The first rung of the career ladder is gone. Investor Carsten Maschmeyer named the problem — the data confirms it, and companies are already restructuring around it.
A economics graduate, top of her class, first week at a consulting firm. The assignment: a market analysis for a logistics client. Two years ago, that meant three days of junior grunt work — research, synthesis, slide deck. Now the senior partner runs a few prompts, refines the output in twenty minutes, and calls it done. The junior watches, wondering what she's supposed to be learning.
That scene isn't exceptional. It's Tuesday.
German entrepreneur and investor Carsten Maschmeyer said publicly what most people in knowledge industries are quietly noticing: AI agents are eliminating the first rung of the career ladder. The tasks that trained entry-level employees at agencies, consulting firms, and law firms — the repetitive, structured, "here's how it's done" work — are being automated at scale. His historical analogy cuts deep: before industrialization, apprentices paid Lehrgeld — a fee to the master craftsman for the privilege of learning. Experience carried a price, not a wage. Maschmeyer worries that model might return — not in the trades, but in the white-collar professions.
It's not a thought experiment. The numbers are already here.
Klingt interessant?
What the Data Shows
Venture capital firm SignalFire analyzed hiring data across the largest public tech companies and mature VC-backed startups between 2019 and 2024. They found a 50% decline in new role starts by people with less than one year of post-graduate experience — and that drop was consistent across every core business function: sales, marketing, engineering, HR, design, finance, legal. Not one department, not a tech-specific blip. Fifty percent, everywhere.
LinkedIn's workforce data tells a similar story: national hiring in the US slowed nearly 9% year-over-year in late 2025 and remains more than 20% below pre-pandemic levels. The National Association of Colleges and Employers projects just a 1.6% increase in hiring for the class of 2026 — which, given the number of new graduates entering the market, is a functional contraction. Nearly half of employers now rate the job market for new graduates as merely "fair." In previous years, "good" or "excellent" were the norm.
Anthropic CEO Dario Amodei has put a number on the direction of travel: AI could eventually eliminate up to 50% of entry-level knowledge worker jobs. The IMF estimates that up to 60% of all jobs in advanced economies will be affected by AI in some form. You don't need to take either figure literally to see that the disruption is already happening — and that it starts at the bottom, not the top.
The evidence isn't just in projections. It's in law firms that have halved associate research hours with AI, agencies running with half their former junior headcount, and development teams that are deliberately not growing because the alternative is faster and cheaper.
The Real Problem: Where Does Anyone Learn Now?
Maschmeyer's historical comparison works because it names the underlying mechanism. Entry-level jobs were never really about affordable labor — they were a bilateral deal. Employers got cheap capacity; employees got the structured exposure that turns a degree into actual competence. The system functioned because both sides benefited.
When AI handles that work faster and cheaper, the deal collapses. It's not that companies have stopped liking junior hires. It's that the math stopped working. The junior analyst who used to take three days to produce a market report now competes not with a more experienced colleague, but with an agent that has no onboarding period, no salary expectations, and no sick days.
What remains is the question Maschmeyer is really asking: where do knowledge workers acquire the foundational craft that used to come from the first two years on the job? The Lehrgeld model — paying for the right to learn — sounds absurd, but it has a perverse logic when AI makes the apprentice years obsolete. Whoever does get hired is expected to operate at a senior level from day one, without the scaffolding that repetitive entry-level work quietly provided.
That's the structural problem: not a shortage of jobs in aggregate, but the disappearance of precisely the jobs that teach you how to make decisions.
What Companies Are Actually Buying
For companies, the picture is less grim but no less significant.
Hiring a junior developer always meant buying the time of a young person to build things. Three months to get oriented, a year to reach productivity, two years until genuine autonomy. That was always a bet on patience as much as talent — and for many organizations, particularly smaller ones, it was simply too expensive to make.
That model is shifting. Not because junior developers have gotten worse, but because the alternative — AI agents that write code, generate documentation, implement routine features, run code reviews — is faster, cheaper, and scales without a headcount cap. You're no longer buying someone's early career years. You're buying outcomes.
That sounds cold. It is cold. But it's the direction. Companies that understand this early are restructuring around smaller, experienced core teams who set direction and own decisions — with AI capacity handling execution. The first rung doesn't disappear because no one wants to start anymore. It disappears because the model that made it function has been replaced by a different one.
SignalFire's Heather Doshay put it plainly: "The bottom rung is disappearing — but that has the potential to uplevel everyone." The optimistic reading is that new graduates will have to arrive already proficient with AI, and that "entry-level" will shift up the skill ladder rather than vanish entirely. The honest caveat she added: that doesn't help the graduating classes of 2024, 2025, or 2026 very much right now.
Maschmeyer's Advice: Take a Trade Seriously
Maschmeyer's suggestion is blunt: if you're deciding on a career path today, genuinely consider a trade. The demand is large and almost no skilled trade is exposed to this particular wave of disruption.
That's not nostalgia. It's a sober read on automation resistance. An electrician, a plumber, a mechatronics engineer — these roles depend on physical execution, on-site problem-solving, and handling the unexpected in environments that no language model can navigate. The answer isn't always on a screen.
For those committed to knowledge work, the signal is clear: raw knowledge is no longer a moat. AI has that too. What isn't automatable is judgment — the ability to turn AI output into a binding decision, to know when the model is wrong, to carry the weight of accountability. Those who sharpen that are building a lead that the next model release won't erase.
The first rung is already gone for most. What you do with that reality is the question — for graduates entering the market, and for companies deciding right now how they build capacity. At Nopex, those capacities — the work that used to fall to entry-level hires — run as specialized AI agents: building features, writing documentation, handling code reviews, automating workflows. The model isn't junior developer on probation. It's outcome.
How AI agents at nopex replace development capacity — nopex.cloud
