At HIMSS 2026, much of the buzz centered on what AI can do for clinicians — but one of the most compelling conversations focused on a group too often left out of that story: patients themselves. In episode 7 of Doceree Dialogue – Chapter HIMSS, Chethan Sarabu, MD, a practicing primary care pediatrician and leader of clinical innovation at Cornell Tech, joined host Ritesh Patel, Ex Chief Growth Officer at Doceree, to explain why 2026 may be the year AI finally starts working directly for patients — and why the rise of large language models is the force making it possible.
Sarabu's perspective is rooted in personal experience. He became a pediatrician after navigating his own chronic condition from childhood, an experience that left him asking a bigger question about the system itself.
"I became a pediatrician because of my own patient journey with a chronic condition from childhood. What that really motivated me to do was understand how we create a healthcare system that empowers children and patients in general to have more of a voice in their care," he said.
That question carried him through years in Silicon Valley, at Stanford, and at health AI startups, exploring technologies that could strengthen the patient-clinician relationship. His route was unusually multidisciplinary: he studied landscape architecture before medical school, a background that sharpened his interest in the links between the health of the environment and the health of people. Living in California through some of the state's worst wildfires, he saw firsthand the toll on the children in his care work that drew him deeper into the intersection of climate change and health through the lens of AI and helped bring him to HIMSS this year. Today, he leads clinical innovation at Cornell Tech, Cornell University's technology campus, while continuing to see patients roughly a fifth of his time.
For years, Sarabu notes, patients have asked for two things: meaningful access to their own health data, and tools genuinely built for how they navigate the system — not engagement features that pay lip service to the idea.
"Patients have long been demanding access to their data and tools that aren't just lip service of patient engagement, but that are meaningfully made for how patients, their caregivers, and family members navigate the healthcare system," he said.
What's changed, in his view, is that the technology has finally caught up to the need.
"We're finally at a time now where the technology has caught up to what's needed," he said.
What that looks like in practice is starting to take shape: a patient who can actually understand what their lab results mean, a caregiver who can navigate a care plan without waiting three days for a callback, a family member who can ask a real question and get a real answer — not a printout and a follow-up appointment.
Central to Sarabu's optimism is a simple observation: healthcare happens in language. A lab result is a number, your hemoglobin A1C, say, but the number is only half the story. The meaning comes from interpretation: Is it good? Is it bad? What does it mean for you?
"The way we think about our health, the way we communicate about it — between patients, patient to clinician, clinician to clinician — is in the realm of language," he said.
Until large language models arrived, he argues, there was no way to make sense of that language computationally, at scale. That, for him, is the real story behind the hype.
"That's why large language models are such a radical force for transformation," he said.
Much of healthcare's AI progress so far has been clinician-facing — ambient scribes and workflow tools that ease the burden on care teams. Sarabu sees 2026 as the year the focus expands to include the people those systems ultimately serve.
"This year, 2026, feels like that inflection point where everyone is getting serious about what AI means directly for patients," he said.
In many ways, it comes back to where he started — a kid navigating a chronic condition, wondering why the system wasn't built more for him. Decades later, he's still asking the same question. The difference now is that the technology may finally be ready to answer it.