“AI could kill all humans.”
“AI could cause human extinction.”
“AI is moving too fast to control.”
Maybe.
But I’m a physician, and I spend my days worrying about a threat to human life that isn’t theoretical. It is already here: American health care.
We spend more than any other wealthy country on health care and still leave too many people sick, undiagnosed or treated too late. We exhaust doctors with administrative work, ask patients to navigate a system few understand and devote extraordinary resources to treating disease after it has progressed instead of keeping people healthy.
The United States spends $5.3 trillion a year on health care, more than $15,000 for every American, according to the Centers for Medicare & Medicaid Services. That consumes 18 percent of our economy.
And we are trying to meet growing demand with a workforce that is already stretched thin. Federal projections suggest that by 2038, the United States could be short nearly half a million physicians and nurses. We cannot solve that problem simply by asking already overburdened caregivers to do more.
What does all that buy us? Not nearly enough. Americans have shorter life expectancy and worse health outcomes than people in many similarly wealthy nations.
We spend a lot of time worrying about what AI might someday do to us. In health care, we should be just as concerned about what is happening to us right now.
That is what makes AI’s potential so important. Used well, it could help us address some of the very problems our health care system has struggled for decades to solve, from undiagnosed disease and preventable complications to care that arrives too late and a system that consumes too much of the time and attention of the people working inside it.
The greatest near-term threat from AI in medicine is not that machines become too capable. It is that we use extraordinary technology to preserve a system that already treats people like machines: doctors as data processors, nurses as routers, patients as records and health care workers as costs to be optimized.
The easy path is to use AI to make that machine run faster: to generate more documentation, process more claims, optimize more schedules and squeeze another percentage point of productivity from an exhausted workforce.
We should demand more.
AI can give physicians more capacity to listen, think and diagnose. It can give nurses more time with the person in front of them. It can identify patients getting sicker before they land in an emergency department, connect information scattered across a fragmented system and take repetitive administrative work away from people we desperately need doing something more important. And between office visits, it can help monitor, educate and support patients, especially those managing chronic disease, identifying problems early and helping solve them before they become crises.
We should bring the same urgency to ending the harm health care causes today that we bring to preventing the harm AI might cause tomorrow. The future threat is real. So is the clear and present danger of the status quo.
That is why framing this as humans versus machines misses the point.
The goal is not to remove humans from medicine. It is to remove the work that keeps humans from practicing medicine.
I have spent my career looking at this problem from almost every side of health care, as a practicing physician and in insurance, integrated health systems, retail health and health technology. The vantage points are different. The problem looks remarkably similar from each of them.
Health care does not suffer only from a shortage of talented people. It suffers from a system that consumes too much of the capacity of the people we have before they get the chance to help patients.
At Lumeris, we are interested in AI not as spectacle, but as service. The question is not simply what AI can do. It is what we want it to accomplish.
Does AI expand what caregivers can do, or simply replace what they do?
If AI allows a physician to care for more people without sacrificing judgment or human connection, that is progress. If it helps recognize disease sooner, reach someone who cannot get primary care or prevent an avoidable hospitalization, that is progress.
If it gives a doctor back an hour only so an organization can fill that hour with more administrative work, we have gained nothing. And if we use AI merely to subtract people from a spreadsheet, we will have missed the point entirely.
This is not an argument for turning medicine over to algorithms. Medicine requires judgment, empathy, trust and accountability. AI can get things wrong. When technology influences consequential clinical decisions, human beings must remain accountable.
But guardrails and ambition are not opposites.
AI presents health care with a choice. We can use it to make a broken system more efficient at being broken. Or we can use it to change what the system asks of human beings in the first place.
The public debate understandably worries that AI could someday make humans matter less. Health care should prove the opposite.
AI should make the human beings in medicine more important, not less.
If it does that, the technology we fear might kill us could help build a health care system that saves more of us.
Dr. David Carmouche is executive vice president and chief medical and commercial officer at Lumeris and co-creator of Tom, its AI-powered primary-care platform. He serves on the federal Healthcare Advisory Committee advising HHS and CMS. The views expressed are his own.
