You watched the Fox News poll drop this week: 56% of Americans say AI is hurting jobs more than helping them. The headline writes itself. The anxiety is real. And the framing is wrong.
Here's the cost of getting this wrong: while everyone argues about whether AI takes jobs, your expertise — the thing that took you 20 years to build — is being ingested, repackaged, and resold without your name on it or a dollar in your account. The job question is downstream. The ownership question is upstream. And upstream is where the money moves.
So the real question isn't "will AI replace me?" It's "when AI does the work I used to get paid for, who gets paid now?"
I'm Matt Cretzman. I build AI systems for experts who refuse to be raw material. Let me reframe what that 56% is actually telling us.
The Poll Measured Fear. It Didn't Measure Extraction.
Fox News asked Americans whether AI helps or hurts jobs. 56% said hurts. 26% said helps. The rest were unsure.
That's a fear survey. It's not a diagnostic.
Because "AI hurts jobs" is a vague shape. It collapses three very different things into one bucket: automation of tasks, displacement of roles, and extraction of expertise. The first two are what every news cycle argues about. The third is the one nobody names — and it's the one quietly reshaping who gets rich this decade.
Tasks get automated. Fine. That's been true since the spreadsheet.
Roles get displaced. Also true, also not new. Industries turn over.
But extraction is different. Extraction is when the expertise inside your head — the pattern recognition, the edge cases, the judgment calls you earned through scar tissue — gets pulled out of you, poured into a model, and sold back to the market as a commodity. You're not replaced by a machine. You're replaced by a machine trained on you.
That's not a jobs story. That's an ownership story. And the poll didn't ask about it because most Americans don't yet have the vocabulary to name what's happening to them.
Three Extraction Economies. Pick One.
In the book I'm writing, On Whose Terms, I argue there are exactly three economies an expert can operate in right now. You're already in one of them. The only question is whether you chose it on purpose.
Economy 1: LLMs extracted you without asking.
If you've ever published a book, a blog, a podcast, a training manual, a Substack, a YouTube video, or a public talk — your work is almost certainly inside the weights of at least one large language model. You weren't asked. You weren't paid. You weren't credited.
In the UK last year, 88% of creators surveyed wanted copyright protection against AI training. The government punted. The model providers kept training. Your work kept being used. That economy's terms are simple: they take, you absorb the loss, the market rewards the platform.
This is extraction on their terms.
Economy 2: You rent yourself out by the hour to train the replacement.
Mercor is the cleanest example. Experts — doctors, lawyers, engineers, PhDs — sign up to answer questions, write prompts, and grade model outputs at $50 to $200 an hour. The CEO has been explicit about the business: they pay experts hourly to dump knowledge into training sets for frontier model labs.
It looks like a side gig. It is a side gig. It's also the cleanest IP-for-stew trade I've ever seen. You get a W-9. They get a permanent asset — your judgment, your frameworks, your years — baked into a model that will serve a million customers after your contract ends. You were paid for the hour. They were paid for the decade.
This is extraction on their terms, with a consent form.
Economy 3: You own the extraction.
The third option is the one that barely exists in public conversation yet, which is why I've spent the last two years building infrastructure for it. You still extract your expertise — you have to, because static experts get passed — but you own the cards, the distribution, the pricing, and the revenue.
Your knowledge gets packaged into systems that work while you sleep. AI agents trained on your frameworks, not someone else's. Products that carry your name. Compounding revenue from a body of work you already did the hard part of building.
This is extraction on your terms.
That's what I built Skill Refinery to make possible. Not as a course platform. As a Knowledge Delivery System — infrastructure that takes what's in an expert's head and turns it into an owned, compounding asset instead of a leaked one.
Why the 56% Feels the Ground Shifting
Here's why I think the Fox poll landed at 56% and not 26%. People can't articulate extraction yet, but they can feel it.
A consultant notices her proposals get copied by ChatGPT in six seconds. A radiologist sees the imaging model hit 94% accuracy on a dataset partially labeled by radiologists just like him. A senior engineer watches Copilot ship the exact pattern he's been refining for a decade, uncredited.
None of those people lost their job this week. But all of them felt the shift. The work they used to be paid a premium for is now a commodity someone else owns the distribution of. The premium is evaporating. The check still comes. For now.
That gap between "I still have the job" and "I can feel the value leaving" is what 56% actually measures. It's anticipatory grief over a status shift most people can't name.
The Numbers That Actually Matter
Forget the jobs poll for a second. Here are the numbers I'd put in front of any expert this week.
The frontier AI labs have raised, cumulatively, north of $100 billion. A meaningful chunk of their asset base is training data sourced from experts who were not compensated or were compensated hourly one time.
A single published book, ingested into a frontier model, will be consulted more times in a month than the author will be read in a lifetime. The author's royalty on those consultations: zero.
A Mercor-style contractor earning $150 an hour for 500 hours walks away with $75,000. The model their work trained generates that in revenue before breakfast on a Tuesday, forever.
A expert who packages their own frameworks into an owned system — even modestly — can compound that into six and seven figures of recurring revenue over the same window, because they own both the IP and the distribution.
Those aren't hypothetical spreads. Those are the spreads being realized right now by the people who understood the game one cycle early.
What I Actually Believe About This
I believe expertise is a gift. Not in a soft sense. In a literal sense — a thing given to you to steward, refine, and pass on with integrity. You don't steward a gift by handing it to the highest bidder for an hourly rate and watching them sell copies of it forever.
You steward it by owning its form. By making sure the knowledge reaches the people it was meant for, under your name, with your standards attached. Compounding is not just a financial concept. It's what integrity looks like over time.
That's the silver thread under the whole project. The extraction economies are not just economically bad deals. They're stewardship failures. Experts trading a legacy for a stipend because nobody showed them the third door existed.
What To Do Before the Next Poll
I'll make this concrete.
First, audit which economy you're in. If your work is on the public internet and you haven't built a system around it, you're in Economy 1. If you're contracting to a data-labeling or prompt-engineering platform, you're in Economy 2. If neither — you're on the sidelines, which is its own kind of loss.
Second, inventory your knowledge assets. Books, talks, frameworks, checklists, decision trees, client case files (sanitized), teaching notes. Most experts I work with have 10x more owned IP than they realize. They just never treated it as IP.
Third, build the delivery system before you build the next piece of content. The asset isn't the content. The asset is the system that turns content into compounding revenue. One is a product. The other is a business.
Fourth, decide on whose terms you want the next decade to run. The extraction is happening either way. The only variable is the signature on the contract.
The Poll Fox Should Have Run
If I could put one question in front of the 1,000 Americans Fox surveyed, it wouldn't be about jobs. It would be this: Do you own the economic output of your expertise, or does someone else?
I suspect 56% would be generous. I suspect the real number is north of 90%. And I suspect most of them have never thought about it in those terms before, which is the whole problem.
That's the conversation I'm trying to move into the open. Not AI versus jobs. Ownership versus extraction. On whose terms.
I'm writing a book about exactly this — On Whose Terms: The New Expert Economy and the Fight for What You Know. If the thesis resonates, join the launch list and I'll send the early chapters as they land.
If you want to see what the third economy actually looks like in practice — the systems, the stack, the build — that's at mattcretzman.com.
The 56% is right to be uneasy. They're just uneasy about the wrong thing.
Keep Building,
— Matt