Everyone is worried AI is making experience worthless. The employment data from millions of workers says the opposite is happening — and the people who understand why are the ones building something now.
TL;DR: Experience is not becoming obsolete in the AI era — it is becoming the scarce asset. Payroll data from millions of US workers shows AI-exposed employment falling 16% for 22-25 year olds while experienced workers in the same jobs held steady or grew (Stanford Digital Economy Lab). Firms that adopt generative AI cut junior hiring 7.7% while senior headcount keeps rising (Hosseini Maasoum & Lichtinger). AI floods the market with codified information; what it cannot generate is tacit judgment earned by doing the work — which is why the premium on verified lived experience is going up, not down. Your job is to package that judgment into a system before your industry figures this out.
There's a quiet fear moving through anyone who has spent decades getting good at something: if AI can generate an answer in three seconds, what was the point of the twenty years?
I've stopped arguing this one from intuition, because in the last year the labor economists did the work for us. The answer is now measurable, and it runs in exactly the opposite direction from the fear.
Start with the cleanest evidence we have. Researchers at the Stanford Digital Economy Lab — Brynjolfsson, Chandar, and Chen — analyzed ADP payroll records covering millions of US workers. Since generative AI adoption became widespread, early-career workers aged 22-25 in the most AI-exposed occupations experienced a 16% relative decline in employment. Experienced workers in the very same occupations? Stable, or still growing. The adjustment happened through employment, not wages — the market didn't discount experience, it discounted inexperience.
A second study makes the mechanism explicit. Hosseini Maasoum and Lichtinger (2025) tracked resume and job-posting data for roughly 62 million US workers at 285,000 firms from 2015 to 2025. Firms that adopted generative AI cut junior employment by 7.7% relative to non-adopters within six quarters — driven by slower hiring, not layoffs — while senior employment at those same firms kept rising. Their name for the phenomenon is the thesis of this dispatch: seniority-biased technological change.
And the workers who pair experience with AI fluency are capturing a premium that is widening fast. The PwC 2025 Global AI Jobs Barometer, built on close to a billion job ads, found jobs requiring AI skills carry an average 56% wage premium — up from 25% just a year earlier — across every industry analyzed, while productivity growth in the most AI-exposed industries nearly quadrupled. The market isn't paying for AI skills alone or experience alone. It's paying most for the combination.
The book puts the core distinction in one place, in chapter three:
A master carpenter understands wood in ways a machine never will. A seasoned real estate broker understands a market in ways an algorithm never will. A surgeon with twenty years of experience understands a patient in ways a diagnostic tool never will.
That knowledge — specific, earned, irreplaceable — has always been valuable.
Economists have a name for this: tacit knowledge — the practical, hard-to-codify skill acquired by doing the work, not reading about it. And it has a measurable price tag. In a randomized field experiment at a US sales call center, Sandvik, Saouma, Seegert, and Stanton (QJE 2020) found that agents at the 75th percentile brought in roughly 48% more revenue per call than those at the 25th — and the driver wasn't raw ability, it was differing knowledge of sales technique. When workers were paired with experienced high performers in short structured knowledge-sharing meetings, sales rose more than 15%, and the gains persisted for at least 20 weeks after the meetings ended. Transferring one experienced person's know-how produced large, durable economic value. That's not a metaphor for mentorship. That's mentorship, measured.
AI was trained on what people wrote about their experience. It was never trained on the experience itself, and the gap between the two is exactly where real expertise lives. AI has read about the fire. You've been in the room when it started.
Here is the part almost everyone gets backwards. The common fear is: "AI can now do what junior people used to do, so expertise as a whole is worth less." The actual dynamic is an inversion.
As AI collapses the cost of producing competent, generic, first-draft answers to near zero, the market floods with plausible-sounding mediocrity. Every industry fills with AI-generated advice that sounds informed but has never been tested against reality. In that environment, the scarce resource stops being information — it becomes verified judgment. The person who actually did the thing, twenty times, and can tell you exactly where it goes wrong, becomes rarer and more valuable precisely because everyone around them is drowning in noise. When anyone can generate content, the person with real scars becomes the trust anchor everyone else is searching for. (If you're wondering whether your specific field still counts, I worked through that question in Is my industry knowledge obsolete?)
There's even a warning bell ringing on the supply side. Theoretical work under review at the American Economic Review — Enrique Ide (2025) — models what happens when AI automates the entry-level tasks through which novices used to absorb tacit knowledge from experts. Short-run output rises; long-run growth and welfare fall, because the intergenerational transmission of expertise breaks. Translated out of economics: the pipeline that manufactures experienced people is being disrupted. The existing stock of deep experience — yours — is about to face less future competition, not more. Scarce assets with constrained future supply appreciate. That is what your twenty years now are.
A specific word for the reader who suspects this whole opportunity belongs to 28-year-olds. The founder data says you have it exactly backwards.
Azoulay, Jones, Kim, and Miranda, working with US Census Bureau administrative data on 2.7 million founders, found the mean founder age for the 1-in-1,000 fastest-growing new ventures is 45.0 — and prior experience in the specific industry predicts much greater rates of success. The result holds in high-tech sectors, in entrepreneurial hubs, and among successful exits. As Azoulay put it in MIT's coverage: "There is no such thing as a 25-year-old biotech entrepreneur." The edge, per the study, comes from accumulated knowledge, business connections, and learning from prior attempts.
Now notice what those three edges have in common: they are all distribution assets. At 50-plus you don't have to build an audience from zero — you have a network of former clients, colleagues, and peers who already know your work. You don't have to manufacture credibility — you have a track record that can be checked. You don't have to guess what your market's real problems are — you've been inside them for decades. The only thing you historically lacked was a way to reach more people than your calendar allowed. That's the one input AI actually supplies. The 28-year-old has energy and no proof. You have proof and, now, a machine that handles the energy-intensive part. In a market that has just started paying a rising premium for verified judgment, that trade favors you.
Which brings us to capacity — the second reason this moment specifically favors experienced professionals.
For most of the last few decades, deep expertise had a hard ceiling: every hour of real judgment had to be delivered one conversation at a time. That ceiling is what's actually breaking, not the value of the expertise itself. AI can now handle the repeatable 80% of delivering your expertise — writing it up, answering common first-pass questions, organizing it into a system someone can follow — while your limited hours go entirely into the 20% that genuinely requires your judgment. The experience isn't being replaced. It's finally being allowed to reach more than a handful of people at a time. (Why this multiplies rather than cannibalizes the mentor's role is the argument of AI won't replace mentors.)
So the practical move isn't to compete with AI at producing information faster — you'll lose that race, and it isn't worth winning. The move is to identify the specific, non-obvious judgment calls that only your particular scar tissue can make, and build a system where AI handles everything around those moments while you handle the moments themselves. Then price it like the scarce asset the data says it is — I laid out how in Pricing your expertise.
This dispatch expands one chapter of The Mentor Economy — the full book walks the gap diagnostic, the three-layer architecture, and the system-building playbook end to end. The book is free; you cover $9.95 shipping. Claim your copy →
Twenty years of hard-won judgment used to be limited by a calendar. Now it can be built into a system. The employment data, the wage data, the founder data, and the tacit-knowledge experiments all point the same direction: the twenty years didn't lose their value. They just found their moment.
More valuable, and the employment data now shows it. Stanford's Digital Economy Lab found that since generative AI adoption became widespread, workers aged 22-25 in the most AI-exposed occupations saw a 16% relative employment decline, while experienced workers in the same occupations held steady or kept growing. AI is replacing the codified knowledge juniors carry, not the tacit judgment seniors carry.
AI has access to more written information than any human, but information isn't the same as judgment. AI can tell you the textbook answer. It can't tell you which textbook answer is wrong for this specific client, in this specific situation, because it never carried the consequences of being wrong.
Outdated facts are easy to update. Pattern recognition — knowing what a problem actually looks like versus what it appears to look like — doesn't expire the same way. The specific tools you used twenty years ago may be gone. The judgment you built using them is still yours.
The data says the opposite. Census Bureau research on 2.7 million founders (Azoulay et al., American Economic Review: Insights) found the mean founder age of the top 0.1% fastest-growing new companies is 45, and prior experience in the specific industry strongly predicts success. Age is accumulated pattern recognition plus a network — both of which AI makes easier to distribute, not harder.
Consulting has always sold experience one hour at a time. What's new is the delivery mechanism. AI now lets one experienced person package their judgment into a system — content, frameworks, first-pass answers — that reaches far more people than a calendar of one-on-one consulting ever could, without diluting the value of the experience itself.
Author of The Mentor Economy and co-founder of MentorMe. He writes about turning hard-won expertise into AI-leveraged one-person businesses.
The Mentor Economy is the full system — free, you just cover $9.95 shipping.
Get Your Free Copy