The most mispriced asset in the American economy right now is not a stock, a bond, or a building. It is a 26-year-old professional with a good degree, three years of experience, native fluency in AI tools, and a job market that has decided it no longer needs her.
Consider what the market is telling us. Unemployment among recent college graduates has climbed to roughly 5.7 percent, well above the national rate of about 4.2 percent, reversing an advantage degree-holders had held for essentially the entire modern era of labor statistics. Underemployment among the same group is at its highest level since the pandemic. Entry-level job postings have fallen by roughly a third since early 2023. Research from the Stanford Digital Economy Lab found that workers aged 22 to 25 in the most AI-exposed occupations have seen a 16 percent relative decline in employment since generative AI spread through the workplace. And surveys suggest around 40 percent of chief executives plan to reduce junior roles further in the next year or two.
Our CEO has written about the “Great Repricing“: the thesis that AI is not primarily destroying jobs but collapsing the scarcity premium on cognitive labor. What the data above describes is the Repricing arriving exactly where economic logic said it would arrive first, at the entry level, where work is most codifiable and least dependent on accumulated judgment.
Most CEOs are reading this as a story about cost savings. A smaller number are reading it as a story about risk, worrying about where their future managers will come from. Almost nobody is reading it as what it actually is: an arbitrage opportunity, and a closing window.
The diamond has a supply problem
Every executive has now seen the chart of the “diamond-shaped” organization. The pyramid that defined corporate structure for a century, wide at the bottom with armies of juniors, is compressing at the base as AI absorbs routine analysis, drafting, basic coding, and administrative work. What remains bulges in the middle: experienced professionals who direct AI, exercise judgment, own client relationships, and catch the confident errors machines still make.
Here is the question the chart never answers: where does the middle of the diamond come from?
Mid-level judgment is not hired so much as it is manufactured. It is manufactured from raw talent, repetitions, feedback, and time. For decades the pyramid was the factory: juniors did the grunt work, absorbed the business by osmosis, made supervised mistakes, and emerged five to seven years later as the seasoned middle. That factory is now being shut down across the economy, one hiring freeze at a time.
The market is already signaling the consequence. Job postings are quietly migrating up the experience curve; in technology roles, for example, the share of postings requiring five or more years of experience has been rising while the share open to two-to-four-year professionals falls. Employers are bidding for finished mid-level talent because they have stopped producing it themselves. Follow that logic five years forward. The cohort that would have become 2031’s experienced middle is being turned away at the door in 2026. The diamond’s most valuable layer is heading into a structural shortage, and nearly every company is responding by making the shortage worse.
This is the setup for the arbitrage. When an entire market abandons an asset class for cyclical reasons while its long-term fundamentals strengthen, the buyers who remain acquire extraordinary value. The asset class in question is young professional talent. The remaining buyers get their pick.
What the repriced cohort actually offers
Let us be precise about what is on sale, because the framing matters. These are not damaged goods. The skills of this cohort have not declined; the market price of their entry-level output has declined, because AI now produces first-draft analysis and boilerplate code at near-zero marginal cost. Those are different things, and confusing them is how competitors will hand you their future leadership pipeline.
What has not been repriced downward, and arguably has appreciated, is everything else this cohort carries. They are the most AI-fluent generation in the workforce by a wide margin: recent London School of Economics research found roughly 83 percent of Gen Z workers using AI on the job, versus 73 percent of Millennials, 60 percent of Gen X, and about half of Boomers. Deloitte’s 2026 global survey found nearly three quarters of Gen Z and Millennial workers now using AI in daily work, up sharply in a single year, with 79 percent using AI to identify their own learning and development opportunities. They are not waiting for the training department; they are already treating AI as an always-on coach.
And this is the deeper point, the one that matters most for transformation. The organizations that win the next decade will not be the ones that bought the best AI tools. The answer is cheap; adoption is what costs you. Winning requires staff who are simultaneously experienced enough to know what good looks like in your business and flexible enough to help author a genuinely new division of labor between humans and machines. That combination barely exists in nature. Experience without flexibility gives you veterans who use AI as a slightly better search engine. Flexibility without experience gives you enthusiasts who cannot tell a brilliant AI output from a confident hallucination.
You cannot hire that combination off the street at any price, because the market has not had time to produce it. You can only manufacture it. And the fastest feedstock for manufacturing it is exactly the cohort the market just put on clearance: sharp, ambitious, AI-native professionals whose flexibility is innate and whose experience can now be accelerated.
The escalator · Training & development as the unlock of the arbitrage
Buying undervalued talent is only half the trade. The other half is compressing the seasoning curve, turning a five-to-seven-year march to mid-level judgment into a two-to-three-year climb. A decade ago that would have been wishful thinking. Today it is a product category.
Modern AI-powered learning platforms have quietly become something very different from the compliance-training portals most executives remember. The current generation can diagnose an individual’s skill gaps against a role profile, generate adaptive learning paths that adjust to demonstrated mastery rather than seat time, place learners in realistic simulations (client negotiations, incident response, financial reviews) where mistakes are free, provide AI coaching available at 11 p.m. the night before a big meeting, and give leadership a live map of capability across the organization instead of an annual guess. Paired with deliberate human mentorship and real P&L exposure, these systems attack precisely the bottleneck that made experience slow to build: the scarcity of repetitions and feedback.
Notice the recursion here, because it is elegant. The same technology that repriced this cohort’s entry-level output is the technology that can accelerate their development into the layer of the org chart where scarcity is growing. AI created the arbitrage on both sides of the trade.
The math, conservatively
Consider a realistic mid-market scenario. A finished mid-level professional hired laterally in 2028-2030, in a market where every competitor is bidding for the same shrinking pool, will plausibly cost $150,000 or more in base compensation, plus a recruiting fee of 25 to 30 percent, plus six months of ramp time before full productivity, plus the retention risk that comes with mercenary hires who joined for the offer and will leave for the next one.
Now price the alternative: a high-potential early-career professional hired today at roughly half that compensation, plus a serious development investment (platform licensing, structured mentorship, stretch assignments) that rarely exceeds a low five-figure annual cost per person, reaching mid-level capability in two to three years inside your business, fluent in your systems, your customers, and your culture. Even after fully loading the development costs and discounting for some attrition, the developed employee typically arrives at the same capability point at a total cost 30 to 50 percent below the lateral hire, with a crucial bonus: she got there your way, shaped by your standards, carrying your institutional knowledge rather than someone else’s habits.
That is the arbitrage in one sentence: the market is pricing this talent on what AI made cheap, while you can capture what AI cannot make cheap, and the spread between those two numbers is wide enough to fund the entire development program several times over.
Answers to a few predicted objections
“Why hire juniors at all when AI does their work?” Because you are not hiring them for their current output. You are hiring them as inventory for the layer of your org chart that is going scarce. The entry-level tasks AI absorbed were never the point of entry-level hiring; they were merely the tuition the employee paid while becoming valuable. AI eliminated the tuition. It did not eliminate the need for the graduate.
“If I develop them, they’ll leave.” Some will. But the retention data on this cohort is more favorable than the stereotype. Deloitte’s research finds these generations prioritizing skill-building, stability, and visible growth paths over rapid title inflation, with most preferring steady development to job-hopping, and strong loyalty effects attached to employers who invest in their capability while the broader market is rejecting them. There is a durable psychological asymmetry in being the company that bet on someone the market discarded. And the alternative to developing people who might leave is employing people who cannot grow, which is not a strategy.
“Isn’t this cohort AI-native but experience-starved?” Yes, and that is exactly the argument for structured development rather than osmosis. A junior who leans on AI without guardrails can skip the repetitions that build judgment; the same junior inside a deliberate program, where AI-assisted work is reviewed, challenged, and connected to real consequences, builds judgment faster than the pyramid ever did, because feedback loops that used to take quarters now take days.
“Maybe this is interest rates, not AI.” Partly, and honest analysis should say so; post-pandemic over-hiring corrections and the cost of capital explain a share of the entry-level freeze. But the direction of travel is not in dispute among the executives doing the hiring, and for the purposes of this argument it does not matter. Whatever mix of forces repriced this cohort, the reprice is real, the mid-level shortage it creates is arithmetic, and the arbitrage is available to whoever moves first.
Growth via growth
There is a reason we find this strategy compelling beyond the spreadsheet. Companies grow in two ways: by acquiring capability and by creating it. The past decade’s playbook overwhelmingly favored acquisition, of tools, of talent, of whole companies. The next decade’s constraint, the genuinely scarce input, is people who combine experience with the flexibility to redesign work itself. That input cannot be acquired at scale. It can only be grown.
Which means the org chart of 2031 is being written right now, in this year’s hiring decisions. The companies that treat the entry-level freeze as a savings line will spend 2029 and 2030 bidding desperately against each other for mid-level talent that does not exist in sufficient quantity, because nobody manufactured it. The companies that recognized the mispricing, hired the sharpest of the repriced cohort, and built the escalator will own the scarcest layer of the modern organization at a discount their competitors will never see again.
The market just put the future middle of your diamond on sale. It will not stay there.
