Some of what you know is being commoditized right now. The most valuable part is not — and the wage data says it's appreciating. Here's the evidence, including the parts that should worry you.
If you've spent twenty years building real skill in a trade, a profession, or a craft, and you've watched an AI tool produce a fluent, confident answer about your field in eleven seconds, you've asked yourself some version of this question. I've asked it about my own work. So let me give you the answer first and the evidence after, because you deserve better than either blind reassurance or blind alarm.
Some of it, yes — the most valuable part, no. The codified layer of your expertise (textbook answers, standard procedures, anything written down) is being commoditized right now. The judgment layer (pattern recognition from years of real situations) is not — and wage data shows it's appreciating. The people actually being displaced are those whose whole job was the codified layer: mostly entry-level workers, not veterans.
Most of what you'll read on this question is vibes. So let's use measurements instead. When Anthropic analyzed millions of real Claude conversations in its first Economic Index, two findings stood out. First, usage leaned toward augmentation over automation — 57% of task usage was people iterating, learning, and validating with AI versus 43% handing tasks over entirely. Second, and more important: only about 4% of occupations showed AI use across three-quarters of their tasks, while roughly 36% showed AI use in at least a quarter of tasks. AI is being applied to slices of jobs, not whole jobs.
Now the honest update, because the picture is moving. By the September 2025 Economic Index report, full-delegation "directive" conversations had risen from 27% to 39% on the consumer side — and enterprise API usage was a striking 77% automation versus 12% augmentation, with 97% of task categories automation-dominant in API traffic compared to 47% among individual users. Read that carefully: businesses automate discrete tasks far more aggressively than individuals do. The automation share of your field's codified tasks is growing, not shrinking. Anyone telling you the augmentation-heavy 2024 numbers settled the question is selling you comfort, not analysis.
Be clear-eyed about this part, because pretending otherwise doesn't help anyone. AI is extremely good at retrieving, summarizing, and applying publicly available information: definitions, standard procedures, best practices, regulatory summaries, common troubleshooting. Economists have a name for this — codified knowledge — and the Brynjolfsson team's mechanism, as summarized by the Dallas Fed, is blunt about where it leads: AI automates codified "book learning" but not tacit knowledge from experience. For an entry-level employee, the codified tasks are the expert part of their job. For you, twenty years in, the same tasks are the inexpert part. Same technology, opposite effect.
If the value you were offering was mainly "I know the textbook answer faster than you can look it up," that value is genuinely collapsing. It was always the most exposed layer of every profession, because it's the layer that's written down and repeatable.
Underneath the textbook layer of every field is a second layer that never gets written down, because it only exists in people who lived it: judgment formed by pattern recognition across hundreds of real, messy, non-standard situations. The client whose case matches no template. The instinct that fires when something is subtly wrong before you can articulate why. In The Mentor Economy I put it this way, and I'll stand by every word:
Industries are made of people, relationships, vocabulary, and hard-won failures that AI was never trained on. The text gets absorbed. The judgment stays scarce.
That was a thesis when I wrote it. It is now a measured result. Dallas Fed economist Scott Davis tested exactly this split against post-2022 wage data, using each occupation's "experience premium" — how much more experienced workers earn than entry-level ones — as a proxy for how much of the job is tacit rather than codified. The result is the single most useful chart in this whole debate:
Per one standard deviation of AI exposure: occupations with a 0% experience premium — jobs where a veteran earns no more than a rookie, meaning the work is almost entirely codified — saw wage growth cut by 0.28 percentage points. At the median premium of 40%, the effect is essentially zero. And at the 90th percentile of experience premium, AI exposure is associated with wage growth about 0.2 points higher. The same technology that erodes pay in codified work is raising it where experience matters most. The median experience premium across 205 occupations is 40%, and it exceeds 100% for lawyers, underwriters, credit analysts, and marketing specialists. If your field pays heavily for experience, AI has so far been your complement, not your competitor.
Reassurance is worthless if it skips the casualties, so here they are. Since ChatGPT's release in late 2022, total U.S. employment rose about 2.5% — but employment fell 5% in computer systems design services and 1% across the most AI-exposed tenth of sectors, per the same Dallas Fed analysis. The displacement is not evenly spread; it is concentrated, and it is concentrated by seniority.
The Stanford Digital Economy Lab, working from high-frequency ADP payroll data, found workers aged 22-25 in the most AI-exposed occupations suffered a roughly 16% relative decline in employment since generative AI adoption spread — while more experienced workers in the very same occupations held steady or grew. Harvard's Hosseini and Lichtinger reached the same conclusion from a different direction: resume data covering 62 million workers across 285,000 firms shows that when a firm adopts generative AI, junior employment falls sharply relative to non-adopters while senior employment barely moves. They call it "seniority-biased technological change." Notably, the junior decline comes from slower hiring, not layoffs — the door is closing on the way in, not opening under people already inside.
There's a longer-run worry worth naming too, because it's the strongest argument against complacency for people like you and me. Enrique Ide's theoretical work on AI and knowledge transmission formalizes it: if AI lets senior workers operate without juniors, apprenticeship-style transmission of tacit knowledge breaks down, and seniority-biased change can "erode the future supply of the very expertise it complements." Translation: the judgment layer stays valuable, but the traditional pipeline for building it is thinning. Which means the people who hold it now — and find new channels to transmit it — hold an asset getting scarcer by the year.
Put the pieces together and the shape of the moment is unmistakable. As AI makes the textbook layer of every field free and instant, the judgment layer becomes the only layer left worth paying for — and every dataset above says the market is already pricing it that way. In a world flooded with fluent, confident, generic answers, a real person with real scars in a specific field becomes more valuable, not less, because the scarcity of genuine judgment is rising exactly as the supply of generic information goes infinite.
The fear most people carry into this question assumes AI and human expertise compete for the same territory. They don't. AI is expanding the territory of instantly available codified information — aggressively, as the enterprise automation numbers show. Your judgment occupies different territory, the kind I've argued makes twenty years of experience your most valuable asset rather than your expiration date. AI can generate an answer that sounds right. It cannot generate the scar tissue of having been wrong in a specific, memorable way and adjusting because of it.
If you're unsure which layer your own expertise lives in, two checks. First, the market's answer: does your field pay experienced people meaningfully more than beginners? If your occupation carries a strong experience premium, the wage data says AI is currently your amplifier. Second, the personal answer: has anyone ever come to you specifically because the generic answer wasn't good enough for their situation? If people seek you out when the standard case doesn't fit, you're operating in the judgment layer — the one AI keeps failing to replace in mentorship for the same structural reasons.
The risk was never that your knowledge becomes worthless. The risk is that you sit on it during the exact window when it's most valuable and least replicable. Every study above points the same direction: the codified layer of your job should be delegated to AI ruthlessly — that's the 43-going-on-77% that businesses are automating anyway — while the judgment layer should be packaged, taught, and scaled. That's not a consolation prize; per the Dallas Fed numbers, it's the appreciating asset. The practical playbook for that split is what I've called monetizing your expertise with AI — encoding your judgment into systems and offers instead of leaving it trapped in your calendar.
The codified-vs-judgment split in this dispatch is the spine of The Mentor Economy — the book walks through how to inventory which parts of your expertise are commoditizing, which parts are appreciating, and how to build a one-person business on the second. Free copy, you cover $9.95 shipping.
Your years in the field are not a liability in the AI era. They are the one asset the evidence says AI cannot manufacture and is quietly making scarcer. The only real question is whether you turn that asset into something others can learn from — or leave it exactly where it's always been: inside your head, helping only the people who happen to work directly with you.
Partly. The codified layer — the textbook answers, standard procedures, and lookup knowledge — is being commoditized fast. The judgment layer built from years of real situations is not. Dallas Fed wage data shows AI exposure hurts wage growth in occupations with no experience premium but is associated with higher wage growth in occupations where experience is most valuable.
The clearest measured displacement is at the entry level. Stanford researchers found a roughly 16% relative employment decline for workers aged 22-25 in the most AI-exposed occupations, and Harvard resume data across 62 million workers shows junior hiring falls at firms that adopt generative AI while senior employment stays flat. Experienced workers in the same occupations have held steady or grown.
Codified knowledge is anything written down — textbooks, procedures, best practices — which AI absorbs and reproduces. Tacit knowledge is compressed experience: pattern recognition across hundreds of real, messy situations that never gets documented. AI substitutes for the first and complements the second, which is why the same tool threatens juniors and amplifies seniors.
Ask whether people have sought you out specifically because the generic answer wasn't good enough for their situation. If clients or colleagues come to you when the standard case doesn't fit — and the median experience premium across U.S. occupations is 40%, exceeding 100% for fields like law and underwriting — you're holding the asset AI raises the value of.
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