Entry-level jobs and AI: what is really happening to first jobs

Ask a recent graduate how the job hunt is going and you will hear a version of the same story: hundreds of applications, a handful of automated rejections, one or two interviews, and a growing suspicion that the jobs they trained for are being done by software. Ask a chief executive and you may hear the mirror image, sometimes stated with pride: the company is not hiring juniors this year because AI handles the work they used to do.

Both stories are partly true and both are being told with more confidence than the evidence supports. This piece is an attempt to separate what is actually known about AI and entry-level work from what is being claimed, to identify which first jobs are changing and how, and to set out what someone at the start of a career, or restarting one, should do differently as a result. It is analysis of a trend that is still unfolding, and where we are uncertain we say so.

What is known

Start with the parts that are well established.

The market for new graduates has been weak for several years. In the US, the Federal Reserve Bank of New York’s tracking of labor market outcomes for recent college graduates has shown their unemployment rate running above the rate for all workers, a reversal of the historical pattern in which a degree bought you a lower rate than the average. Underemployment (graduates in jobs that do not require a degree) has stayed high. In the UK, graduate vacancy trackers from the Institute of Student Employers and similar bodies have reported flat or falling graduate recruitment among large employers, and the Office for National Statistics has recorded a sustained fall in job vacancies from their 2022 peak.

Postings for junior roles in specific fields have fallen hard. Indeed’s Hiring Lab data on software development postings shows them running well below pre-pandemic levels since 2023 and not recovering with the rest of the market. Similar patterns have been reported for customer support, some marketing and content roles, and junior positions in professional services.

Certain employers have said, on the record, that they are reducing junior hiring because of AI. Firms in technology, consulting, law and finance have made public statements to that effect, and some have paired them with announcements of flat or reduced graduate intakes. Those statements are real even if the causal claim inside them is untested.

And there is some early academic work. A widely discussed 2025 study by researchers at Stanford, using payroll data from a large US provider, found that employment for workers in their early twenties in occupations most exposed to generative AI had declined since late 2022, while employment for older workers in the same occupations and for young workers in less exposed occupations had not. The authors were careful about causation and the finding has been debated, but it is the most direct evidence so far that something specific is happening at the junior end of AI-exposed work.

What is claimed but not established

Now the parts that get repeated as fact and are not.

That AI is the main cause of the weak graduate market. The timing is suggestive but confounded. The same period saw interest rates rise sharply, which hit the technology and professional services sectors that had over-hired in 2021 and 2022. Companies that cut junior intakes in 2023 were mostly reversing a hiring binge, and many said so at the time. Offshoring of junior work to lower-cost countries accelerated in parallel. Untangling AI from those forces is genuinely hard, and most economists who have tried say the AI-specific effect is real but smaller than the headlines.

That whole occupations are disappearing. They are not, at least not yet. Employment in accounting, law, software and marketing remains large and mostly stable. What is changing is the composition of the work and, at the margin, how many people at the bottom rung are needed to do it.

That the effect is uniform. It is not. The evidence is concentrated in a narrow set of desk-based, text-heavy, process-heavy junior roles. Entry-level jobs in healthcare, the trades, logistics, hospitality, education and care are largely unaffected by generative AI and several are short of people.

That companies know what they are doing. Some of the “we don’t need juniors” claims will age badly. A firm that stops training juniors does not have mid-level staff in four years, and several industries have run this experiment before with offshoring and regretted it. There are already signs of employers quietly restarting graduate programs after finding that the work did not do itself. Whether this becomes a broad correction is one of the open questions.

Which entry roles are changing, and how

The useful question is not “is AI taking jobs” but “what did the first job in this field consist of, and how much of it can now be done faster by a tool in the hands of someone more senior.”

Junior software development. The classic first job (implement a well-specified feature, write tests, fix small bugs) is the work that coding assistants do best. Employers still hire juniors, but fewer of them, and they expect new hires to arrive able to use the tools and to review what the tools produce. The market for bootcamp graduates with no other background has thinned badly; our honest read of that is at /training/are-coding-bootcamps-still-worth-it/.

Customer support and service. First-line chat and email support was already being automated before generative AI; it has accelerated. The remaining human roles skew toward escalation, complex cases and voice, which are harder for a new entrant to step into.

Paralegal and legal research. Document review, first-draft research memos and contract summaries are exactly what current tools handle competently. Law firms have not stopped hiring paralegals, but the junior work is shifting toward checking, client contact and process management.

Content, copywriting and junior marketing. Volume content production has been hit hardest, and freelance rates for generic writing have fallen. What is left values judgement, brand knowledge, distribution and measurement more than raw drafting.

Data and financial analysis. The mechanical parts (cleaning data, building standard reports, first-pass reconciliation) are increasingly tool-assisted. The junior analyst role is not vanishing, but the expectation has moved toward interpretation and communication earlier in a career. We set out the day-to-day at /careers/what-does-a-data-analyst-actually-do/.

Across these, the common pattern is that the tasks used to be how juniors learned the job. A trainee lawyer who never does document review, or a developer who never writes boilerplate, misses the repetitions that build judgement. Employers have not solved this. The ones who are thinking about it are redesigning junior roles around supervised use of tools; the ones who are not are simply hiring fewer people and hoping.

The roles that are not changing much

It is worth stating plainly, because graduates in particular tend to forget it: most entry-level jobs in the economy involve a body in a place doing something physical or interpersonal. Nursing assistants, apprentice electricians, warehouse leads, chefs, early-years educators, HVAC technicians, care workers, truck drivers. Several of these fields report shortages, offer paid training, and lead to careers that pay comparably with the desk jobs graduates are chasing. The shift in employer attitudes toward degree requirements, covered at /news/skills-based-hiring-is-growing-what-it-means-for-people-without-degrees/, cuts both ways: it opens doors for people without degrees and it reminds people with degrees that the paper is worth less on its own than it was.

None of this means a graduate in English should retrain as a plumber. It means the map of “safe” first jobs has changed, and the desk job is no longer automatically the safer choice.

How to position yourself if you are starting out

The advice below is drawn from what employers who are still hiring juniors say they want, filtered through some skepticism about what they say versus what they do.

Be visibly fluent with the tools, and visibly able to check them. The junior who can produce a first draft with an AI assistant and then explain what is wrong with it is more useful than either the junior who refuses to use the tools or the one who trusts them. In applications and interviews, describe specific work where you used a tool, found its errors and fixed them. That story is currently rare and it lands.

Bring evidence of finished work, not potential. Employers that have cut junior intakes are hiring fewer, more proven people. A portfolio, a project with real users, a published analysis, a contribution to someone else’s project: anything that shows you can complete something without supervision. This matters more than grades and more than the name on the degree.

Target smaller employers and less glamorous sectors. The graduate schemes at big consultancies, banks and tech firms are where the cuts are concentrated and where the competition is worst. Mid-sized companies, regional firms, public bodies and unfashionable industries still need juniors, often cannot afford senior hires, and get far fewer applications.

Use internships, co-ops and placements aggressively. Work-integrated routes have always been the reliable path into a first job. In a market where employers are nervous about untested hires, a placement that lets them see you work is worth more than ever. Apply early and be willing to take one that is not perfect.

Treat the first job as an apprenticeship, whatever it is called. The roles that survive are the ones where you learn to exercise judgement. Ask in interviews how juniors are trained and what they are expected to be doing in two years. An employer with a clear answer is worth taking a lower salary for. One that shrugs is telling you it plans to use you until the tools improve.

Build the skills that are hard to automate and easy to see. Speaking clearly in a meeting, writing an email that gets a decision, managing a small project to a deadline, handling a difficult customer. These are not soft skills, they are the parts of junior work that employers report are hardest to replace, and they are learnable in almost any job, including part-time and service work you might be doing to pay the rent.

If you are a career changer rather than a graduate

Career changers have an advantage that is easy to overlook: they already have the judgement and workplace skills that new graduates lack, and what they need is the domain knowledge and tool fluency. In a market that is short of juniors who can be trusted, someone with ten years in another field who has retrained credibly is often a better bet for an employer than a 22-year-old, particularly at a smaller company. Lead your application with what you have done, and treat the new skills as the thing you have added, not as the thing you are.

What to watch over the next couple of years

Three signals will tell you whether this is a structural change or a cyclical dip with a good story attached.

Whether graduate intakes at large employers recover as interest rates and the broader market settle. If they do, the AI story was overstated. If they stay down while general hiring recovers, it was not.

Whether the “we don’t need juniors” firms start reporting mid-level shortages and reopen training routes. Early signs of this exist; whether they become a trend is the open question.

Whether junior roles are being redesigned rather than removed: new titles, new training structures, apprenticeship models in desk-based industries. That would suggest employers are adapting rather than just cutting.

In the meantime, the practical position for someone starting out is unchanged by any of the debate: apply widely, favor employers who can say how they train people, show finished work, and learn the tools well enough to be the person who catches their mistakes. That was good advice before anyone was worried about AI. It is now the whole game.