Entry-level jobs are supposed to be where people learn the skills they will need later in their careers. New hiring data suggests AI is changing that sequence. In occupations most exposed to AI, employers are increasingly asking junior candidates to bring judgment, leadership and strategic thinking much earlier.
An entry-level job has traditionally come with an understood trade.
The employer gets someone with limited experience. The employee gets a place to build it.
That bargain becomes harder to maintain when technology can perform many of the basic tasks that once gave junior workers their first experience.
A 2026 analysis from PwC examined 2.4 million US entry-level job advertisements and found that the most AI-exposed junior roles were seven times more likely than the least AI-exposed roles to demand skills that had traditionally been associated with experienced workers. The analysis of how AI is changing entry-level work identified skills including strategic decision-making, motivational leadership, team building, stakeholder management, process management, mentorship and data-driven decision-making.
The underlying numbers make the finding easier to understand.
Among newly appearing skills in the most AI-exposed entry-level occupations, 52% had historically been concentrated in experienced roles. In the least AI-exposed occupations, only 7% of newly appearing skills met that definition.
That difference produces the roughly sevenfold gap.
The finding does not mean companies are turning every graduate into a manager. It suggests that the content of some junior jobs is changing. As AI takes over more routine parts of the work, employers appear to be asking the person who remains to handle tasks requiring more interpretation, judgment and responsibility.
For people starting their careers, that creates an uncomfortable question: how do you acquire senior skills if the entry-level work that used to develop them is disappearing?
The job title may stay junior while the work inside it changes
Consider how a junior analyst might have worked several years ago.
A large part of the job could involve collecting information, cleaning spreadsheets, summarizing reports, preparing slides and producing first-pass analysis for someone more experienced to review.
None of those tasks was glamorous.
They did, however, teach the analyst how the organization worked. Repetition exposed them to different clients, decisions and mistakes. A senior colleague could see where they struggled and gradually hand them more difficult work.
AI can now perform meaningful parts of that preparation.
The junior analyst may arrive at the interesting part of the assignment much sooner. Instead of spending hours collecting information, they may begin with an AI-generated summary. Instead of manually producing the first analysis, they may review what a model has already produced.
That can make the employee more productive.
It also changes what the employer needs from them.
If the machine handles more of the mechanical work, the human contribution moves toward deciding whether the analysis makes sense, understanding what matters to the business and explaining what should happen next.
Those are exactly the kinds of capabilities that used to become more important after someone had accumulated experience.
The technology therefore does not simply make an old entry-level job faster. It can change what qualifies someone to perform the job at all.
The 7× number comes from a very specific definition of a senior skill
This is one statistic where the methodology matters enough to understand before drawing conclusions.
The researchers defined an entry-level job as a posting asking for zero to two years of experience.
They then compared the skills appearing in those jobs in 2025 with the skills appearing in the same occupations in 2019.
A skill counted as “new” when it appeared more than ten times in 2025 entry-level postings for an occupation but five or fewer times in equivalent postings in 2019.
The researchers then asked whether those newly requested skills had historically belonged to more experienced workers.
A skill was classified as traditionally senior when, within the same AI-exposure group, it appeared more than 50 times in experienced job postings in 2019 but five or fewer times in entry-level postings.
That produced a striking result: 52% of the newly appearing skills in the most AI-exposed entry-level occupations had previously been concentrated in experienced roles, compared with 7% in the least exposed occupations.
Examples included strategic decision-making, people management, stakeholder management, mentorship and leadership.
This methodology is more interesting than simply counting how often words such as “leadership” appear in job descriptions. It is looking for skills that have migrated down the career ladder over time.
In other words, employers are not merely asking juniors for more skills.
They are increasingly asking some juniors for skills that six years earlier were rarely requested from people at their level.
Entry-level work has always been part job and part apprenticeship
One reason this matters is that junior jobs serve two economic purposes.
They produce useful work today, but they also develop workers who will be capable of doing harder work tomorrow.
A new accountant learns by reviewing transactions before advising clients on difficult financial decisions. A junior developer fixes small bugs before making architectural choices. A researcher gathers evidence before deciding which evidence should drive a recommendation. An early-career marketer executes campaigns before being responsible for brand strategy.
The simpler work creates repetition without putting the most important decisions in inexperienced hands.
Over time, the worker develops pattern recognition.
They see what goes wrong, which details matter, how senior colleagues make decisions and how customers respond. Eventually, the person can exercise judgment because they have experienced enough situations to develop it.
Generative AI complicates that system because many of the tasks most attractive to automate are also the tasks traditionally assigned to beginners.
The company can gain efficiency by removing repetitive junior work.
The long-term problem appears if it removes the learning mechanism at the same time.
A business still needs experienced people in five or ten years. Those people have to gain experience somewhere.
The most interesting finding is that some entry-level jobs are growing, not disappearing
The entry-level AI debate is often framed as a question of whether junior jobs will survive.
The data suggests a more complicated outcome.
Within the most AI-exposed US occupations, entry-level postings that had become “seniorised” grew 35% between 2019 and 2025. Entry-level roles in the same highly exposed group that had not added traditionally senior skills declined 10%.
That difference is important.
It suggests that the entry-level jobs performing relatively well are not necessarily the jobs AI has left untouched. Some are the jobs being redesigned around more demanding human capabilities.
The full labor-market analysis also found that early-career vacancies in the most AI-exposed occupations had broadly flatlined across the advanced economies for which sufficient experience data was available, while less-exposed groups continued to grow.
The report explicitly warns against treating that relationship as proof that AI caused the slowdown. The international analysis included Canada, Singapore, the United Kingdom and the United States, with US postings representing roughly 73% of the sample. Interest rates, industry changes, post-pandemic hiring corrections and other structural factors can also affect entry-level recruitment.
The combination of findings is still revealing.
The clearest divide inside highly exposed entry-level work is not simply between jobs that use AI and jobs that do not.
It is between junior roles that are becoming more demanding and those that remain closer to the old model of entry-level execution.
Independent payroll data shows pressure on young workers too
Job advertisements tell us what employers say they want.
Payroll records provide another view of what is actually happening to employment.
A Stanford Digital Economy Lab study using payroll data covering millions of US workers found no evidence of widespread economy-wide employment displacement from generative AI through June 2026.
The results for younger workers were different.
Among workers aged 22 to 25 in highly AI-exposed occupations, employment stood about 19% below the level it would have reached if it had kept pace with similarly aged workers in less-exposed occupations. More experienced workers did not show a comparable gap.
The researchers found that the difference appeared primarily through lower hiring of young workers rather than unusually high rates of existing young employees being dismissed.
They also found an important distinction between different uses of AI. Employment weakness was concentrated in occupations where AI was being used more to automate human tasks. In occupations where AI was used primarily to complement workers, employment was flat or rising, particularly among more experienced employees.
The study is a working paper, and its authors do not claim that the entire weakness facing younger workers can be attributed to AI. Earlier versions of the research also found that some employment changes preceding 2024 were likely influenced by other economic factors.
That caution matters.
There has been a broad slowdown in early-career hiring, particularly after the unusually strong labor market that followed the pandemic. AI is arriving during that adjustment rather than in an otherwise unchanged economy.
Still, the payroll evidence and the job-posting evidence raise a similar concern from different directions.
In AI-exposed work, younger employees appear to be facing a labor market that increasingly rewards the capabilities associated with experienced workers.
AI creates an unusual advantage for beginners at the same time
There is a paradox here.
Some of the best experimental evidence on generative AI shows that less experienced workers can benefit from the technology more than experts.
A large study of 5,172 customer-support agents published in The Quarterly Journal of Economics found that AI assistance increased productivity by about 15% overall. The gains were substantially larger among less experienced and lower-skilled workers, while the most experienced workers benefited much less.
The researchers found evidence that the system helped transfer some of the practices used by stronger workers to employees who had not yet developed the same experience.
This suggests AI can function as a powerful training aid.
A beginner can receive assistance at the exact moment they encounter a problem. They can see examples of stronger responses and deal successfully with situations they might previously have needed to escalate.
That creates two very different possibilities for entry-level work.
A company can use AI to reduce the number of junior employees because experienced workers can now complete more work themselves.
Or it can use AI to make junior employees productive faster and accelerate how quickly they acquire expertise.
Both outcomes are technologically possible.
Which one becomes more common depends on how companies redesign work.
Removing routine work is easy. Replacing what workers learned from it is harder.
Organizations have a strong incentive to automate tasks that are repetitive, time-consuming and relatively easy to evaluate.
From a productivity perspective, that makes sense.
The difficulty is that inefficient-looking tasks sometimes perform a second function.
A junior lawyer reviewing large numbers of documents is not only moving documents through a process. They are repeatedly seeing how legal arguments, contracts and disputes are structured.
A new developer debugging ordinary problems is not only fixing bugs. They are building an understanding of why systems fail.
A junior consultant preparing basic analysis is also learning what senior colleagues consider important enough to challenge.
Not every repetitive task deserves to survive for educational reasons. Forcing employees to spend hundreds of hours on work a machine can perform better would not be a sensible talent strategy.
The alternative, however, cannot simply be to remove the repetition and assume judgment will appear automatically.
Companies need another way to create the experience.
That could mean giving junior employees more supervised decision-making, allowing them to review and challenge AI output, rotating them across different kinds of problems or making senior reasoning more visible instead of leaving beginners to infer it from years of routine work.
The training architecture has to change alongside the job architecture.
Employers increasingly expect AI fluency before people arrive
The career ladder is also changing from the candidate side.
Young workers are entering the labor market having used generative AI throughout much of their education, but employers increasingly expect them to convert that familiarity into professional capability.
Handshake’s 2026 research on students and early-career hiring found that 85% of graduating seniors used AI, with more than a third using it daily. At the same time, more than 10% of active internships mentioned AI-related terms by March 2026, while the share of full-time postings mentioning AI had nearly doubled over the previous year to 4.2%.
The education system has not moved at the same pace.
Fifty-eight percent of graduating seniors said they would need stronger AI skills to succeed at work, while only 28% said their academic program had meaningfully integrated AI.
That creates another version of the experience problem.
Employers increasingly want graduates who are comfortable with AI, capable of exercising judgment and ready to contribute quickly.
But students cannot acquire all of those abilities simply by learning how to operate the technology.
Knowing how to use an AI assistant is different from knowing whether its recommendation makes sense in a real business situation.
The second capability depends heavily on context and experience.
The new entry-level advantage may be evidence of judgment
For years, junior candidates competed largely through credentials that signaled potential.
A degree showed education. An internship showed some exposure to professional work. Technical tests showed whether someone possessed a particular skill.
If entry-level jobs continue moving toward more advanced human capabilities, candidates will face a different challenge: demonstrating judgment before they have a long employment history.
That makes practical work more useful.
A developer who can explain why they rejected an AI-generated implementation demonstrates more than someone who simply says they know how to use coding assistants.
A marketer who can show how they interpreted weak campaign data and changed direction demonstrates decision-making.
A researcher who can explain why several apparently credible sources were excluded demonstrates evaluation rather than information retrieval.
Projects, internships, freelance assignments and other real work can provide evidence of those capabilities, although access to those opportunities is itself uneven.
This matters because the strongest human skills are often difficult to prove through a résumé keyword.
Leadership, resourcefulness and judgment become credible when there is a decision behind them.
For businesses using outside specialists, the same movement is already visible later in the labor market. Flexable’s research into the growth of jobs fixing AI-generated work found rising demand for professionals who can evaluate and improve work after AI has already produced the first version.
Entry-level hiring may be moving in a similar direction. Generating the first output matters less when the difficult part of the job is knowing whether that output deserves to be used.
Companies can create a talent problem by optimizing only for today’s productivity
Reducing junior hiring can make financial sense in the short term.
If one experienced employee using AI can perform work that previously required an experienced employee plus two junior staff members, the immediate productivity calculation is attractive.
The long-term calculation is harder.
Five years later, the business still needs experienced employees.
If companies across an industry reduce the number of beginners they train, the supply of people with five years of experience eventually changes too.
This is why entry-level jobs have always had value beyond their immediate output. They are part of the system that produces future managers, specialists and experts.
Employers appear to understand that risk.
Separate early-career research found that 83% of hiring managers considered entry-level talent important to their organization’s future success, while 51% of managers expecting to increase early-career hiring said one reason was the need to build a future leadership pipeline.
The tension is clear.
Companies want the efficiency available from AI today while still needing experienced humans tomorrow.
Solving both problems requires something more sophisticated than eliminating junior work wherever automation is possible.
Entry-level hiring is weak for reasons beyond AI
Any report on this subject also needs to resist assigning every difficult trend facing graduates to generative AI.
The US labor market has undergone several large changes since 2019.
Companies hired aggressively during parts of the pandemic recovery and later corrected those staffing levels. Interest rates rose sharply. Technology firms went through a major hiring contraction. Employee turnover fell, creating fewer openings as experienced workers stayed in their positions longer.
Early-career recruiting data reflects those broader pressures.
The Class of 2026 entered a market where overall early-talent job postings were below their pre-pandemic level, while student pessimism about finding work had increased substantially.
Some researchers looking at early-career hiring have therefore been more cautious about attributing the entire slowdown to AI.
That is the correct caution.
The 7× finding is strongest when interpreted as evidence about changing skill requirements, not as proof that AI alone caused the decline in entry-level opportunities.
It tells us what employers increasingly request inside highly exposed junior jobs.
It does not by itself establish why every missing junior vacancy disappeared.
What the 7× figure actually means
The headline statistic comes from an analysis of 2.4 million US entry-level job advertisements.
Entry-level postings were defined as jobs requiring between zero and two years of experience.
Occupations were grouped into quartiles based on an AI exposure index that estimates how closely the abilities required in each occupation overlap with capabilities AI systems can perform.
That is another important distinction.
A job being highly AI-exposed does not mean the employer has already automated it. Exposure measures the potential for AI capabilities to affect the tasks associated with that occupation.
Researchers then compared skills requested in 2025 with the same occupations in 2019.
In the most AI-exposed quartile, 52% of unique new skills appearing in entry-level postings qualified as traditionally senior. In the least-exposed quartile, the equivalent share was 7%.
That ratio is the basis of the 7× headline.
The report also notes a methodological limitation: experienced job postings are more numerous than entry-level postings, which can make some skills more likely to appear in the experienced sample. The researchers counted each unique skill once in the historical comparison and argued that the difference was unlikely to materially change the result.
None of this means every new graduate now needs to be capable of managing a department.
What it does mean is that employers in highly AI-exposed fields are adding capabilities to junior jobs that were historically much more common further up the career ladder.
That changes the definition of being ready for an entry-level job.
For employers, the challenge is making sure higher expectations come with a credible way for young workers to acquire the experience those expectations require.
For people entering the workforce, the challenge is different. Learning to use AI is increasingly part of being employable, but it may not be the hardest part.
The harder task is developing the judgment to know what to do after AI gives you an answer.