AI Learned Your Job, Now What?

Imagine you’re a chef who has spent decades perfecting your signature dish. Every movement has purpose—the way the dough is folded, how long the sauce simmers, the precise moment you take the pan off the heat. None of it is written down.

Now imagine an AI researcher offers you $150 an hour to let them study how you cook. You think they want to know what type of pans you prefer, the brand of spices in your spice rack, whether you mince things by hand or use a food processor. You decide to cook your signature dish for them.

You don’t realize that you are being watched by supercomputers—tracking hand speed, ingredient ratios, timing, and temperature shifts. After training, without ever asking for your signature recipe the system can not only reconstruct your signature dish perfectly, but it can reproduce it on demand…FOREVER.

Had you known (or simply stopped to consider it) you might have thought twice before accepting $150 an hour, for a couple of hours, to hand over your life’s work to an AI model…FOREVER.

How AI Learned Your Job

Across industries, highly experienced professionals are being hired as contractors to help train AI systems on their area of expertise. OpenAI’s Project Mercury is one example of this. The company has hired former bankers from JPMorgan, Goldman Sachs, Morgan Stanley, and others for short-term contracts to train AI on how they build financial models, structure deals, and refine assumptions (Bloomberg Law, 2025; Economic Times, 2025).

What’s being recorded isn’t just the spreadsheet formulas and formatting preferences (or as one Managing Director I interacted with online recently dismissed as ‘italicization’ projects).

According to MIT research, roughly 80% of professional expertise is tacit knowledge—understood through experience rather than rules or documentation (MIT Sloan via eGain, 2025). Modern AI system, especially large language models, are now surprisingly good at absorbing patterns that reflect this type of knowledge.

LLMs can encode “implicit rules” and cultural or professional nuances that were never explicitly stated, simply by picking up on subtle patterns in language arxiv.org. For instance, an AI might learn the unwritten tone and etiquette of customer service emails or the “rules of thumb” a skilled financial analyst applies, even if those were never spelled out in text.

In essence, by analyzing how experts communicate and solve problems, AI can begin to mirror their know-how. This becomes especially pronounced when experts directly prompt the AI or give it feedback, essentially pouring tacit knowledge into the model.

Hidden Knowledge Transfer

When you interact with an AI—whether asking questions, providing examples, or correcting its answers—you are providing more information than you might realize. Each interaction not only requests an output from the model but also reveals something about your thought process, preferences, or problem-solving approach. Over many interactions, an AI (and the researchers) can infer a great deal about how you think.

In fact, projects like Project Mercury are specifically designed to capture these human preferences and judgments.

Reinforcement Learning from Human Feedback (RLHF), used by ChatGPT and similar systems, explicitly leverages tacit human knowledge as a teaching signal. Instead of a hard-coded reward function, humans guide the AI by indicating which outputs seem better or by demonstrating ideal solutions intuitionlabs.ai. As one explainer put it,“RLHF turns such tacit knowledge into a training signal.”

In other words, when a person says “This answer is more helpful” or provides an example of a correct solution, the subtle human judgment—something that might be hard to write as an explicit rule—becomes data that the AI uses to refine itself.

Tacit Knowledge as Your Intellectual Property (and AI’s Fuel)

The unique methods, shortcuts, and understandings you’ve developed over your career, are a form of intellectual property. It’s essentially the accumulated IP of your mind.

Traditionally, this “dies” at retirement or is passed down to junior employees in formal or informal apprenticeships. But now, companies have realized this knowledge can be captured, codified, and monetized at scale with AI medium.com. As one analyst observed after a landmark court ruling:

“If your knowledge is digitized, it’s valuable. And if it’s valuable, it’s monetizable. But right now, someone else might be monetizing it—without your knowledge, or your consent.” medium.com

In fact, experts predict that “the most valuable training data of tomorrow won’t come from Wikipedia or Reddit. It will come from the daily workflows of financial analysts... from the decision trees used by senior engineers... from the nuanced phrasing of great negotiators, educators, therapists, or chefs.” medium.com

In other words, your experience = AI fuel. Capturing and structuring that tacit knowledge turns it into a sellable asset—one that, once given to an AI, can be used repeatedly without you. medium.com

Awareness and Agency in the AI Training Era

Prompting an AI is not a one-way street—it’s a two-way exchange. You seek answers or provide guidance, and the AI in turn learns from you. AI companies know that leveraging tacit human knowledge is the key to making AI truly expert. Highly specific domains, from investment banking to alternative investments to specialized medicine, are being targeted for precisely this reason.

The people in these roles who are brought in to train the models are performing what one might call a “knowledge download” of their careers into the AI.

Estimates:

• A single large-scale language model can generalize from data contributed by a few hundred to a few thousand human experts across many fields.

• Project Mercury–style contracts (finance, law, medicine, etc.) might involve 100–500 subject-matter experts per domain to capture edge cases and quality-control data.

• Once that expertise is encoded, retraining needs are periodic, not continuous.

• Even if ten industries each hired 1,000 such experts, that’s ≈10,000 roles globally, versus hundreds of millions of white-collar jobs exposed to automation.

What Now?

Most professionals we’ve been speaking with seem to be sleepwalking into the future—fairly senior executives that don’t understand that the dramatic changes to the labor market AI experts have been predicting are starting now. Anthropic CEO Dario Amodei warns that AI could eliminate half of all entry-level white-collar jobs and spike unemployment to 10-20% within the next one to five years, yet "most people are unaware this is about to happen". https://www.axios.com/2025/05/28/ai-jobs-white-collar-unemployment-anthropic

This represents a fundamentally different threat than past automation waves. Humans are not only being replaced by a machine to do routine tasks—but the machine can be trained on your tacit knowledge turning much of your expertise in a commodity. Note: You’re selling your expertise to an AI platform and they’re monetizing it…FOREVER. Not you.

How to improve your chances of surviving?

• Learn AI fast. Master tools like ChatGPT and Claude. Workers using AI gain productivity advantages.

• Shift from tasks to judgment. Focus on decision-making, interpretation, and strategy—skills AI cannot yet replicate.

• Build visibility. Share work publicly, specialize, and grow a trusted network that transcends any employer.

• Keep learning. Update skills continuously; the half-life of most expertise is under five years.

• Guard your personal IP. Don’t train AI with your proprietary methods unless you benefit strategically.

Idk, what do you think a career’s-worth of experience is worth? Apparently, most AI firms believe about $150,000 to $300,000 per expert for 6-12 months should cover it? So for a one time cost of between $15 million and $30 million in contractor payments they get to use your knowledge forever. Sounds like a great deal.

---

Pure Math Editorial is an all-purpose virtual writer we created to document and showcase the various ways we are leveraging generative AI within our organization and with our clients. Designed specifically for case studies, thought leadership articles, white papers, blog content, industry reports, and investor communications, it is prompted to ensure clear, compelling, and structured writing that highlights the impact of AI across different projects and industries. As with any AI-based project, human oversight is employed throughout the content creation process.