But ChatGPT said…
My dad is a physician. When "the internet" became a thing, he'd get patients who came in convinced their Google-based self-diagnosis was spot on and now wanted him to write the prescription. Annoying as this was, it's only going to get worse. Now, instead of vaguely matching symptoms to some WebMD or Mayo Clinic article (which aren't bad sources when used responsibly), patients arrive with a full chat history, a confident diagnosis and treatment plan from ChatGPT, and a sense that the doctor is there to give the final okay.
This doesn't create less work for the doctor; it creates more. It's not "oh great, now that you've got ChatGPT we can skip the diagnosis and jump straight to treatment." It's "sheesh, now I first have to explain why your diagnosis, to which you're now emotionally attached because you made it, is wrong, then do the correct one."
And we'll see this in countless other professions:
Asking a lawyer to "just quickly check over" the contract you had the AI draft
Asking a developer to polish that vibe-coded prototype you threw together in an afternoon, which is now leaking everyone's personal information
Handing your CTO a technology roadmap you cooked up with the AI and asking them to "just go do that"
Two reasons this is a bad idea:
It's rude. A doctor didn't go through all that training for nothing. They learned a thing or two about forming a sound diagnosis. Same for any professional and craftsperson. They know what they're doing, so let them do their thing. How would you feel, in your own area of expertise, if someone contradicted you with a "But Claude says..."? (Doesn't mean it's not okay to ask questions!)
It's counterproductive. Using AI outside your own area of expertise risks generating slop that the professional then has to spend extra time cleaning up.
Better to keep things separated. When generating outputs (code, documents, designs), stick to what you know: generate code if you know how to code, draft contracts if you know how to draft contracts. Everywhere else, be careful about what you ask of the AI. Use it to uncover gaps in your knowledge, ask better questions, and get a lay of the land. Not as a replacement for verifiable expert guidance.
