With an estimated 140,000 health AI companies competing for health system attention, the challenge for executives is no longer whether to adopt AI but how to evaluate which platforms will deliver at scale.During a featured session at Becker’s 16th Annual Meeting in April, sponsored by GW RhythmX, three health system leaders joined the company’s founder and CEO for a live demonstration of precision care AI and a discussion of what separates durable AI deployment from the noise.The panelists were:

• Deepthi Bathina, founder and CEO of GW RhythmX (Get Well + RhythmX) in Palo Alto, Calif.

• Darren Shafer, DO, president of Presbyterian Healthcare Services in Albuquerque, N.M.

• Chris Akeroyd, chief information officer at Lee Health in Fort Myers, Fla.

• Eric Alper, MD, senior vice president, chief quality officer and chief clinical informatics officer at UMass Memorial Health in Worcester, Mass.

Below are four takeaways from their conversation.1. No time to wait

All three health system leaders described competitive and operational pressures that make waiting untenable. Presbyterian Healthcare Services faces acute access challenges in New Mexico, where patients travel hours to see specialists. Lee Health is managing rapid population growth in southwest Florida with a shrinking provider base. UMass Memorial Health is competing against larger systems to the east. In each case, the panelists framed AI adoption not as a technology experiment but as a strategic necessity.“I don’t have time to wait,” Akeroyd said. “The time is now to do these things and start creating that road and that thought process of where we’re going to be.” Dr. Shafer echoed the sentiment from a clinical standpoint: with over 200 primary care physicians and advanced practice clinicians already using the platform daily at Presbyterian, he said he no longer understands why any system would delay.

2. Clinical validation is key

Bathina framed the core evaluation question for health system leaders as one of depth: how much investment sits beneath the interface?

She described GW RhythmX’s precision care AI as drawing from 10 data sources, including structured and unstructured EHR data, payer formularies, thousands of clinical guidelines, social determinants of health (SDOH) and referral network data. A clinician panel also conducts ongoing validation before any recommendation reaches a provider.

The platform spans multiple large language models and is fully embedded in Epic and Cerner workflows.

For Akeroyd, that depth is the deciding factor. “The more information it brings, the more validation it has, the more explainability, the more trust it’s going to create,” he said.

By contrast, he warned that point solutions layering a thin UI over a public model will create integration debt that health systems may spend years managing.

3. Solving for cognitive burden

The session’s live demo illustrated two capabilities that drew strong reactions from the panelists. The first was a differential diagnosis agent. When given a real-time clinical trigger such as a rising creatinine level, the platform surfaced ranked diagnostic possibilities with supporting evidence from the patient’s longitudinal record, relevant lab trends and applicable clinical guidelines, all within seconds.Dr. Alper, who has spent two decades focused on patient safety, described diagnostic error as one of healthcare’s most persistent unsolved problems, particularly as advanced practice providers take on more of the clinical workload.“The ability to have AI that is able to help you with differential diagnosis and be that second opinion is really going to be incredibly helpful,” he said.The second was a specialist consult agent that replicated the clinical guidance a nephrologist or cardiologist would provide, validated by specialists on GW RhythmX’s clinical panel. This directly addressed the reality that in many markets, specialist appointments are months away.

4. Hyper-personalization

The panelists discussed a dimension of precision care that goes beyond the physician-facing workflow. It’s the ability to personalize patient education and engagement at the point of care.Bathina described sitting in a Presbyterian clinic and watching a physician use the platform to generate culturally specific, microwave-only recipes for a Latino patient who needed dietary guidance to support a new medication. This all happened in real-time at the point of care.Dr. Alper raised the use case of an intelligent after-visit summary agent that patients can query conversationally, asking what to prioritize, when to start a bowel prep, what a specific instruction means,  rather than navigating a 30-page printed document.For Akeroyd, personalization also connects directly to revenue integrity. Accurate documentation at the point of care reduces denials and ensures reimbursement reflects the complexity of care actually delivered.

The panelists closed with a shared framework for health system leaders evaluating AI platforms. Organizations need to identify their largest operational problems first, insist on EHR integration as a baseline requirement and hold vendors accountable for the depth of clinical validation beneath the interface.

The post What precision care AI looks like at enterprise scale — 4 takeaways appeared first on Becker's Hospital Review | Healthcare News & Analysis.