You are not the customer. You are the curriculum. Every time you use Al to write the email, fix the code, build the deck, the system watches. It learns how someone in your role thinks. It learns what good output looks like for your job. Then it sells that knowledge to whoever wants to do your job…

The idea of work acting as a curriculum for AI is a major focus for Tristan Harris and the Center for Humane Technology. They argue that as people use these tools to refine code or draft emails, they are participating in Reinforcement Learning from Human Feedback. This process turns everyday professional tasks into high-quality data points that teach the model how to replicate expert logic. While tech companies often market these features as productivity boosters, the underlying goal for many developers is to reach a level of automation where the software can perform the task independently.

This creates a tension between immediate efficiency and long-term job security. Research from firms like Anthropic shows that while AI can theoretically handle a huge portion of knowledge work, there is still a gap between what the math can do and what is happening in the actual economy. However, early trends suggest that entry-level roles in fields like accounting and programming are seeing shifts in hiring as companies experiment with these tools. Critics of this transition warn that focusing purely on automation might hollow out the talent pipeline, leaving businesses without a way to train the next generation of human leaders.