GenAI Won’t Lower Medical Costs By Replacing Doctors And Nurses

A person in a lab coat stands with arms crossed.

getty

In earnings reports and internal communications, the CEOs of Amazon, Salesforce, IBM and Shopify (alongside dozens of other publicly traded companies) have linked generative AI to lower costs and reduced headcount.

The World Economic Forum now estimates that AI will displace 92 million jobs by 2030 while Goldman Sachs reports that AI is already erasing a net 16,000 U.S. jobs per month.

The logic behind replacing humans with technology is easy to understand. If GenAI allows a software engineer to write faster code, a lawyer to review more documents or a customer-service agent to resolve more cases per hour, companies can produce the same output with fewer people.

For centuries, technology has lowered costs by substituting mechanical power for human labor. Agriculture is the classic example. Tractors and other machines allowed the United States to produce more food with far fewer farmers. Over time, the share of the U.S. labor force working in farming fell from roughly 41% in 1900 to less than 2% today. Meanwhile, food has become far more affordable relative to income: Americans spent roughly 20% of disposable income on food in the 1960s, compared with less than 10% today.

That cost reduction strategy through the application of GenAI still makes sense in most industries. However, it is the wrong way to think about making American medicine more affordable.

Why Healthcare Is Different

In healthcare, pushing doctors and nurses to care for more patients in less time produces limited savings. The major costs in medicine arise when preventable and controllable illness progresses to a life-threatening problem.

When that happens, patients usually require surgery, hospitalization and expensive medications. Poorly controlled hypertension can lead to stroke. Unmanaged diabetes frequently leads to kidney failure and lower-limb amputation.

Preventing these dangerous medical problems will have a far great impact on healthcare affordability than reducing headcount. Just look at the math: National healthcare spending is currently $5.7 trillion annually and is projected to reach $9 trillion by 2034. Of that, physician and clinical services account for only one-fifth of total medical expenditures, according to KFF estimates. Direct wages for doctors and nurses make up an even smaller share.

Thus, even if GenAI reduced clinician labor costs by 10%, total medical spending would fall by only about 2% or 3%. And that assumes every dollar saved from reduced staffing would flow back to patients, employers or taxpayers.

In contrast, based on CDC data, effective control of chronic disease could reduce life-threatening complications like heart attacks, strokes and kidney failure, saving upward of $1 trillion annually (nearly 20% of total medical spending).

Staying healthy is what every American would want for themselves and their family, and it is what broad application of generative AI now makes possible.

Where The Biggest Savings In Healthcare Can Be Found

The opportunities for better health are not new to medicine. Better management of chronic disease, improved access to care at night and on weekends, and more consistent coordination of care have been recognized for decades.

What is different now is that patients can obtain all three through GenAI applications:

1. Chronic Disease

Long-term illnesses like hypertension, diabetes, heart failure and kidney disease affect an estimated 76% of American adults and drive a disproportionate share of medical spending. The CDC says chronic diseases account for most illness, disabilities and deaths in the United States.

Yet despite enormous spending, chronic disease remains poorly controlled. Only about 1 in 4 U.S. adults with high blood pressure has it under control. Millions of patients with diabetes, heart disease and kidney disease experience similar gaps between what medical science makes possible and what the healthcare system reliably delivers.

GenAI can help close the gap. It can explain diagnoses and treatments in language patients understand, increasing adherence to evidence-based medical recommendations. It could connect with bedside and wearable monitors to continuously track blood pressure in patients with hypertension, blood glucose in people with diabetes, and early signs of deterioration in those with chronic heart failure. Rather than waiting months until the next scheduled doctor’s visit, GenAI could alert patients when a chronic condition is not under control and allow clinicians to make necessary medication changes months earlier.

2. Access

The typical medical office is open only a fraction of the week, but medical problems arise regardless of time or day. At night or on the weekend, patients are often forced into an impossible choice: wait until morning and risk getting worse or go to the emergency department and face hours of waiting and far higher costs for care that may not have required a hospital visit at all.

GenAI could assess the patient’s problem, alert the individual when symptoms require immediate attention and explain when it is safe to wait until the doctor’s office opens. Clinicians worry that AI tools will make mistakes. But compared with a patient having to guess what to do or consult the internet for information, the risks of consulting GenAI would prove far lower. Recent studies demonstrate that large language models answer medical questions at a level comparable to (and sometimes better than) clinicians, although today they are best used with clinician oversight.

3. Care Coordination

Coordinating doctor visits, controlling chronic disease and managing medications have become growing challenges for our nation's seniors. More than 90% of adults 65+ live with at least one chronic condition, and nearly all take prescription medications. Furthermore, Medicare patients with multiple chronic conditions see an average of 14 different physicians annually. For this population, care coordination is not a minor issue. It is central to safety, quality and cost.

Many of these medications are lifesaving when first prescribed. But over time, some become unnecessary or risky, especially as patients age or their health changes. Continued use can contribute to serious complications, including falls and fractured hips. Despite these risks, prescriptions are often refilled automatically for years.

A second problem occurs when a different physician prescribes a new medication without knowing the patient’s full drug list. The result can be a dangerous “drug-drug interaction.”

Generative AI can provide oversight by identifying medications that no longer appear indicated based on national guidelines and warning clinicians when two drugs may interact. Reducing pharmaceutical errors would increase patient safety and lower medical costs.

What Stands In The Way

The barriers obstructing the use of GenAI in medical practice will not be technological. They will be human.

The first barrier is financial. As long as doctors and hospitals are paid primarily for the volume of work they do, GenAI tools designed to empower patients and provide medical expertise in the home will be seen as a threat. Under capitation — a single payment to a group of doctors to provide the totality of medical care for a population of patients — the incentives align.

The second is systemic. GenAI will not lower costs simply because patients download an app or doctors add another tool to the electronic health record. The model of medical care delivery will have to change in parallel. Medical care today remains too fragmented, too uncoordinated and too dependent on calendar-based office visits. To maximize value, clinicians will need to use generative AI to provide patients with expertise continuously, identify medical problems earlier and coordinate care across clinicians and care-delivery settings.

The third is cultural. American medicine still prizes intervention more than prevention. The specialist who opens a blocked artery is celebrated more than the primary-care team that prevents the heart attack in the first place. And any technology that performs work once reserved for physicians will be perceived by most clinicians as a threat.

If healthcare leaders rate the current medical care system as excellent, then they will fail to pursue medicine’s biggest opportunities and view replacing people with technology as the only path to lower costs. But once they recognize the hundreds of thousands of patients who die annually from complications of poorly controlled chronic diseases, the inability to obtain medical expertise at night and on weekends, and the cracks through which individuals fall every day, they will see massive opportunities to increase quality, improve access and lower costs.

Once they implement GenAI solutions, they will create a virtuous cycle. As avoidable demand falls, clinicians will gain more time for complex patients. Better care will produce fewer complications. Fewer complications will lower costs. And lower costs will make it possible to invest even more in prevention, access and coordination.

Replacing doctors and nurses is not the future of American medicine. Helping them keep people healthier is.