Johns Hopkins Partners With OpenEvidence to Rethink AI's Role in Medical Training and Care

Imagine a medical student who can pull up the latest cancer trial data in seconds, but has never had to sit with a dying patient's family and explain what comes next. That tension, between speed and wisdom, is exactly what Johns Hopkins Medicine and the AI company OpenEvidence say they want to get right as they begin a new partnership announced this week.

The two organizations are joining forces to build what they call "human-first AI systems." The idea is straightforward even if the technology is not: combine the best available medical evidence, a patient's own history, and Johns Hopkins' clinical best practices into tools that help doctors think faster and more clearly, without replacing the judgment that makes medicine a human profession in the first place.

"On the clinical side, we're interested in exploring how we can bring evidence-based AI more directly into clinical workflows, helping clinicians access and apply the latest medical knowledge in the context of individual patient care," Sara Murray, M.D., Johns Hopkins Medicine's chief medical informatics and AI officer, told Fierce Healthcare in an email. She said the goal is to find high-impact use cases, build tools that fit naturally into how clinicians already work, and then rigorously test whether they actually improve decisions and outcomes.

That last part matters. Plenty of health tech rolls out with big promises and little follow-up measurement. Murray and her colleagues are signaling they want evidence, not just enthusiasm, before these tools become embedded in daily care. The partnership is still young. Murray acknowledged the specific capabilities haven't been locked in yet, but the ambition is clear: move past standalone search tools and toward AI that's woven into the actual delivery of care.

OpenEvidence is already a familiar name in exam rooms and nursing stations. The platform functions as a clinical decision support system and medical search engine, and as of September it counted 1.12 million licensed, verified U.S. clinicians as users, including physicians, nurses, nurse practitioners and physician assistants. It has struck similar partnerships with Mount Sinai Health System, Cedars-Sinai, Penn Medicine, NewYork-Presbyterian and its affiliated medical schools, AdventHealth, and most recently Northwestern Medicine, where its tools are being built directly into the electronic health record.

The company's growth has investors paying attention too. OpenEvidence recently closed a previously unreported $250 million funding round led by Byers Capital and Andreessen Horowitz, pushing its valuation to $15 billion, according to Forbes. That kind of capital flowing into a clinical AI company raises a fair question: who ultimately shapes how doctors are trained when a commercial vendor is at the table. OpenEvidence CEO Daniel Nadler did not disclose any commercial terms of the Johns Hopkins deal but confirmed it is, at minimum, a research and product collaboration.

Redesigning What It Means to Think Like a Doctor

Sanjay Desai, M.D., Johns Hopkins Medicine's Vice Dean for Education, frames this moment in almost historical terms. Johns Hopkins, after all, is the institution credited with inventing the modern model of bedside medical training back in the 1890s. Desai believes AI now demands a similar act of reinvention.

"We are all entering, at an unbelievably fast pace, this new world of education and clinical delivery in an AI era," Desai told Fierce Healthcare. He added that academic medical centers have a responsibility to study these tools carefully, because new technologies always carry consequences, intended or not. The task, as he sees it, is making sure the physician workforce stays competent even as machines take over more of the information-gathering grunt work.

That's where the concept of "cognitive offloading" comes in. Think of it like calculators in a math classroom. Once students stopped doing long division by hand, educators had to ask a hard question: which skills still need to be taught by rote, and which can be handed off to a tool? Desai wants Johns Hopkins to answer that question deliberately rather than by accident. He described sorting physician competencies into three buckets: things AI can safely take over, things where physicians will work alongside AI, and things that remain inherently human, like comforting a frightened patient or making a judgment call in an ambiguous situation. The risk, he warned, is graduating doctors who've never had to build certain skills because an algorithm always did it for them, only to find those skills missing the moment they're working without supervision.

Despite those warnings, Desai sounds genuinely optimistic about the upside. He thinks AI could finally break medical education out of its one-size-fits-all mold. Instead of every student following an identical track regardless of their experience gaps, AI could track what clinical cases a student has actually seen, flag missing experience, fill it with simulation, and let learners move at their own pace. "It's hard to find a group of people, I think, in the world that are more curious and enjoy learning as much as those that pursue medicine, and this brings the joy back," Desai said. He also sees potential for shifting training away from a strict time-based model, where years in a program are the main credential, toward one based on demonstrated competency.

Why It Matters

Travis Zack, M.D., Ph.D., OpenEvidence's chief medical officer, put the stakes plainly in a statement: the real question isn't whether AI will change what doctors do, it's who gets to decide how that change happens, and under what principles. That framing cuts to the heart of why this partnership matters beyond one hospital system. AI is already moving into exam rooms and classrooms whether institutions are ready or not. The choice facing medicine now is whether that transformation gets shaped thoughtfully, with patient safety and physician competence as the measuring stick, or whether it simply happens to the profession while everyone scrambles to keep up. For patients, the difference could mean the gap between doctors who use AI to listen more closely, and ones who've quietly lost the skills to listen at all.