The First Rung Is Gone. And What Comes Next Worries Me More.
Entry-level job postings are down 35% since 2023. AI is removing the grunt work where judgment develops — creating an experience gap that won't show up in the workforce until 2030. Research from Harvard, King's College London, McKinsey, and the WEF.
AI is removing entry-level jobs. We all know that.
But the deeper problem — the experience gap that follows — is what nobody is talking about yet.
The most dangerous thing AI has done so far isn't taking senior jobs. It's quietly removed the first rung of the career ladder. And most leaders I talk to haven't noticed what comes next.
I've been thinking about this not just as a trend, but as something with real consequences for the people entering the workforce right now, and for the organizations that will depend on them in 10 years.
Here's what I've observed, what the data shows, and what I think we should actually do about it.
The numbers are harder than most people think
Entry-level job postings in the U.S. are down 35% since January 2023. In the UK, tech graduate roles dropped 46% in a single year. Across Europe, LinkedIn, Indeed, and EURES all report a 35% decline in junior tech positions. Stepstone analyzed 4 million job ads across Europe — entry-level is at its lowest share ever recorded.
Even the most prestigious employers are pulling back. Deloitte cut graduate hiring by 18%. EY by 11%. PwC by 6%. All within one year. The UK's Big Four — firms that have reliably absorbed new graduates for decades — are no longer doing that at scale.
When I look at these numbers together, I don't read it as a cycle. I read it as a structural shift.
Why it's happening — and it's not just the economy
The easy explanation is economic: high interest rates, post-COVID tightening, conservative hiring. And yes, those factors matter. But they don't explain the pattern.
Researchers at King's College London looked specifically at UK firm data and found the drop in junior hiring started exactly when LLMs became widely available — November 2022. Not before. Even within the same firm, roles with higher AI exposure saw steeper cuts than less-exposed roles in the same organization. That separates the AI signal from the macro noise.
A Harvard study confirmed the same: early-career headcount at AI-adopting firms fell 7.7% over six quarters since early 2023. Senior staff? Barely touched.
The reason becomes obvious when you think about what junior jobs actually were. Summarizing documents. Writing first drafts. Basic research. Simple code. Data cleanup. That was the deal — you did the grunt work, you absorbed how things actually worked, and eventually you moved up. McKinsey estimates 60–70% of those tasks are now automated or on their way there.
The grunt work didn't just disappear. The learning pathway inside it disappeared too.
The flip side — and why it worries me more
Here's the argument I keep hearing: "It's fine — juniors can use AI to punch above their weight. They can skip the grunt work and go straight into higher-level roles."
And in the short term, that can be true. 73% of HR professionals in a UK survey said AI allowed them to hire graduates directly into more senior roles. 55.5% of early-career developers already use AI tools daily — more than their senior colleagues.
But when I think about what that actually means 5 or 10 years from now, I don't see an opportunity. I see a problem building underneath the surface.
The grunt work wasn't just inefficiency. It was where you developed judgment. When you spent two years debugging code you didn't fully understand, you built instincts. When you manually constructed a financial model before automating it, you learned what the numbers actually meant. When you wrote 50 mediocre first drafts, you learned what good looked like.
AI removes all of that. And that's the part most people are underestimating.
The long-term risk nobody is pricing in
Stanford and Harvard research put it clearly: AI is strongest at codified knowledge — the kind you learn from textbooks. It is far weaker at tacit knowledge — the judgment, intuition, and contextual awareness that only develops from years of doing the work. Since new graduates enter the workforce primarily with codified knowledge, they are disproportionately vulnerable when those early rungs disappear.
So what happens in 2030 and 2035, when the people who were 22 in 2023 should be stepping into senior leadership roles? They'll have AI skills. They may even have impressive track records of output. But will they have the judgment to lead through a crisis they haven't seen before? Will they know when the model is wrong?
We risk creating a generation of architects who have never laid a brick.
The research at King's College London adds an economic dimension: this expertise gap could reduce annual economic growth by 0.05–0.35 percentage points — manifesting not now, but 10–20 years from now, when today's missing junior cohort should be occupying senior positions.
What actually helps
This isn't a problem without answers. But the answers require being honest about what experience is actually for — and refusing to believe that AI output is a substitute for the judgment that produces it.
Don't outsource the thinking. Use AI for speed — but force yourself to understand what it's producing. Debug it. Challenge it. Ask why. That discomfort is where the real learning happens. The output is easy. The judgment underneath it is the work.
Find people who built things before AI — deliberately. Seek out mentors with 15–20 years of experience who worked before these tools existed. Their judgment — hard-won through failures you've never had to make — is exactly what a model cannot give you.
Own outcomes end to end. Take on projects where you're fully accountable for what succeeds or fails. AI can execute. It cannot replace the growth that comes from having your name on something — and living with what happens next.
Build deep domain knowledge, not just AI breadth. The people creating real value combine genuine domain expertise with AI tools. One without the other is increasingly fragile. AI amplifies what you already know — it can't replace knowing it.
For leaders: rebuild the pipeline deliberately. If your organization has cut junior roles in favor of AI efficiency, ask yourself what you're eating. The short-term gain is real. The long-term cost is a senior leadership pipeline that doesn't exist in 2033. Structured apprenticeships, near-peer mentoring, and project-based learning aren't nice-to-haves. They're how you avoid the shortage.
The entry-level gap is real and it's already here. The experience gap that follows is the problem nobody is talking about yet.
AI gives you the output. Experience gives you the judgment to know if the output is right. You need both. And right now, we're building a generation that may only have one.
Sources: Harvard Business School Working Paper (2025) · Klein Teeselink et al., King's College London (2025) · Stanford Social Innovation Review (2025) · McKinsey State of AI (2025) · Brynjolfsson, Chandar & Chen, Stanford / ADP payroll data (2025) · Autor & Thompson, MIT (2025) · WEF Future of Jobs Report (2025) · Burning Glass Institute (2024) · SignalFire State of Tech Talent (2025) · HiBob UK Worker Survey, n=2,000 (2025) · Stack Overflow Developer Survey (2025) · Stepstone Group EU 4M job ads analysis (2025) · Institute of Student Employers UK (2024/25) · ECB Blog: AI & European firm hiring (March 2026) · U.S. Bureau of Labor Statistics (2025)