Companies Are Letting Experienced People Go. AI Companies Are Buying Their Expertise to Train Better Models.

Across industries, companies are restructuring, automating, and reducing headcount. Among those leaving are experienced professionals who have spent 10, 20, sometimes 30 years accumulating domain knowledge.

At the same time, another market is growing rapidly.

The Expert Data Market

AI companies need expert data.

Not just generic data annotation.

They need doctors, engineers, lawyers, financial professionals, researchers, and other specialists to solve difficult problems, evaluate model outputs, design rubrics, explain reasoning, and provide the judgment that frontier models still struggle to reproduce.

AI labs pay data companies. Data companies build expert networks. Experts contribute their knowledge. And that knowledge helps train better models.

From a business perspective, this makes perfect sense.

In fact, when I first looked at this market, my immediate reaction was:

This is a great business.

There is valuable human expertise sitting across industries. AI companies are willing to pay for high-quality expert knowledge. A company that can identify, organize, validate, and transform that expertise into useful training and evaluation data sits in a very valuable position.

Knowledge Without the Knower

But then another question occurred to me.

What happens to the people whose expertise we are turning into model capability?

For most of modern corporate history, expertise and the expert were difficult to separate.

If a company wanted access to someone's 20 years of engineering judgment, it generally had to hire the engineer.

If it wanted an experienced lawyer's judgment, it hired the lawyer.

The knowledge traveled with the person.

AI begins to change that equation.

Human expertise can increasingly be captured, structured, evaluated, and transformed into something machines can learn from. That is an extraordinary technological achievement.

But it may also create a new economic question. What if expertise becomes increasingly valuable, while employing the person who owns that expertise becomes less valuable?

A Paradox Worth Naming

I don't think the answer is simply "AI will replace experts."

Reality is much more complicated.

Models still need human experts. New categories of expert work are emerging around training, evaluation, verification, and AI supervision. And every major technological shift has historically destroyed some tasks while creating others.

But there is a paradox here that I find difficult to ignore: We may be entering an economy that values human expertise enormously — while needing less human labor to deliver that expertise at scale.

An expert might be paid once to solve a problem, evaluate an answer, or demonstrate good judgment.

But once some of that knowledge contributes to model capability, that capability can potentially be reproduced across millions of interactions.

The economics are powerful.

The social implications are less clear.

Who Captures the Value?

Who captures the value created when human knowledge becomes machine capability?

The AI company?

The data company?

The enterprise deploying the model?

The expert?

And if experienced professionals help build systems that eventually perform more of the work they once did, should we rethink how they participate in the value being created?

I don't have a clean answer.

And I don't think stopping AI development is the answer either.

The Question We Should Be Asking

But perhaps the next phase of the AI conversation shouldn't only be about how much smarter models can become.

It should also be about: What happens to human capital when human expertise becomes training data?

That feels like a question worth asking before the technology gets much further ahead of the institutions around it.