Partner Perspective: Success with AI Depends on Clean Data & Selecting the Right AI for the Job

As AI continues to take hold in healthcare, executives must ensure they are using the right AI tools for the right jobs, according to Drew Ivan, Chief Strategy Officer at Rhapsody. While generative AI (GenAI) garners attention for its potential in clinical documentation, other AI methods, such as machine learning (ML), remain critical for tasks requiring precision and consistency.



AI in healthcare is not a new phenomenon, Ivan noted, but recent advancements in computing power and data availability have fueled its acceleration. “The first AI algorithms predate fax machines,” he said. “The difference now is that we have the computational power and cloud storage to process massive datasets.” These developments have allowed AI to flourish in various healthcare applications, ranging from radiology to patient identity management.

One of the most widely embraced applications of GenAI is ambient listening technology, which generates clinical notes from physician-patient interactions. “This is a perfect example of tool meets job,” Ivan said. “Physicians love it because it reduces their documentation burden, and health systems benefit from improved efficiency.” However, he cautioned that not every AI problem should default to GenAI. “If there is a single correct output, machine learning is often a better fit,” he explained. “GenAI thrives in scenarios where variability is acceptable, such as summarization.”

AI in Radiology

Radiology has been an early adopter of AI, largely due to the structured nature of imaging data. AI can identify areas of interest for radiologists, prioritize high-risk cases, and improve diagnostic efficiency. “Radiology remains the most mature AI use case in healthcare,” Ivan said, noting that roughly 85% of FDA-approved AI algorithms pertain to imaging. “What AI doesn’t do is diagnose—it highlights areas for human review. That’s a critical distinction.”

He shared an anecdote about a radiologist using AI-assisted imaging. In a trial, the radiologist initially reviewed scans independently before AI analysis. “The AI flagged an area of concern the radiologist had overlooked,” Ivan said. “It turned out to be something significant, and the physician was relieved. That’s the kind of collaboration that makes AI so valuable in healthcare.”

Beyond imaging, the use of GenAI for clinical decision support raises concerns. Unlike imaging applications, which operate within defined parameters, GenAI introduces unpredictability. “The issue isn’t that AI makes mistakes,” Ivan said. “It’s that it can’t always explain why it reached a conclusion, which creates risk when applying it to patient care.”

This explainability issue has become a focal point in AI governance. AI-driven decisions—such as approving or denying medical claims—must be transparent and auditable. “In the European Union, regulations require AI-driven decisions to be explainable,” Ivan said. “The same expectation is growing in healthcare. Physicians and patients want to understand why an AI system made a particular recommendation.”

He also pointed out that expectations of AI often exceed reality. “We hold AI to a higher standard than humans,” he said. “If a human doctor misses a diagnosis, it’s considered an unfortunate but natural occurrence. If an AI system makes an error, it’s seen as a fundamental flaw.” This mindset affects adoption rates, particularly in patient-facing applications.

Competitive Edge

However, he predicted that AI-enabled care will soon become a competitive differentiator. “If a physician is using AI to gather and summarize information, they will have a broader view of a patient’s health history than one who isn’t,” Ivan said. “Just as no one would visit a doctor without an electronic health record system today, AI-enabled care will become the standard.”

While AI’s presence in healthcare is expanding, Ivan stressed that its success depends on data quality.