I want to talk about something that doesn't show up in the AI productivity reports.

Not the hours saved. Not the diagnostic accuracy gains. Not the administrative burden reduced. All of that is real, and most of it is genuinely good. But there's a cost accumulating quietly on the other side of every AI interaction that nobody is measuring yet. And in medicine, unmeasured costs have a way of showing up where we least expect them.

I'm talking about the necessary struggle.

What We're Outsourcing

Sitting with uncertainty is uncomfortable. Asking a better question instead of accepting the first answer takes effort. Changing your mind when new evidence shows up requires a kind of intellectual honesty that has to be practiced to be maintained.

These are muscles. And muscles weaken when we stop using them.

The concern isn't theoretical. When every symptom gets answered before it's fully felt, every moment of disagreement gets resolved by an algorithm's opinion, every flash of curiosity gets satisfied before the question finishes forming — we are not saving time. We are simply skipping the part where critical thinking actually develops and we are letting a tool carry us over terrain that was supposed to build us.

Why Clinicians Are a Special Case

In medicine, I believe AI will make clinicians better. The clinical judgment was built over years of training, sharpened by cases that didn't resolve cleanly, refined by the moments where the data and the patient in front of you told different stories. For a trained mind, AI is an extension. It surfaces what might have been missed. It handles what doesn't require expertise so expertise can go where it's needed.

But that's only true because the foundation was already in place.

The clinician who arrives at AI with judgment intact uses it as a tool. The clinician who arrives without it — or whose critical thinking has slowly eroded through years of outsourcing — is in a different situation entirely. One that affects every patient they see.

This is part of what Dr. Javier Mendoza and I think about constantly at Mentor-IA: not just how to teach clinicians to use AI, but how to ensure that the introduction of these tools strengthens clinical judgment rather than quietly substituting for it.

The Broader Problem

What worries me more, though, is what's happening outside of clinical settings.

The average person now has access to AI tools that can answer medical questions, interpret symptoms, assess risk, and generate treatment considerations with remarkable fluency. That access is not inherently bad. Informed patients are better partners in their own care. Preparation leads to more meaningful conversations with physicians. Curiosity, when it leads someone to come to an appointment with better questions, is unambiguously good.

But there's a version of this that goes differently and the AI answer replaces the physician conversation instead of preparing for it. Where the symptom gets resolved by an algorithm before the patient has even fully registered what they're feeling. Where the question gets answered so quickly and so completely that the patient never develops the curiosity that would have led them somewhere more important.

That version is the one I'm paying attention to.

What the Right Relationship Looks Like

The healthiest relationship with AI is dialogue, not delegation.

Use it to come in better prepared; more informed, more specific, more ready for a real conversation with your physician. Let it sharpen the question you bring to the appointment. Let it help you understand what you're reading, contextualize what you're feeling, or identify what you don't yet know. That's the version of AI access that improves care.

What it shouldn't do is replace the part of you that knows how to wonder. The part that sits with a question long enough to understand it. The part that pushes back when the first answer doesn't quite fit.

That is the whole point.

The Question Nobody Else Thought to Ask

The future of healthcare will not belong to whoever has access to the most powerful AI. It will belong to whoever still knows how to ask the question the AI didn't think to generate.

That question comes from a trained mind, a curious patient, a clinician who hasn't stopped wondering. It comes from people who used the tool and kept thinking afterward.

Don't let the tool do the thinking for you. Let it make your thinking better.

That's the difference. And it matters more than any productivity metric will ever show.

——————————————————————————————————————

Dr. Rafael Grossmann is a trauma surgeon, digital health innovator, and global keynote speaker focused on the intersection of technology and human-centered medicine. He speaks on AI in healthcare, physician burnout, and the future of patient care.