Entry-level job postings dropped 35% between 2022 and 2024. Not because the economy collapsed. Because AI got good enough to do the work those roles existed to do. AI killing entry-level jobs isn’t a prediction anymore. It’s already happened in enough sectors that new graduates are feeling it in real time.
If you just graduated, or you’re about to, this is the thing nobody at your commencement ceremony mentioned. The entry-level jobs that were supposed to be your on-ramp? A lot of them are gone. And the ones that remain are being competed for by more people, paying less, and demanding more.
This isn’t doom. It’s just the situation. And if you know what’s actually happening, you can do something about it.
How AI killing entry-level jobs actually plays out
The data is pretty clear. A 2024 analysis by Anthropic found that AI tools disproportionately affect early-career workers, the people who spend most of their time on the tasks AI does cheapest and fastest. Research writing, data formatting, basic code, first-draft copy, simple analysis. The work that used to take a junior hire a week, a senior person can now do in an afternoon with Claude.
The World Economic Forum’s Future of Jobs Report puts it plainly: employers expect 39% of core skills to change by 2030. That’s not a rounding error. That’s almost half of what companies currently hire for becoming either less relevant or replaceable.
Some CEOs are already saying it out loud. Duolingo cut contractors. IBM paused hiring for roles AI could handle. Klarna replaced 700 customer service workers with an AI agent and announced it publicly. These aren’t flukes. They’re signals about where the cost-cutting math leads when AI gets cheap enough.
Why the traditional advice is actively bad for you right now
“Get a degree, build your resume, apply for entry-level roles, work your way up.” This was reasonable advice in 2010. Right now it’s like telling someone to fax their CV.
The ladder still exists for some people in some industries. But the bottom rungs are being sawed off. Companies are increasingly using AI to skip the junior layer entirely. A team of five senior people with AI tools can now do what used to require fifteen, including the three junior hires who would’ve learned by osmosis.
The old model assumed companies needed bodies at every level. Cheaper bodies to do simpler things, more experienced bodies for the hard stuff. AI broke that assumption. A senior marketer who knows how to prompt Claude is now doing first drafts, research, and reporting herself. She doesn’t need an assistant. The assistant job is gone.
Waiting for the market to return to normal is a losing bet. This is normal now.
This came from a book.
Don't Replace Me
200+ pages. 24 chapters. The honest version of what AI means for your career, written by someone who actually builds this stuff.
Get the Book →What’s replacing entry-level: the gap job and the superjob
Here’s the part that doesn’t make it into the doom articles.
Every time AI gets better at something, a gap opens up between what it outputs and what humans actually need. Those gaps don’t disappear. They become jobs. Messy, ill-defined, weirdly compensated jobs at first. Then real ones.
AI can write a product description. It can’t tell you if the tone is right for the brand, if the claim would make the legal team nervous, or if the regional sales team will actually use it. Someone who understands AI output AND has enough domain sense to bridge it to human needs? That person is currently drowning in work and underpaid because there isn’t yet a job title for what they do.
That’s the gap job. It’s the space between AI output and human outcome.
The superjob is the evolved version. It’s what happens when someone builds enough expertise in a domain, learns AI fluency on top of it, and can now do work that used to require a whole team. A single person who can research, analyze, write, edit, project manage, and ship. These hybrid roles are already emerging, and they pay well because there aren’t many people who can do them yet.
Neither of these shows up on a job board. Not clearly, anyway. You have to build your way in.
What industries are being hit hardest right now
Not every sector is seeing the same damage. Some industries are cutting entry-level roles fast. Others are slower. Knowing the difference matters if you’re deciding where to aim.
The hardest-hit categories right now are the ones built on structured, repeatable, computer-based work:
| Industry | Entry-level roles most affected | Why |
|---|---|---|
| Finance | Junior analysts, data entry, reporting | AI handles the modeling and summarizing |
| Legal | Paralegal research, document review | LLMs are fast and cheap for first-pass reading |
| Marketing | Junior copywriters, social media assistants | AI drafts fast; one senior person can edit |
| Tech | QA testers, junior developers | AI catches bugs and writes boilerplate code |
| Customer service | First-line support, ticket routing | Chatbots handle volume; humans handle exceptions |
The common thread: anything that was mostly information processing on a computer. Those tasks didn’t require judgment, physical presence, or deep relationships. They required time and attention. AI has time and attention in unlimited quantities.
The less-affected sectors involve physical work, high-stakes human relationships, or decisions with real legal and ethical weight. Construction, skilled trades, social work, medicine, senior legal counsel. Not immune, but slower-moving targets.
If you’re choosing a field right now, that table is worth thinking about.
The playbook: build, don’t apply
Sending out 200 applications is not a strategy. It’s a slot machine. And right now, the slot machine is mostly broken.
What works instead is building visible evidence that you can do the thing. This is Rule #14 from Don’t Replace Me, the field guide for exactly this situation: ship something ugly. Not polished. Not perfect. Visible. A project you did with AI tools. A case study of a problem you solved. A newsletter you started. A spec work piece you built for a company that didn’t hire you.
The work doesn’t need to be impressive. It needs to prove you’re capable of doing instead of just studying. Hiring managers who are drowning in AI-polished resumes that all look identical respond to people who’ve actually built things.
For people who are worried about whether AI will take their career entirely, the honest breakdown of what’s actually at risk is worth reading before you panic.
What you should actually learn right now
AI fluency. Not coding. Not a $997 prompt engineering course. Actual working familiarity with the tools.
Pick one: Claude or ChatGPT. Use it every day for the next 30 days on real problems in your field. Not “what is photosynthesis” nonsense. Real problems. Draft this cover letter. Analyze this dataset. Summarize these 40 pages. Write a first draft of this report. The goal is to develop judgment about when AI helps and when it hallucinates, what to trust, and what to check.
That judgment is an actual skill that non-technical people can build without a computer science degree. Most people don’t have it yet. That’s the gap you can be in.
Here’s a rough stack for someone starting from zero:
| Skill | What it means | How to start |
|---|---|---|
| AI fluency | Using LLMs effectively for real work tasks | Use Claude/ChatGPT daily for 30 days |
| Domain knowledge | Deep understanding of one field | Double down on your actual area |
| Output judgment | Knowing when AI is wrong | Compare AI output to expert opinion repeatedly |
| Human bridge | Translating AI output for non-technical people | Practice explaining, editing, contextualizing |
| Visible portfolio | Proof you can build things | Ship one project per month, post it publicly |
The combination of these is what Dee calls the “combination moat” in the book: the intersection of what you know, what AI can do, and the human judgment that holds it together. AI can’t replicate the full stack, because it can’t tell when its own output is wrong in context.
The honest conversation about what’s actually hard here
Some roles are genuinely going away. Not all of them. Not most of them. But entry-level paralegal work, basic data entry, first-pass content writing, simple QA testing, junior financial analysis. If your plan was to get a foot in the door with one of those and work your way up, the door is narrower than it used to be.
That’s real, and it sucks, and anyone who tells you it doesn’t is either selling something or not paying attention. The actual data on AI job displacement isn’t as catastrophic as the headlines suggest overall, but for entry-level white-collar work, it’s genuinely significant.
Here’s the thing though. The people who adapt now, while the field is still unsettled, have a real advantage. The superjob category is growing fast. There are organizations desperate for people who can work with AI and think critically about its output. Most of them can’t find those people yet because most people are still pretending the old model works.
You don’t need to wait for the job market to figure itself out. You can start building toward the gap right now.
Five things you can do this week
Not eventually. This week.
Open a Claude or ChatGPT account and use it for a real work task today. Not a test. A real thing you need done.
Do a gap audit in your field. Find three things people in your industry complain about that AI should theoretically be able to help with. Write down why it doesn’t work perfectly. That gap is where you can position yourself.
Build one ugly thing. A simple case study. A one-page analysis of something you care about. A spec project for a company you’d like to work for. Post it somewhere public.
Stop sending generic applications. For every role you really want, send a personalized note and attach proof of relevant work. Ten targeted applications beat 200 generic ones.
Read the room on anxiety. If the fear is blocking you from doing anything, that’s worth addressing directly. There’s a reason AI anxiety is its own problem right now. You can’t build while you’re frozen.
The entry-level market is harder than it was three years ago. That’s true. It’s also not the end of your career before it starts. The people who treat this moment as a prompt to adapt, rather than a reason to panic, are going to look back on 2025 as the year they got ahead.
The ladder is being rebuilt. You just have to climb what’s there instead of waiting for the old version to reappear.
Frequently asked questions
Is AI really killing entry-level jobs, or is this just hype?
It’s real, not hype. Entry-level job postings in white-collar fields dropped significantly between 2022 and 2024, with some analyses showing declines of 35% or more. The WEF projects that 39% of core job skills will change by 2030, and early-career workers are disproportionately affected because their work overlaps most with what AI can currently do. It’s not every industry, but it’s enough sectors that it matters.
What should new grads do instead of applying for traditional entry-level jobs?
Build visible proof of work. One project that shows you can do something real is worth more than a polished resume right now. Combine that with genuine AI fluency, meaning actual daily use of Claude or ChatGPT for real tasks, and you’re ahead of most applicants. Target companies where your specific combination of domain knowledge and AI skills fills a gap.
Do I need to learn to code to survive the AI job market?
No. Coding helps, but it’s not required for most non-technical roles. What actually matters is AI fluency: knowing how to use the tools, how to judge their output, and how to bridge AI results to human needs. That’s a skill non-technical people can build in weeks, not years. See our breakdown of AI skills for non-technical people.
Which entry-level jobs are most at risk from AI?
Basic data entry, first-draft content writing, simple financial analysis, entry-level legal research, customer service routing, basic QA testing. Roles that consist mostly of structured, repeatable tasks on a computer are the most exposed. Roles that involve physical presence, complex human relationships, or highly contextual judgment are less at risk.
What are "superjobs" and how do I get one?
Superjobs are emerging hybrid roles where one person does the work of what used to require a team, using AI as a force multiplier. They combine domain expertise, AI fluency, and human judgment. You get there by building skills across multiple areas, shipping real work publicly, and positioning yourself as someone who bridges AI output and human needs. The guide to future-proofing your career has the full framework.
Is it too late to adapt if I'm already a year or two into the job search?
No. The window is closing slowly, not overnight. The people getting ahead right now are the ones who start building today rather than next year. A year of real AI fluency and a portfolio of actual work puts you ahead of most people who spent that same year sending applications. Start this week, not next month.
