Artificial intelligence has arrived in American medicine with considerable speed and considerably less consensus. Physicians are being asked to adopt new tools, trust new systems, and reassure patients who are increasingly encountering AI in their own care, often without fully understanding what they're interacting with. The conversations happening inside hospitals and clinics are more complicated than the press releases suggest.
Dr. Shikha Jain, a board-certified internist and oncologist, has been thinking carefully about where AI fits into clinical practice. Her view is neither dismissive nor uncritical. She sees specific, concrete ways the technology could make medicine better, and specific, concrete ways it could make things worse.
The case for optimism, in her telling, comes down to friction. Much of what undermines good care has nothing to do with physician skill or patient need. It is the documentation that piles up, the fragmented records, the follow-up that falls through the cracks. "If AI is built and deployed responsibly," she said, "it can help reduce some of that friction." She envisions tools that assist with pattern recognition and administrative work, potentially easing a burnout crisis that has pushed physicians out of practice at a rate the health system can ill afford. For patients, earlier risk detection and better access to information about their own diagnoses could matter enormously, particularly for those whose geography or circumstances have historically limited their options.
What she is not interested in is replacement. The goal, as she put it, is using technology "to make medicine more human, not less." That framing runs counter to how AI is sometimes marketed, where efficiency and automation are the headline benefits. Jain is after something different: a physician who has more time to be present, more capacity to listen, more ability to act on what they know about the person in front of them.
Her concerns, though, are serious. She is skeptical that AI is being built with equity as a design principle rather than an afterthought. "Technology is never neutral," she said. "If AI is trained on biased data, deployed without transparency, or used primarily to increase efficiency rather than improve care, it could reinforce the very disparities we are trying to eliminate." The patients most likely to benefit from better tools are often the same patients least likely to be represented in the data those tools were trained on. That gap is not theoretical.
She also raises a quieter concern about what happens when clinicians begin to lean too hard on algorithmic recommendations. Clinical judgment is built from training, experience, and an understanding of a particular patient's life and circumstances. An algorithm has none of that context. "Medicine requires context, nuance, and trust," she said. "Those cannot be outsourced to an algorithm." Her prescription for moving forward includes transparent development practices, governance structures with genuine accountability, and, critically, diverse voices involved in shaping these tools before they reach anyone's bedside. The technology may be moving quickly. The question is whether the institutions deploying it are prepared to move thoughtfully.