Conversational AI and Population Health: Transforming Patient Engagement

Conversational AI agents have the potential to transform routine patient portal support calls into clinical engagement opportunities. In this episode of our Partner Perspective Interview Series, Bill Hudson, CIO at Hippocratic AI, and Dan Stoke, VP of Healthcare at CTG, discuss their partnership integrating Hippocratic AI agents into CTG’s service desk operations. CTG handles 500,000+ annual calls on behalf of its health system clients. The technology enables population health outreach at an unprecedented scale while maintaining human support options. Implementations typically require 6-8 weeks, with use cases ranging from password resets to proactive care gap closure and preventive health screen reminders.

Health systems are deploying AI agents to handle routine patient portal calls while simultaneously addressing clinical care gaps, creating opportunities to engage patients at an unprecedented scale during what were previously transactional interactions.

CTG, an IT consulting company with more than 35 years in the healthcare industry, provides service desk support to health systems and has partnered with Hippocratic AI to integrate conversational AI healthcare agents into patient portal support operations. The collaboration targets the more than 500,000 annual calls that CTG handles for password resets and navigation—interactions that represent over 50% of all patient portal inquiries.

Podcast: Play in new window | Download (Duration: 50:20 — 37.8MB)

Subscribe: Apple Podcasts | Spotify

The partnership aims to transform non-clinical interactions into moments for meaningful patient engagement. When patients call to reset their MyChart password, they typically have an underlying clinical need—checking test results, scheduling appointments, or reviewing care instructions.

“Nobody calls to get the MyChart password reset because they’re just wanting to reset their password for fun,” Bill Hudson, CIO at Hippocratic AI, said. “They’re calling in because they want to make an appointment with their physician. They care about test results.”

Conversational AI agents can complete the administrative task while also identifying opportunities to address any clinical care gaps. After resetting a password, the agent might inform the patient about outstanding lab results or overdue preventive screenings, like mammography or colorectal cancer, using information already available in the health system’s records.

This approach allows organizations to maintain human representatives for patients who prefer traditional support while automating routine requests. Dan Stoke, VP of Healthcare at CTG, emphasized the importance of patient choice in the deployment strategy, noting that some patients will continue selecting human assistance while others engage comfortably with AI agents.

Addressing Population Health at Scale

Hudson, who previously served as CIO at John Muir Health and Integris Health, views the technology as a solution to a challenge that has occupied much of his recent career: managing population health effectively. Care managers typically oversee between 400 and 600 patients, limiting their ability to reach entire at-risk populations.

Conversational AI agents enable outreach to significantly larger patient cohorts. A care manager can use EHR registries to identify patients with multiple comorbidities and care gaps, then leverage AI agents to support them in contacting them at scale to preserve clinician capacity for higher-acuity care.

“I have access to scale and abundance, this infinite scale that I haven’t had before to be able to have a thorough conversation with you to get to understand what your specific healthcare needs are,” said Hudson.

The agents can reference prior, consented interactions documented in the health system record, creating continuity across calls. A patient who mentioned walking their dog in one conversation might hear the agent reference that detail in future calls, connecting health behaviors to personal motivations.

Clinical Safety Without Diagnosis

Hippocratic AI’s generative AI healthcare agents handle clinical workflows without providing medical diagnoses or prescribing. The AI agents are able to support across the patient journey, such as scheduling appointments, following up on blood pressure readings, explaining MRI preparation requirements, checking whether patients need claustrophobia management for imaging procedures, and more.

“We’re taking those entry-level types of tasks that you perform with the front desk office or with the medical assistants,” Hudson said. Those routine touchpoints can be turned into moments where patients feel supported, informed, and nudged toward better health. “That can be everything from following up on your results to making sure that you’ve checked your blood pressure or you’ve taken your blood glucose levels.”

The agents include safety escalation protocols. If a conversation reveals a patient experiencing concerning symptoms, the system transfers the call to a licensed clinician. Hippocratic AI has built a proprietary Polaris safety constellation architecture that has been tested by their network of over 7,000 U.S.-licensed clinicians. Safety thresholds are established in collaboration with health system partners and align with their clinical guidelines.

Hudson distinguished between providing diagnosis and understanding clinical context. An agent scheduling an MRI needs to know the proper sequence of tests and whether preliminary procedures are required, but this clinical knowledge differs from diagnosing medical conditions.

Implementation and Integration Strategy

Hippocratic AI works with health systems to identify specific use cases before building integration strategies. The company can pull data from Epic, Cerner, Meditech, custom systems, and Salesforce, sometimes drawing from multiple sources for a single deployment.

The integration approach focuses on speed to value. Teams copy existing interfaces, apply appropriate filters, and send data to Hippocratic AI, which structures and normalizes the information for use in conversational interactions. The data then flows back into the EHR in formats clinicians can use.

“What I used to see on my side as a CIO, was that work on a project like this would take three or four months,” Hudson said. “We’re standing use cases up in about six to eight weeks because we’re using that deep experience we have around integrations and that data normalization.”

CTG evaluated multiple AI vendors before selecting Hippocratic AI. Stoke noted that while many AI companies focus on clinician-facing tools or back-office operations, few prioritize direct patient engagement. Hippocratic AI’s emphasis on addressing clinical workforce shortages while working backward from patient needs distinguished the company in their assessment.

ROI Considerations and Measurement

Return on investment varies by use case. Some applications, such as reducing hospital readmissions or improving medication adherence in Medicare Advantage populations, deliver clearer financial returns. Others provide strategic value with longer timeframes.

Population health management represents a strategic investment. Most health systems recognize the need to improve risk management capabilities as payment models shift toward value-based care, but current care management teams cannot reach all at-risk patients. AI agents enable outreach to ten times as many patients at the same cost as traditional methods.

“Every single CFO, every single CEO in the provider healthcare space knows that we have to get good at risk,” Hudson said. “[The key is] being able to manage these risk patients at scale because you have access to an affordable model to reach them, an abundant model where I can reach out to all of my patients, not just 5% of those patients I’m taking risk for. The return on that is a long tail, but the return is huge.”

Implementation Best Practices

Successful implementations require clinical representation from the outset. Organizations should identify desired outcomes—whether reducing costs, improving patient engagement, or decreasing appointment no-shows—and work backward to determine appropriate use cases.

Security teams must be involved early, ideally as part of existing AI governance frameworks. The implementation should align with the organization’s strategic mission and account for the existing technology ecosystem.

“You can’t come to the party without understanding the client’s needs,” Stoke said, reflecting on health system sensitivities. “You’ve got to come with something to solve an existing problem, and don’t show up if you can’t consider or appreciate the ever growing technology suite they are working with.”

AI agents complement rather than replace existing tools. MyChart and patient portals excel at some interactions, but some clinical needs require conversational engagement. Scheduling an MRI, for example, may require discussing claustrophobia, metal implants, or coordination with physicians—conversations that text-based systems often handle poorly.

Organizations should start with high-volume, well-defined use cases before expanding to more complex applications. Password resets and navigation provide an entry point, but the technology’s value emerges when organizations move toward care gap closure, appointment scheduling, and preventive care outreach.

The CIO’s Role in AI Adoption

Hudson said CIOs should help operational leaders understand available tools and their capabilities while ensuring proposed solutions integrate into clinical workflows.

Population health teams often recognize they cannot reach all patients under their care with existing resources. CIOs can demonstrate how Epic registries combined with AI agents enable outreach at greater scale, allowing care managers to focus on the most complex patients while AI handles broader population engagement.

The CIO must also establish appropriate boundaries. If a use case involves diagnosis or requires licensed clinical judgment, the technology should not be deployed until it demonstrably meets the necessary safety standards.

Hudson emphasized the partnership nature of successful implementations, describing the CIO and operational leaders as running a race together. The CIO helps the organization understand both the possibilities and limitations of technology, but implementation succeeds only through collaboration.

Take it Away

• Deploy conversational AI agents for high-volume, routine patient portal interactions like password resets to create opportunities for clinical engagement during administrative calls
• Ensure conversational AI agents include safety escalation protocols that transfer calls to licensed caregivers when conversations reveal potential medical emergencies or conditions requiring clinical judgment
• Include clinical, security, and IT stakeholders from project inception, aligning AI deployments with existing governance frameworks and strategic organizational goals
• Position conversational AI agents as complementary tools to existing platforms like patient portals, deploying them for interactions that benefit from conversational engagement
• Establish clear boundaries between administrative support and clinical diagnosis to avoid deploying AI for tasks requiring licensed clinical judgment

Hudson expressed enthusiasm about the potential to fundamentally change healthcare delivery through the partnership. “I’m very passionate about population health, and I’m very passionate about health outcomes. I’m really excited about the opportunity to help drive change in this space.”