Blog article
*This article originally appeared on Healthbeat*
Dr. Dave A. Chokshi, keynote speaker at the New York Academy of Sciences’ AI in Healthcare conference, shares his thoughts on the potential of artificial intelligence in healthcare.
Published April 13, 2026
Dave A. Choksi, MD
Artificial intelligence will revolutionize our health, according to technology leaders. Dario Amodei, CEO of Anthropic, declared that AI “has the potential to deliver biological advances of the next 50 to 100 years in 5 to 10 years.” “With perhaps 10 gigawatts of computing power, AI could figure out how to treat cancer,” enthuses Sam Altman, CEO of OpenAI.
Back at the clinic, our enthusiasm becomes even more cautious. Part of this is due to the checkered history of medical technology, where humanism has too often been distilled into medicine in the pursuit of profit. Another hesitation is because in the real world, the problem is very often delivery, not discovery. Treatments that are not reaching the patients who need them most. Prevention that can never be put into practice. But what if AI could actually help change that?
Let me use one of my long-time patients as an example. I recently had a doctor’s appointment for his 45th birthday. We mourned the struggles of midlife (I’m also turning 45 this year) and celebrated the improvement in his mental health in recent months. And since colon cancer is affecting more and more young patients, I recommended a colonoscopy.
He consented to a stool test, which came back positive, but declined a further potentially life-saving colonoscopy. When I reached out to him to understand why, I learned that he had lost his job and health insurance and felt he had bigger fish to fry. “I’m doing well, Doctor,” he told me. His reassurance only reinforced my own inability to reassure him.
My patients are never alone. Despite the proven benefits of early detection and treatment, approximately half of Americans with a positive stool test do not complete follow-up within 6 months.
This yawning gap is repeated throughout American medicine. Less than half of people with high blood pressure receive effective treatment. Only about one-third of patients are eligible for the potential drug. cure Hepatitis C patients receive it. Fewer than one in five patients who would benefit from treatments such as buprenorphine for opioid addiction receive treatment.
Even as more effective treatments are discovered, we are unable to reach those who need them most.
AI can help patients navigate the complex journey to treatment
The discovery-delivery gap is not new. Thirty years ago, healthcare pioneer Don Barwick emphasized “the gap between what we know and what we actually do.” But we now have a real opportunity to address this failure by judiciously leveraging AI to save lives.
Let’s take a look at my patient’s colon cancer test results. Several steps are required after the stool test, including notifying the patient, adjusting the schedule, ensuring coverage, and explaining bowel preparation. In some cases, preparation is inadequate and the patient is sent back to square one.
The complexity of this process explains why many health systems employ patient navigators and why more affluent people seek concierge care to cut through the thickets. Artificial intelligence can help scale this type of personalized, ongoing engagement. Imagine a system that calls my patients in their native language, schedules colonoscopies, and answers questions about bowel preparation. Nurses and social workers can focus their valuable time on more complex cases, such as helping patients enroll in Medicaid.
The test for AI apps is whether they enrich the human relationships that are at the heart of great care. Too often, medical technology undermines human relationships, reducing eye contact, depersonalizing interactions, and turning drugs into transactions. Patients are less likely to trust recommendations. In this environment, trusted intermediaries like community health workers are becoming more important, not less, and technology needs to work for them.
Human relationships have always been key to bridging innovation and implementation, such as vaccines and treatments. The United States, for example, tamed tuberculosis through painstaking efforts to find and monitor infected people in communities and provide subsequent support to complete treatment plans.
AI can help bring treatments to patients who need them
Today, thanks to new treatments such as antiviral drugs, we have a chance to eradicate modern scourges like hepatitis C. Hepatitis C is an insidious disease. Each year, tens of thousands of Americans develop cirrhosis, a disease in which liver function deteriorates, eyes turn yellow and toxins cloud consciousness. However, a simple pill administration for 8 to 12 weeks can eliminate the virus from the body. AI could help ensure that such breakthroughs actually reach patients.
Identification of cases is especially important when people can transmit the infection without knowing they are infected. However, our tools for identifying these “silent carriers” are often blunt. Currently, the national recommendation for hepatitis C is for all Americans between the ages of 18 and 79 to have at least one blood test. In contrast, a machine learning algorithm tested in Israel was 100 times more efficient, yielding one diagnosis in 10 tests instead of one in 1,000.
Similar opportunities exist in HIV prevention. As powerful new tools emerge, like the long-acting injectable lenacapavir, AI could help ensure protection reaches those most at risk.
Eradicating diseases like HIV and hepatitis C is within our reach, as is curing smallpox and polio, but it requires more than just technology.
What AI cannot do is solve the all-too-human problem that some of the most important breakthrough treatments, from antiviral drugs for hepatitis C to $28,000-a-year lenacapavir to new GLP-1 treatments for diabetes and obesity, are not widely accessible due to high prices. To do this, we need to overcome the chaotic human system and forge a better path.
Mere mortals have done it before. In Louisiana, the so-called “Netflix model” facilitated expanded access to hepatitis C drugs for a fixed subscription fee. The Congressional Budget Office also estimates that a national enrollment model for hepatitis C could save the federal government about $6.6 billion over 10 years, in part by avoiding Medicare-funded hospitalization and intensive care costs. We can save lives and money.
My patient was finally able to complete a colonoscopy because her social worker helped her find medical insurance she could afford. He and I both breathed a sigh of relief when we found a small precancerous polyp. But he plans to have another colonoscopy in five years. Let’s hope medical science has advanced by a century by then. At the very least, we can make it easier for patients like him to receive treatments that are already known to save lives.
Dr. Dave A. Chokshi will be the keynote speaker at this year’s AI in Healthcare conference, May 12-13 at the Icahn School of Medicine at Mount Sinai.. Learn more and book your place at this influential conference.
About the author
Dr. Dave A. Choksi is a physician at Bellevue Hospital and the Sternberg Family Professor at the City University of New York. He previously served as New York City’s health commissioner.
