A 2026 survey by Resume.org found that 21 percent of companies have already frozen entry-level hiring specifically because of AI. Nearly half (47 percent) expect to eliminate entry-level roles entirely at their companies by 2027. This is not a tech industry story. Finance and information services have been shedding an average of 9,000 jobs per month since 2023, according to Fortune. Before the pandemic, those same industries were adding 44,000 per month. Entry-level finance positions have fallen by 24 percentage points, with one analysis noting that an analyst is now expected to do the work that a team of three did in 2020. In the legal sector, large firms are reducing paralegal headcount after deploying AI platforms to review contracts and summarize case law. In marketing, teams that once hired three junior writers now hire one senior strategist and use AI to produce the volume.
The conventional reading of these numbers is that AI is eating entry-level jobs, and young workers should be afraid. That reading is incomplete. The deeper story is that AI is making experienced professionals more valuable, not less, because it is producing work that looks polished and correct but is often subtly wrong, and only someone with real domain knowledge can tell the difference.
This piece is about why the combination of professional experience and AI fluency is the most valuable position in the workforce right now, why the traditional path to building that experience is disappearing, and what that means for both the experienced professional and the parent of a teenager who will graduate into this market.
The jagged frontier is the reason experience matters more
In September 2023, a research team spanning Harvard Business School, MIT Sloan, Wharton, and the BCG Henderson Institute published a landmark study on how AI affects knowledge worker performance. The paper, led by Fabrizio Dell'Acqua and co-authored by Ethan Mollick among others, introduced the concept of the "jagged technological frontier." It was formally published in the peer-reviewed journal Organization Science in March 2026.
The core finding: AI assistance improved performance on some tasks and worsened it on others, even within the same knowledge workflow and at a seemingly similar level of difficulty. The frontier is "jagged" because the boundary between where AI helps and where it hurts is unpredictable and uneven. The researchers found that experienced, incentivized knowledge professionals performed worse when given access to AI for tasks that fell outside the frontier.
That last sentence is worth reading again. Experienced professionals did worse with AI on certain tasks. Not because AI is bad, but because the output looked good enough to trust. The danger of a confident, well-structured, subtly wrong answer is that it passes every surface-level check. Only someone who has done the work before, who knows what a correct answer looks like from having produced hundreds of them manually, catches the error.
This is why experience has become more valuable, not less. AI did not eliminate the need for judgment. It raised the stakes of not having it.
What experience gives you that AI fluency alone cannot
There is a specific gap that no amount of prompt-writing skill or tool mastery can fill: the ability to recognize when an AI output is almost right.
A new graduate can prompt well. They can structure a request, specify an output format, and iterate on a response. Technically, they may even be faster with AI tools than someone ten years into their career. But when the model returns a project plan with the right structure but the wrong priorities, or a financial analysis that uses plausible yet incorrect assumptions, or a set of curriculum-aligned test questions that appear rigorous but miss a standard, the new graduate has no internal reference to flag these issues. They have never run the project, built the model, or taught the standard. They do not know what "right" looks like from experience, only from what the tool told them.
We see this in our own work. When we needed to generate hundreds of math practice questions aligned to state standards, we ran the same 69 questions across four AI models and scored each one against 18 standards. The best model cleared 62 out of 69. The worst cleared 48. That evaluation was only possible because the people doing it had years of experience in curriculum alignment and instructional design. A technically fluent operator without that background could have run the same test and not known which failures mattered and which were cosmetic.
The evaluation gap is the gap, not the prompting gap. AI fluency without domain experience produces fast, confident, unverified output. Domain experience without AI fluency leaves value on the table. The combination of both is what the next three years will reward.
Why is three years the window, and why does the advantage grow
AI capabilities are accelerating. The research organization METR estimated in January 2026 that the doubling time for AI task capability has shortened to roughly 4.3 months, up from about 7 months in the 2022 to 2024 period. Models are getting dramatically better at producing outputs that look correct to a non-expert.
That acceleration is precisely why experienced judgment becomes more valuable, not less. As AI outputs become more polished and more convincing, the cost of an undetected error rises. A subtly wrong strategic recommendation that reads as authoritative. A compliance document that uses the right legal structure but misapplies a regulation. A lesson plan that aligns with the wrong grade-level standard. These are the errors that experienced professionals catch, and new users miss, and as the outputs get smoother, the errors get harder to see.
The World Economic Forum's Future of Jobs Report 2025 projects that 39 percent of existing workforce skill sets will either transform or become outmoded between 2025 and 2030. The skills that transform are the ones that add AI fluency to existing expertise. The skills that become outmoded are those that AI can replicate without a judgment layer: data entry, basic code generation, routine drafting, and first-pass research. The WEF estimates that 7.5 million data-entry roles alone will disappear by 2027.
Senior roles are more insulated, not because they are harder to automate technically, but because they require the judgment that comes from having done the work. The experienced professional who adds AI fluency to that judgment is in the strongest position in the market. The experienced professional who avoids AI is leaving value on the table. The new graduate who has AI fluency but no domain experience is the most exposed.
The broken ladder
There is a structural problem underneath the hiring numbers that most people are not talking about, and it affects more than just the current crop of job seekers.
Entry-level work was never just about getting paid. It was the apprenticeship. A junior analyst learns what a good financial model looks like by building bad ones and being corrected. A first-year teacher learns classroom management by failing at it in real time with a mentor watching. A new consultant learns to spot a flawed recommendation by presenting one and having a partner take it apart. The grunt work was the training ground. It was where judgment was built.
AI is now handling much of that grunt work. Research reported by CNBC in 2026 found that hiring of workers aged 22 to 24 dropped 9 percent immediately after ChatGPT launched in industries including finance, insurance, and professional services. By mid-2025, those industries saw a 12 to 15 percent decline in employment, roughly 150,000 fewer early-career jobs. The decline was almost entirely due to fewer hires, not layoffs. Companies did not fire their juniors. They stopped replacing them.
Job postings on Handshake declined 15 percent in 2025, while applications per posting jumped 30 percent. Fortune reported that entry-level unemployment hit 9.7 percent, the worst in 37 years. The Brookings Institution found that AI could automate more than 50 percent of tasks in entry-level positions, five times the risk faced by more senior roles. And these numbers span industries: finance, law, marketing, consulting, tech, insurance, and professional services are all affected.
This means the ladder that built experience is being removed at the same time that experience is becoming more valuable. The generation entering the workforce between now and 2030 faces a real paradox: they need domain judgment to work effectively with AI, but the entry-level roles where that judgment was traditionally developed are being automated or eliminated.
This is not a generational criticism. It is a systemic problem that affects every organization that hires.
Why teens need professional experience earlier
If the bottom rung of the career ladder is disappearing, the answer is not to wait and hope it returns. The answer is to move the experience earlier.
Teens and young adults entering the professional workforce over the next three to five years will graduate into a market that expects both technical fluency and professional judgment from day one. The hiring data already reflects this: employers want hybrid skills, technical competence plus business acumen, and they are not willing to train for the basics the way they once did.
This is the work we do with our teen learners at 24/7 Teach. Our programs give teens access to technical and analytical professional experience years before they would normally encounter it. Real projects, real presentations, real feedback from mentors, and real portfolios. Not simulated work. Not classroom exercises dressed up as projects. Actual deliverables reviewed by working professionals.
The reason is precisely what this post argues. If experience is what gives you the judgment to work effectively alongside AI, and if the traditional path to gaining that experience is narrowing, then the teens who arrive at the workforce with two or three years of real project work already behind them have a structural advantage over peers who are starting from zero. They are not operating at an entry-level skill set and mindset. They have already built the judgment layer that entry-level work used to provide.
Our teen graduates have been accepted to Harvard, Columbia, NYU, Cornell, Spelman, Howard, Stanford, and others, collectively earning more than $5.7 million in scholarships. But the outcome that matters most for this argument is not the acceptance letter. It is the portfolio of real work and the professional instincts they built before they ever applied.
Note: This article was researched and written by Justice Jones with AI assistance, then reviewed and edited by our team. The studies cited belong to their original authors. The examples from our own work reflect our organizational practice.
"Isn't this just telling people not to worry?"
No. It is telling experienced professionals to move faster.
The advantage of experience plus AI fluency is real, but it is not automatic. An experienced professional who avoids AI tools is not protected by their experience. They are simply slower than someone who has both. The market will not wait for anyone to get comfortable. The window is now because the tools are accessible, the cost of learning is low, and the gap between "experienced professional with AI fluency" and "experienced professional without it" is widening every quarter.
The Harvard research found that even highly skilled knowledge workers could not reliably tell which of their tasks AI could handle and which it could not. That means experience alone is not enough. You also need the fluency to test, evaluate, and learn where your judgment is needed and where AI can carry the load. That testing is a skill, and it is learnable.
What to do with this right now
For the experienced professional: your experience is an asset, not a liability. But it is a depreciating asset if you are not adding AI fluency to it. Pick one recurring task in your workflow this week, something you know well enough to spot errors, and run it through an AI tool. Evaluate the output against what you know a good result looks like. That is the start of the judgment-plus-fluency combination the market is rewarding.
If you want a structured path rather than piecemeal experimentation, that is the work our AI Fluency for Professionals bootcamp is built for: 12 weeks of real AI workflows from your actual job, reviewed by mentors who have done the same.
For the parent of a teenager: the career ladder your child was going to climb is being rebuilt while they are still in school. The teens who enter the workforce with real professional experience, real projects, and real analytical judgment will be the ones who do not have to start from scratch. That is not an abstract advantage. It is a structural one. Take a look at what our teen programs build, and ask whether your teen is getting that exposure now or waiting for a first job that may not exist by the time they graduate.
For the hiring manager: the most valuable person on your team is probably the mid-career professional who already has the judgment and is willing to learn the tools. Invest in them. The research indicates they will deliver the highest return from AI adoption. The new hire who is fluent in AI but has no domain judgment will produce fast, confident, wrong work, and you will not catch it until it costs you.
Human-led, AI-facilitated. The experience is yours. The tools are here. The combination is what lasts.
About the author
Justice Jones is an instructional designer, AI strategist, and former K-12 principal. He co-founded 24/7 Teach to close the gap between what schools teach and what teens and professionals actually need to succeed, and he serves as CSO of Naomi-AI. 24/7 Teach has supported more than 50 organizations and placed 600 or more adults in new careers, and its teen graduates have collectively earned more than $5.7 million in scholarships.