AI Doesn't Replace the Expert. It Multiplies Them.

Every few months, a new wave of AI coverage lands on the same thesis: AI is coming for your job. The specifics shift, this month it’s coders, last month it was analysts, but the frame stays the same. I think that framing is backwards, and it’s leading a lot of companies to make genuinely bad decisions.

Here’s what I’ve watched play out in practice: an experienced engineer with solid AI tooling doesn’t produce 10% more output than before. They produce something closer to 5x. The multiplication is real, but it flows through the experienced practitioner, not around them. The judgment call is still human: the architecture decision, the “this looks wrong,” the recognition that the AI hallucinated something plausible-sounding but incorrect. What AI removes is the friction between a good idea and a working implementation. That’s only valuable because the person holding the idea knows what a good idea looks like.

This is where the replacement framing does real damage. Companies trying to cut headcount by swapping experienced practitioners for AI agents are discovering they’ve removed exactly the judgment they needed to catch the agent’s mistakes. The agent is confident and fluent, and it’s wrong in ways a senior person would have caught and a non-technical manager won’t.

AI does change workforce composition, but not in the direction most coverage assumes. It’s not “fewer humans doing work.” It’s fewer humans doing the rote parts and more humans doing the parts that require judgment. One experienced practitioner with deep domain knowledge and the right AI tools outproduces a team of agents with no one steering them. You still need the expert. You just need fewer of them to do more.

The workforce question isn’t “who does AI replace?” It’s “which people will AI multiply?” Right now, the companies asking the second question are pulling ahead.