A graduate can now be told to bring experience to an entry-level interview, which is a lovely little trick if you enjoy circular logic. Software has stripped out the job that used to provide the experience, then the employer acts surprised that the candidate arrives short of it.
This is the quiet absurdity in a lot of AI adoption talk. The easy wins are obvious enough: draft the memo, check the spreadsheet, pull the background notes, answer the routine customer query, even write the first pass of code. Those jobs were never glamorous, but they were where people learned how work actually behaves when a manager is breathing down your neck and the deadline is not interested in your theory.
The first rung was never decorative
The tasks AI removes first
The first rung on the career ladder was mostly built out of repetitive work. Junior staff drafted basic documents, checked numbers against source files, gathered background material, handled inbox sludge and did the kind of routine coding that senior engineers had neither the time nor the patience to do. Under supervision, that work taught process, judgement and the small habits that stop people from making expensive mistakes later.
Software can swallow those tasks fastest. A model can produce a passable first draft in seconds. OCR and machine learning can pull figures from forms and reconcile them against records much faster than a tired clerk. Chatbots can triage customer questions before a human ever sees them. GitHub Copilot and similar tools can spit out boilerplate code and routine test scaffolding. Drudgery disappears first because drudgery is easiest to automate.
IBM said in 2023 that about 30% of customer service roles could be affected by AI automation. This is not a niche back-office issue. Customer service is one of the classic entry points for young workers who need a first job, not a polished LinkedIn narrative. If a company removes the work that teaches people how the business operates, it should not act shocked when the pipeline starts coughing.
Companies are deleting the bridge they depend on
The people who used to learn by doing
In 2026, the World Economic Forum warned that cutting junior roles can leave companies with weaker succession plans, thinner knowledge transfer and senior people stuck doing work that used to train the next generation. Too many firms skate past this. A junior role is not just cheap labour with a laptop. It is the bridge between classroom knowledge and the judgement you only get from being corrected in real time.
Take the usual path in a legal team, a newsroom or an engineering department. You start by checking references, tracing numbers, tidying drafts and learning what a good one looks like. You sit near people who know where the shortcuts are and, more importantly, which shortcuts are traps. Half the education is invisible. It happens in the edits, the raised eyebrow, the quick explanation after the meeting.
Remove those seats and the organisation still has work, but it no longer has apprentices. Senior staff pick up the boring jobs because the machine handled the cheap labour, then they lose time they should have spent on mentoring, planning and decision-making. The company gets a cleaner payroll line and a messier future. Tacit knowledge starts leaking out the bottom because nobody is coming through behind the people who hold it.
Future managers, editors, analysts and engineers do not learn judgement by reading about it. They learn it by doing enough small things badly, then less badly, then well enough that somebody trusts them with something expensive.
The savings have a sting in them
The expensive hunt for talent
The short-term case for automation is embarrassingly neat. Fewer junior hires means lower salary bills, fewer training hours and, on paper, a leaner operation. Finance departments love a clean saving. A company can save money by refusing to grow its own people, then spend far more buying them from someone else.
The Society for Human Resource Management said in 2023 that the cost-per-hire for experienced roles can run at 1.5 to 2 times the cost of filling entry-level positions. This is before you add recruiter fees, sign-on bonuses, relocation, the months spent convincing a scarce candidate to jump ship and the productivity lost while they settle in. If your firm has spent years deleting junior posts, you are not saving talent costs. You are outsourcing them to competitors with better patience.
There is a nastier twist. The employers who keep their entry-level pipelines open will produce the mid-level staff everyone else eventually needs. They will train the people who know the systems, the habits, the awkward corners and the judgment calls. Then the firms that cut their own pipeline will come back later, wallet open, asking to buy the finished product. This is not efficiency. This is paying premium prices for a crop you decided not to plant.
What survives the automation wave
The work that still needs a human to grow into it
The companies that come through this properly will not treat junior roles as sentimental relics. They will redesign them. Some work will be automated, because of course it will. No sane firm wants people manually keying data that software can check in a blink. But the entry point still has to exist, and it has to teach something that the machine cannot hand over on day one.
This means fewer roles built around pure grunt work and more roles shaped around oversight, interpretation and escalation. The junior analyst validates the model’s output. The assistant editor learns how to spot the sentence that sounds fine and is actually nonsense. The new engineer is trusted with small systems, then learns how things fail when the assumptions are wrong. Internships and apprenticeships stop being ceremonial and become structured learning with real feedback attached.
If companies skip that step, they will get the small saving first and the talent shortage second. By the time they notice, the bill will belong to the firms that were boring enough to keep training people while everyone else congratulated itself on trimming payroll.
