The path into a career used to have a reliable first step. You graduated, you took an entry-level role heavy on routine tasks, and you learned the trade while doing the grunt work. That first rung is now harder to reach. This is not the familiar layoffs story about senior staff being cut. It is something narrower and, for anyone just starting out, more personal: companies are quietly hiring fewer juniors, and part of the reason is that AI now does much of the codifiable work that early-career hires were once brought on to handle.

The good news up front: this is a real shift, not the end of ambition. Openings still exist, and the people who understand what changed are better positioned to find them.

The data on the shrinking first rung

Start with a genuinely unusual signal. For most of modern history, college graduates enjoyed lower unemployment than the general workforce. That relationship has flipped. According to the Federal Reserve Bank of New York's "Labor Market for Recent College Graduates" data, unemployment for recent grads (ages 22 to 27) sat around 5.3% in early-to-mid 2025, well above the roughly 4% national rate, and it climbed further as the year went on. The same NY Fed data put underemployment, the share working in jobs that do not require a degree, above 42% in late 2025, the highest reading since 2020. The Federal Reserve Bank of St. Louis reached the same conclusion in its 2025 analysis: recent grads are bearing the brunt of the current labor-market shift.

Underneath the unemployment numbers is a hiring pattern. Employers are posting fewer roles aimed at people with no experience. In a July 2026 survey, Gartner found that AI automation is reducing some entry-level hiring at nearly a quarter of organizations. Labor-analytics reporting through 2025 pointed the same way, with entry-level postings down meaningfully from their 2022 peak.

The causes: AI, but not only AI

The most cited evidence linking this specifically to AI comes from the Stanford Digital Economy Lab. Their study "Canaries in the Coal Mine?" by Erik Brynjolfsson, Bharat Chandar, and Ruyu Chen, using payroll data from ADP, found that early-career workers aged 22 to 25 in the most AI-exposed occupations experienced roughly a 16% relative decline in employment since generative AI took hold, even after controlling for firm-level shocks. The declines cluster in fields such as software development and customer service, exactly where a chatbot can now draft, debug, or answer at scale. Brynjolfsson's blunt summary is that what younger workers know overlaps heavily with what large language models can already do.

Here is the honest nuance you deserve. The Stanford authors themselves, in a later update, noted that once you add the broadest set of controls, the AI-attributable decline becomes statistically clear mainly from 2024 onward, meaning earlier softness was likely driven by other forces. Those forces are real: higher interest rates cooled interest-rate-sensitive hiring, and the tech sector was correcting a pandemic-era over-hiring binge. AI is a major thread in this story, not the whole cloth. Anyone telling you it is 100% robots, or 0%, is selling something.

What is clearest is a change in what firms hire for. The routine, codifiable tasks that once justified a junior headcount, first-pass research, boilerplate drafting, basic data entry, ticket triage, are increasingly automated. The roles that remain lean toward judgment, communication, and supervising the machine's output. That reframes the whole challenge of breaking in.

Where the demand still is

The first rung has not vanished everywhere. It has moved.

Healthcare is the standout. The U.S. Bureau of Labor Statistics projects healthcare and social assistance to add the most jobs of any sector through 2033, with home health and personal care aides alone projected to grow about 21% and add over 800,000 jobs, and nurse practitioner roles projected to grow more than 46%. Much of this work is hands-on and demographic, driven by an aging population, and hard to automate away.

Cybersecurity is another durable on-ramp. CyberSeek, the workforce-data project run with NIST, CompTIA, and Lightcast, reported over 514,000 cybersecurity and cyber-adjacent job listings in the U.S. in its June 2025 update, with a supply-demand ratio of 74%, meaning roughly a quarter of roles go unfilled. Many employers here now accept certifications and adjacent IT experience in place of a four-year degree.

Skilled trades round out the picture. They are physical, local, and resistant to being done by a model in a data center. If you want to see how a specific field scores on automation exposure, our job risk by role breakdown is a useful starting point.

A playbook to break in anyway

The strategy that worked for the last generation, apply broadly to entry-level postings and wait, is weakest exactly where AI bites hardest. Here is a more resilient approach.

  1. Aim at augmented roles, not automatable ones. Position yourself as someone who directs and checks AI output rather than someone who produces the raw draft the AI now generates. Fluency with the current AI tools is fast becoming a baseline expectation, not a bonus.
  2. Build proof of work you can point to. A portfolio, a shipped side project, a public analysis, or a documented internship beats a resume full of coursework, because it demonstrates the judgment employers are now screening for.
  3. Target the growing lanes deliberately. Healthcare, cybersecurity, and the skilled trades are hiring at the entry level today. A focused certification in one of them can open a door faster than another year of generalist applications.
  4. Stack durable, human skills. Communication, client-facing problem solving, and cross-functional coordination are the tasks least exposed to automation. Our guide to the skills to build maps these in detail.
  5. Use the side door. Apprenticeships, contract-to-hire roles, internal transfers, and smaller firms often hire juniors that large, heavily-automated employers no longer do.
  6. Know your own exposure. Before you commit to a field, understand how automatable its entry-level work is. You can gauge that directly with our AI displacement risk diagnosis.

None of this is a promise that the market is easy. It isn't. But the first rung is being rebuilt, not removed, and it now rewards people who can do the judgment-heavy, human, AI-augmented work that a model cannot finish on its own. Aim there, and you are climbing toward where the ladder is actually being extended.

Sources & further reading