As far as the AI can see
Be in the room where it happens as healthcare AI experts examine what’s coming in 2027 and beyond for cardiology.
Artificial intelligence has reached a turning point in healthcare, with increasingly sophisticated tools capable of supporting clinical decision-making, improving efficiency and helping patients manage chronic conditions. But as AI technologies advance, the question is shifting from what they can do to how effectively — and safely — they can improve patient care.
“The models are now sophisticated enough to support meaningful clinical reasoning,” said Rashmee Shah, MD, a former health scientist at Meta and currently associate professor at the University of Utah in Salt Lake City. “The degree to which this technology will help people live longer, healthier lives, free of cardiovascular disease, depends on implementation, policy and regulation.”
Shah will co-moderate the Saturday afternoon session “What Comes Next for AI: Hear From the Builders of Tech’s Future.” The session features Daniel Nadler, PhD, CEO of OpenEvidence, and Shiv Rao, MD, CEO and co-founder of Abridge. The two technology leaders will explore current and emerging AI applications and their potential impact on cardiology in 2027 and beyond.
Although numerous companies tout novel AI tools, the tools that will make the cut and experience widespread adoption are those that solve real problems, said session co-moderator Pierre Elias, MD.
Healthcare AI applications primarily look to enhance diagnosis (find disease earlier, more accurately), navigate decision-making (predict options, risk, outcomes) or improve operational efficiency (automate prior authorizations, reimbursements).
Vetting these technologies is critical, said Elias, an assistant professor in cardiology and biomedical informatics at Columbia University in New York and CEO and co-founder of Pathway Labs. He subscribes to a checklist of inquiries that helps “separate substance from hype”:
- Identify the specific problem the tool claims it will solve. How will the outcome of this problem or process change once implemented — for better or worse?
- Ask how the product was developed and tested. Was it validated in a setting or workflow similar to yours, with or without patients? What is its scope of use and its privacy and security protections? Looking beyond the company’s polished demo, Elias said peer-reviewed external validation and real-world experience from colleagues define a product’s proficiency.
- Examine what could go wrong and the potential fallout. How widespread are the consequences, and who is responsible for monitoring or correcting errors?
- Start using it and measure the outcomes. Then, answer the ultimate query: Does it solve the problem in a way that benefits patients and the practice?
“The question is not simply, ‘How smart is the AI?’ It is, ‘Does it help deliver better care with an acceptable burden and risk?’” he said.
For cardiology in particular, Shah said AI’s biggest impact is likely to occur between visits, such as in monitoring patient progress and coordinating care.
“Much of cardiology is chronic disease management, including hypertension, heart failure, lipid management and anticoagulation,” she said. “Hopefully, AI will make it easier for the key stakeholder (the patient) to manage their health with less friction.”
Shah predicted the technologies that will scale earliest and be implemented most are those that ease the task load of clinicians or patients while keeping them in the loop, requiring final sign-off or both.
“AI-assisted front doors to healthcare will also see broad adoption. These include agents that help patients with scheduling, refills and prior authorizations — the administrative work that consumes significant time for clinicians and patients. Eventually, agent-to-agent workflows will be mainstream,” Shah said.
Meanwhile, fully autonomous applications face an immeasurable obstacle course in regulation and accountability, she said. The session panelists, Nadler and Rao, will address these challenges, along with other hurdles such as safety and scalability.
