become primarily gig economy roles - companies hire recent graduates as contractors for specific tasks, avoiding long-term employment commitments

• We see a two-tier system emerging where only elite university graduates get traditional entry-level positions, while others compete for lower-paid "AI assistant" roles

• Companies simply stop hiring

Looking at this rather wide option space, I am not necessarily arguing for or against Amodei's statement that LLMs could wipe out half of all entry-level white-collar jobs. That may well happen - but it does not need to happen, provided governments, businesses, and civil society acknowledge that entry-level employment isn't just an economic issue but a question of social stability. This requires coordinated interventions.

---

As always, I'm interested in your thoughts: Feel free to message me. Or, if you prefer, you can share your feedback anonymously via this form.

All opinions are my own. Please be mindful of your company's rules for AI tools and use good judgment when dealing with sensitive data.

---