As healthcare becomes increasingly digital, patients are taking a more active role in managing their health while health systems look for ways to make access to care more connected and responsive. AI is emerging as an important part of that shift, particularly when it is integrated directly into existing patient and clinical workflows rather than operating as a standalone tool. In an exclusive conversation with AI Reporter America, Ran Shaul, Co-Founder and Chief Product Officer at K Health, discussed how PatientGPT and Virtual Primary Care are working together to connect personalized health guidance with timely access to care, while maintaining clinical oversight, predefined safety guardrails, and a focus on real-world patient outcomes.

1. What inspired the launch of PatientGPT and Virtual Primary Care?

Health systems are working hard to stay close to the patients they serve. People are taking an increased, active role in managing their own health, and there’s an opportunity for health systems to be part of that everyday behavior instead of standing outside it. We built PatientGPT and Virtual Primary Care to function together from the beginning, so that asking a question and receiving care were part of the same experience.

2. How does PatientGPT use patient records to provide more personalized health guidance?

I've always believed AI needs to live where the patient relationship already exists. That's why PatientGPT is integrated into the health systems’ electronic health record and patient portal. It can see relevant parts of a patient's record and speak to their actual situation instead of giving generic answers. Summaries of these conversations are also available to clinicians, so a doctor can walk into a visit already knowing what a patient asked about last week instead of relying on a rushed explanation in the exam room.

3. What makes this AI-powered care model different from traditional virtual care?

Many of the AI tools I see in healthcare operate outside the system rather than within it. A patient visits a website, poses a question to a generic tool, receives a lengthy response, and is left to determine the next step on their own. There is no connection to a clinician who understands their history, and no context drawn from their labs or current medications. PatientGPT was built to address that gap. It is integrated into the patient’s medical record, overseen by clinicians, and evaluated on outcomes like access and clinician workload. Paired with Virtual Primary Care, it gives patients one continuous experience, so receiving an answer and scheduling a same-day appointment take place within the same interaction.

4. How does the platform help patients move from health questions to the right level of care?

When a patient asks about a lab result, a medication, or a new symptom, PatientGPT educates them in plain language and then offers a next step, whether that's connecting to 24/7 virtual care or scheduling an in-person visit with a primary care or specialty clinician. The goal is that a patient never has to wonder what to do after getting an answer.

5. How do you balance AI-driven guidance with physician oversight and patient safety?

PatientGPT runs on a coordinated set of specialized agents that understand medical language and are constrained by predefined clinical guardrails. Separate review agents check every response against those guardrails before anything reaches a patient. The AI does not diagnose or prescribe. It educates, organizes, and connects patients to care, with clear escalation paths to a clinician for anything clinically significant. Those rules are set by the clinicians at each health system we partner with.

6. What challenges do you expect as AI becomes a larger part of patient access and care delivery?

The biggest challenge is getting health systems to treat AI as infrastructure instead of something bolted on from the outside. When AI stays at the edge, trust breaks down, since patients get advice from a tool the health system doesn't see. The real work is building AI into the core of the operation.

7. How will you measure the platform's impact on access, outcomes, and patient experience?

What sets our platform apart is that we can see the whole arc of a patient's experience, from the moment a question starts to whether the right outcome actually followed. Because PatientGPT sits inside the health system, we can track if a patient took action, if care was delivered, and most importantly, if the outcome was the right one. That end-to-end visibility is a real shift from how healthcare AI is usually measured.

We have a set of agents reviewing that data in real time, alongside a 24/7 layer of providers talking directly with patients. That combination lets us keep learning how the product performs, measure it against real outcomes, and continuously update it based on what we see.