The broken skills ladder is AI absorbing the entry-level tasks juniors learned on. See the data on vanishing junior roles and how mentorship rebuilds them.
TL;DR: The broken skills ladder is AI absorbing the entry-level tasks that once built experience. The fix is deliberate skill-building and mentorship, the apprenticeship the job no longer provides.
The broken skills ladder is what happens when AI absorbs the routine, entry-level tasks that early-career workers used to learn on. The bottom rungs people once climbed to build experience are being pulled up out of reach, so junior hiring falls while senior hiring holds steady. Stanford researchers found that workers aged 22 to 25 in the most AI-exposed occupations experienced a 16 percent relative decline in employment, even as older workers in the same roles stayed stable or grew.
The honest read is not that entry-level jobs are vanishing outright. It is that the apprenticeship built into them is disappearing, and the replacement is deliberate mentorship and structured skill-building.
The phrase was popularized by a Deloitte report titled The broken skills ladder, which describes the mechanism plainly: "AI is automating the tasks that have traditionally enabled workers to build low- or mid-level proficiency." The rungs are the small, repeatable jobs (cleaning data, drafting memos, reviewing documents, fixing simple bugs) that taught newcomers how work actually gets done.
The Stanford Social Innovation Review puts the trap in sharper terms. AI "has cannibalized the routine, low-risk work tasks that used to teach newcomers how to operate in complex organizations," producing a paradox where "you need experience to get a job, and you need a job to gain the experience."
That is the whole problem in one sentence. When the tasks that built experience are automated, the ladder still has a top, but the bottom rungs are gone.
Not exactly, and the precise answer matters. Entry-level postings are up in raw volume: about 11 million early-career jobs were posted in 2025, up from 7.3 million in 2018 and 3.2 million in 2012. So the flat claim that entry-level work is gone does not hold.
What is happening is more specific and more troubling. Early-career employment in the fields most exposed to AI is measurably shrinking, and the roles that remain are being redrawn to demand skills that used to arrive years later.
The anchor study comes from Stanford's Digital Economy Lab. Using ADP payroll data, Erik Brynjolfsson and colleagues found that workers aged 22 to 25 in the most AI-exposed occupations saw a 16 percent relative decline in employment, while experienced workers in the same jobs held steady.
The decline is not a one-time shock that has passed. As of April 2026, employment for that group is shrinking about 3.8 percent per year, up from 2.8 percent two years earlier. Brynjolfsson's summary is blunt: "Whatever it is, it's not going away."
A second, independent dataset points the same way. A Harvard working paper, Generative AI as Seniority-Biased Technological Change, tracked nearly 62 million workers across 285,000 firms and found that entry-level employment at AI-adopting companies fell 7.7 percent within six quarters of adoption, relative to firms that did not adopt. The drop came from slower hiring, not layoffs, and senior employment kept rising.
PwC's research reframes "disappearing" as "morphing." Entry-level roles in highly AI-exposed occupations are now 7 times more likely to require skills that historically appeared later in a career.
In those most-exposed occupations, 52 percent of the new skills showing up in entry-level postings were ones traditionally tied to experienced workers, versus just 7 percent in the least-exposed fields. Openings for these redrawn roles grew 35 percent since 2019, while traditional entry-level openings shrank 10 percent.
Put together, the picture is consistent. The first job still exists, but it now asks for the experience the first job used to provide.
The squeeze is not evenly spread. It concentrates in occupations where the daily work is text, code, and structured data, because those are the tasks current AI does well. Deloitte's analysis of US job postings from 2022 to 2025 found firms hiring fewer entry-level data scientists, software developers, and similar roles, even while senior-level hiring in the same fields stayed strong.
That pattern lines up with the Stanford payroll data, where the sharpest early-career declines showed up in the most AI-exposed occupations rather than across the board. Fields built on hands-on, physical, or highly relational work have been far less affected so far. The lesson is not to flee any one field, but to notice which of your tasks are the automatable ones and to build the judgment that sits above them.
AI is not replacing most work outright. Across usage measured by the Anthropic Economic Index, augmentation still makes up 57 percent of activity and automation the remaining 43 percent. The problem is where that automation lands.
It lands hardest on exactly the routine tasks that formed the training ground: first drafts, data cleaning, document review, and straightforward debugging. Those were never just chores. They were the reps through which juniors learned judgment, context, and how an organization really operates.
Worse, the informal mentoring that used to backfill the ladder is eroding too. The LinkedIn 2025 Workplace Learning Report found that only 15 percent of workers said their manager helped them build a career plan in the past six months, a 5-point drop from 2024. The same report notes that 50 percent say managers lack the support to drive career development.
So the two ways people used to build experience, doing the entry-level tasks and being coached by a busy manager, are both thinning at once. That is why the response now has to be deliberate rather than passive.
The ladder is not dead, but its shape has changed. The old model assumed you would be hired into a junior role, absorb the craft by doing low-stakes tasks, and get promoted as you accumulated tacit knowledge. That arrangement depended on employers paying you to learn on the job, and AI has made the cheapest version of that bargain harder to justify.
What replaces it is a ladder you assemble yourself. You still climb from novice to expert, but the early rungs now come from deliberate practice, mentorship, and work you can point to, rather than from a job that quietly trains you in the background. The people who adapt fastest treat learning as an explicit part of the work instead of a byproduct of it.
If the apprenticeship no longer comes bundled with the job, you have to assemble it yourself. The most reliable substitute for a vanished on-the-job apprenticeship is structured mentorship paired with deliberate skill-building, and there is broad appetite for it: Deloitte found that 72 percent of workers and 73 percent of executives believe organizations should do more to create opportunities to gain experience.
Start with what does not work. Stacking generic credentials is a weak fix. The US has nearly 1.1 million unique credentials and spends about $2.1 trillion a year on them, yet only 12 percent deliver significant wage gains. A certificate signals that you sat through material; it rarely proves that you can operate.
The better move is to treat experience as something you produce on purpose, not something you wait to be handed. That means seeking out real problems, shipping work you can point to, and finding someone experienced who will tell you where your reasoning breaks down.
What builds real experience is a tighter loop of practice, feedback, and correction, which is what a mentor provides. Focus your energy on:
Because mentorship no longer arrives automatically, you have to go get it. The good news is that the channels are more open than they have ever been.
Start with your existing network and widen it on purpose. Ask managers, alumni, and people one or two steps ahead in the roles you want, and be specific about the help you need rather than asking someone to "be your mentor."
Then go where practitioners gather. Professional associations, industry communities, and targeted outreach on LinkedIn all put you in front of people who have already climbed the rung you are missing. Consistency matters more than a single perfect ask.
Finally, use tools that make mentorship available on demand. An AI mentor can give you feedback, run practice scenarios, and pressure-test your thinking between conversations with human mentors, which is where a platform like GPTnius fits: it turns deliberate skill-building into a daily loop instead of an occasional favor.
Start building your own rungs with an AI mentor
None of this rebuilds the old ladder. It builds a new one, rung by rung, on your own terms.
The broken skills ladder describes how AI automates the routine, entry-level tasks that early-career workers once used to build experience. Deloitte, which popularized the term, notes that AI is automating the tasks that let workers build low- or mid-level proficiency. The rungs new hires climbed, such as data cleaning and first drafts, are being pulled up out of reach.
Not exactly. Entry-level postings actually rose in raw volume, with about 11 million early-career jobs posted in 2025. What is shrinking is early-career employment in the most AI-exposed fields, where Stanford found a 16 percent relative decline for workers aged 22 to 25. The bottom rung is being pulled up out of reach, not removed entirely.
Stanford researchers found a 16 percent relative decline in employment for workers aged 22 to 25 in the most AI-exposed occupations, while experienced workers held steady. A separate Harvard working paper found entry-level employment at AI-adopting firms fell 7.7 percent within six quarters of adoption. Both point to a shrinking bottom rung, not a collapse of all entry-level work.
Treat experience as something you produce on purpose. Pair deliberate skill-building with structured mentorship, seek out real problems you can ship work on, and get feedback from someone more experienced. Generic credentials are a weak substitute, since only 12 percent deliver significant wage gains. A tighter loop of practice, feedback, and correction is what actually builds judgment.
Because mentorship no longer arrives automatically on the job, you have to seek it deliberately. Start with your existing network, asking managers, alumni, and people one or two steps ahead in the role you want. Then go where practitioners gather, such as professional associations and LinkedIn, and be specific about the help you need. AI mentors can add on-demand feedback between those conversations.