In a healthcare landscape flooded with claims data and AI hype, Dan Riskin is focused on one basic goal: accuracy. As founder and CEO of Verantos, he’s built a company around one idea—if we want to use real-world evidence to guide patient care, the data can’t just be big. It has to be right.
While much of the industry leans on billing codes and traditional natural language processing (NLP), he sees the flaws firsthand. “If you want the most accurate information, you don’t go to billing information,” he said. “You go to the information intended to convey clinical issues, which is doctor narratives describing care.”
But parsing those medical narratives can be quite tricky, especially for AI. One example Riskin gave is a note that reads “patient with MA.” Is that migraine with aura or Medicare Advantage? Most systems can’t tell. Riskin says Verantos applies full-record inference, which links symptoms, medications, and even imaging, to decode intent.
“We go beyond NLP to use AI-based inference where we’re taking into account the rest of the record,” he explained. That means connecting data across prescriptions, notes, and lab values, and then scoring it for traceability and completeness. Verantos routinely hits 80% accuracy, far above the industry’s 50% norm, according to Riskin.
Riskin’s obsession with rigor is personal. A clinical professor of surgery at Stanford University with extensive experience in surgery, critical care, palliative care, and clinical informatics, he’s been in the room when decisions had life-or-death consequences. “We want to be on the front page of the New York Times saying we are doing the most advanced research in the world to benefit patients the most,” he said.
As generative AI makes headlines, Riskin wants tech health systems to learn from every patient interaction without drawing the wrong conclusions. “The scariest thing would be to do it wrong,” Riskin stressed. “But if we do this right and create the right frameworks… I think the future is extremely bright in healthcare.”
