A generation that never had to struggle: what AI is doing to entry-level jobs
A two-tier job market is emerging
There is a quiet problem happening right now in offices, factories, and professional services firms. Entry-level jobs – the kind that school leavers and graduates used to cut their teeth on, are disappearing. And the reason is artificial intelligence.
This is not a future prediction. It is already happening.
The numbers are hard to ignore
Since ChatGPT was launched, vacancies for graduate jobs, apprenticeships, internships, and junior roles that do not require a degree have dropped by 32%. That data comes from Adzuna, a job search site.
It gets worse. According to Randstad’s Workmonitor report, four out of ten employers say they plan to hire fewer graduates in 2026 specifically because of AI. The big accountancy firms - EY, PwC, KPMG and Deloitte – all reduced their graduate intake last year.
These are not small changes. They are a structural shift.
What is replacing the bottom of the ladder?
In the past, most organisations looked like a pyramid. Lots of junior people at the bottom, fewer senior people at the top. Juniors did the basic work. They made mistakes. They learned from those mistakes. Over time, they moved up.
That model is being replaced by what some consultants call a “diamond”. Fewer junior roles. A large middle group of experienced staff who are trained to use AI. And a small number of senior people who act as checkpoints.
Here is how one consultant put it: “I can orchestrate a load of agents together to do an entire process from beginning to end, where I only need humans as checkpoints.”
That sounds efficient. But it creates a serious problem. If young people never do the basic work, how do they learn to judge situations? How do they learn to make decisions under pressure? How do they learn to recover from mistakes?
The risk: a generation that never struggled
One business leader called this “a generation that has never struggled”. The concern is not about being kind to young people. It is about the opposite.
Struggling through difficult tasks – tedious ones, frustrating ones, ones that might fail, is how human beings build judgement. It is how they learn to tell the difference between a good decision and a bad one. It is how they build resilience.
If all the difficult, messy, entry-level work gets handed to AI, then young people enter the workforce without having gone through anything truly hard. They become good at asking AI for answers, but bad at knowing which questions to ask in the first place.
The labour market is becoming two-tier
A leader from Microsoft described the trend as a polarized labour market. On one side, there is growing demand for people with deep, specialist expertise. On the other side, there is very little demand for generalists or beginners.
He also predicted that only a small number of workers will use advanced AI tools. For everyone else, the most valuable skill will not be technical knowledge. It will be adaptability – the ability to change approach when circumstances change.
But here is the catch. Adaptability is not something people are born with. It is something they learn by being put in unfamiliar, uncomfortable situations. In other words, by struggling.
What needs to happen
Employers cannot wait for the government to fix this. Politicians move slowly. The goal should be to use AI to help people do their jobs better, not to replace them. That is a sensible statement. But it is not a plan.
The real responsibility sits with employers, HR teams, and business leaders. Here is what that means in plain terms:
1. Protect some entry-level roles on purpose: Do not automate every single basic task. Leave some work for juniors to do manually, even if it is less efficient in the short term.
2. Let young people make mistakes: If a junior employee never fails at something small, they will never learn how to fix problems when they are large. Managers need to allow controlled failure.
3. Teach judgement, not just tool use: AI literacy is useful. But knowing how to use ChatGPT is not the same as knowing whether an answer is sensible. Judgement comes from practice, not from a training video.
4. Do not replace the bottom of the ladder and then complain that there is no mid-level talent: That is already happening. Companies cut junior hires, then struggle to find experienced staff. The two things are directly connected.
5. Measure the right things: Do not only measure how fast work gets done. Measure whether junior staff are being given difficult, non-routine decisions to make. If the answer is no, then the learning pipeline is broken.
A simple warning
Here is the bottom line. Every time an AI agent replaces a junior employee’s basic task, something is saved – time, money, effort. But something is also lost. That something is the messy, slow, human process of learning how to think under pressure.
If that loss happens across thousands of companies for five or ten years, the result will be a workforce full of people in their late twenties who have never truly struggled. They will be technically competent. They will be able to prompt an AI. But when the AI gives a bad answer, or when the system fails completely - they will not know what to do.
That is not a future anyone should want.
Employers can still change course. But the window is closing. Every hiring cycle that replaces a junior role with an AI agent makes the problem worse.
The question is not whether AI is useful. It is clearly useful. The question is whether organisations are willing to keep some room for human struggle - not because it is inefficient, but because it is the only way to build the next generation of capable, confident workers.