The average person visits the doctor just 2.4 times per year, yet health happens every day—can conversational AI in healthcare bridge that gap? At HLTH 2025, healthcare technology leaders Kristen Valdes, Founder and CEO of b.well, and Farid Vij, CEO and Founder of Citizen Health, tackled this question head-on in their session, “Inside the CMS Conversational AI Assistant Pledge: How Early Adopters Are Delivering.”
If you missed the session, or caught it live and want to revisit the key insights, we’ve broken down everything you need to know about how the CMS Health Tech Ecosystem is transforming patient care by putting personalized AI health assistants in every patient’s pocket. From eliminating portal barriers to creating “Waze for healthcare” through consented community data, here’s what these early adopters are building and why this time feels different from every other attempt at personal health records.
The CMS Pledge as a New Foundation
Building on three decades of interoperability work, the voluntary CMS Health Tech framework establishes a new technological floor for patient data access. Twenty-one companies have pledged to become CMS-aligned networks, committing to adopt FHIR (Fast Healthcare Interoperability Resources) standards that enable complete medical record data transfer, including physician care plans, clinical notes, and comprehensive payer information.
The framework eliminates portal dependency by prohibiting requirements for portal accounts or appointments to access health records. This matters because 75% of patients drop off when asked to log in. Instead, modern identity verification methods like Clear or ID.me provide secure, frictionless access.
The pledge also supports three priority use cases:
- Kill the Clipboard: QR code-based digital health cards containing insurance information and longitudinal health records that patients can share with any provider
- Diabetes, Obesity, and Prevention Management: Targeted data exchange for conditions that burden the healthcare system
- Conversational AI Health Assistants: Personal AI agents that translate codified medical records into accessible, contextual information in plain language
Building Trustworthy Conversational AI in Healthcare
With 15-30% of queries to foundational LLMs already health-related, patients are using AI whether the industry participates or not. The question is whether healthcare will provide trustworthy, medically-grounded tools or leave patients to navigate on their own. Yet there are ways the industry can ensure patients use AI safely while building the trust necessary for widespread adoption.
Every AI response must be grounded in actual medical records with clear data provenance, what we call “citation or silence.” The AI either cites exactly where information was pulled from or remains silent if it cannot verify. This transparency is critical because medical records contain errors. A patient might show as both BRCA-positive and BRCA-negative in their records. Rather than resolving contradictions, trustworthy AI presents them to patients, who can flag discrepancies. The system remembers these corrections, building accuracy over time. b.well extends this further by letting patients see and correct the “memories” being created about them.
Conversational AI in healthcare must also operate in co-pilot mode, never providing medical advice. This human-in-the-loop approach aligns with CMS governance principles, emphasizing human involvement in healthcare decisions. Patients are already bringing ChatGPT outputs to appointments. The goal is to design systems that providers want to engage with and present information in ways that build clinical confidence. When providers ask patients, “Where did you get that? Is there a provider version?”—That’s success.
Finally, with thousands of EHR systems in the U.S., each coding information differently, accessing data isn’t enough. A semantic interoperability layer significantly improves AI accuracy by ensuring consistent interpretation across systems, making trustworthy AI not just possible, but practical.
The Power of Consented Community Data: “Waze for Healthcare”
The most transformative aspect of conversational AI in healthcare may be connecting patients with similar conditions through what Citizen Health calls “Waze for healthcare.” Unlike social media groups or patient forums, this approach uses AI agents securely communicating with each other on behalf of their patients, grounded in verified medical record data rather than anecdotes.
For rare disease patients—who see an average of nine providers concurrently and visit doctors 20 times more than average, yet still experience 99% of care at home—finding others with similar experiences is life-changing. The question, “Why is there no one else like me?” could finally have an answer through agent-to-agent communication.
How Agent-to-Agent Communication Works:
- Your AI agent knows your complete medical history
- With explicit consent, it can query other agents on the platform
- Information returns de-identified and private
- Results are based on verified identity and validated medical record data
- Within seconds, patients learn what treatments worked for others, which doctors helped, what side effects occurred, and where centers of excellence exist
This creates a verified, privacy-preserving way to access the collective wisdom of patient communities, knowing that insights come from actual medical records, not just remembered experiences or anecdotes.
Patients can eventually connect directly with individuals who’ve had similar experiences, getting the richest experiential data to inform their next steps. For complex patients exhausted from retelling their stories at every new appointment, this technology offers both practical support and emotional relief.
Why This Time Is Different
Personal health record apps have emerged every few years over the past decade, but most have failed to gain traction. This moment is different for four key reasons:
Conversational AI Changes Everything
Natural language interaction makes health data finally accessible and usable for average consumers. People don’t want to log into portals and click through menus; they want to ask questions and get personalized answers.
Healthcare Is Vertical, Not Horizontal
There won’t be one solution for everyone. The CMS pledge recognizes this by inviting multiple companies to participate. Different solutions will serve different needs: rare disease patients, chronic disease management, preventive care, and specific populations. It’s about pulling data and the context you put around it.
The Technology Finally Exists
FHIR standards, modern identity verification, advanced LLMs, semantic interoperability, and mobile-first design have converged to make patient-centered data access technically feasible.
Unprecedented Urgency
The largest healthcare organizations have pledged to bring this technology to market in 2026, with a sense of urgency never before seen in healthcare interoperability.
What This Means Moving Forward
The convergence of data interoperability and conversational AI in healthcare represents a shift from focusing on healthcare (those 2.4 doctor visits) to supporting health in the 365 days when health happens outside clinical settings.
For patients, this means access without portal barriers, medical jargon translated to plain language, learning from others with similar conditions while maintaining privacy, and sharing comprehensive health information with new providers without retelling painful medical histories.
For providers, it means better-informed patients arriving with relevant questions, access to longitudinal records from all sources, reduced administrative burden, and AI assistants that support rather than replace clinical judgment.
For the healthcare system, this means better-informed decision-making, improved chronic disease management, elimination of redundant tests driven by information gaps, and a shift from paternalistic care models to true patient partnership.
The Invitation
The CMS-Aligned Network pledge is voluntary and open to all. With 21 companies already committed, the technology standards established (FHIR, modern identity, no portal requirements), and live demos available, the momentum is undeniable.
Patients are already using AI for health questions. The only question is whether the healthcare industry will provide trustworthy, medically-grounded tools or leave patients to navigate on their own.
The future of healthcare is about making health information accessible, understandable, and actionable for every person, every day. And that future is arriving faster than most people realize. Healthcare stakeholders can join the movement by applying to the b.well network and starting onboarding today through our developer portal.
