Learn how conversational AI agents are capable of naturalistic dialogue while maintaining strict adherence to conversation flows and clinical guardrails.
Leadership in Artificial Intelligence and Compassionate Specialty Care Engagement with Morgan Gill from Oshi Health
Lack of patient engagement and non-adherence in specialty care creates severe operational and financial impacts for providers while harming patient health
Gastroenterology is among the most complex specialties, with high non-adherence rates, stigma, and treatment dropout making patient engagement especially difficult
AI voice agents can close this gap, but there have been no previous controlled evaluations of AI-initiated voice outreach conducted in a real-world setting
- RadiantGraph, an AI-powered proactive care platform designed specifically for the healthcare industry, was able to present findings at the 2026 International Conference on Artificial Intelligence in Medicine (AIME), in Ottawa, Canada July 7-10, showing that AI-powered voice outreach increased appointment completion rates by 3.4x overall, including a 24.7x increase among patients who had created an account but not yet scheduled an appointment. It also enabled a 9x increase in outbound call volume compared to human agents. Equally important, no safety incidents or escalations were identified during the study.
The study, which was a collaboration between RadiantGraph and Oshi Health, a virtual-first gastrointestinal (GI) specialty clinic, used RadiantGraph AI Voice Studio to deploy conversational AI agents capable of naturalistic dialogue while maintaining strict adherence to conversation flows and clinical guardrails. Voice agents in the study were able to answer nuanced questions, inform members about their personal program, explain costs, and more without requiring a live representative. This represents one of the first controlled evaluations of AI-initiated voice outreach in a real-world setting aimed at driving healthcare engagement and program enrollment.
Oshi Health is the only nationwide multidisciplinary (GI) medical clinic combining specialized medical, dietary, and gut-brain care proven to get patients better, faster. Available to adults in all 50 states with the convenience of telehealth and in-network with most insurance providers, Oshi's care accelerates the speed to diagnosis and symptom control—significantly improving people’s quality of life. Morgan Gill shared her experiences and insights about their work, including why certain patient groups responded positively to AI outreach. As a senior growth & analytics manager at Oshi Health, she digs into the data and harnesses it to improve patient outcomes by bringing virtual digestive care to more people. She has a background in strategy & operations, finance, and emergency medical care. Based in Jackson Hole, she spends time outside of work exploring the mountains or teaching Pilates.
Adherence is critical, as 50-60% of chronic disease patients mismanage or discontinue their treatment within the first year, leaving half or fewer adhering to their prescribed therapy after 12 months. GI conditions in particular provided an optimal testing ground for the platform since related diseases and treatments are especially complex. GI conditions carry enormous burdens, yet engagement is lacking. Many patients are also impacted by stigma that can limit adherence (up to 84% in IBD).
There’s a huge access problem in gastroenterology. Nearly 40% of metro counties and the large majority of non-metro counties have no gastroenterologist at all, and average wait times for an in-person GI visit have run around 40 days in major metro areas - that's the gap Oshi's virtual, multidisciplinary model was built to close, which means we have a lot of patients to serve. Oshi's own published data shows patients waiting a median of about six days for a first appointment instead of ~40, and the large peer-reviewed cohort study found over 92% of patients reported symptom improvement.
Proactive, personalized outreach — including AI-assisted outreach — is what makes a model like this scale nationally without losing the "high-touch" feel. It's not virtual care instead of personalized care; it's the mechanism that makes personalized, proactive care possible at scale.
Healthcare research suggests that people may be more willing to discuss sensitive topics around their health with an AI versus an actual human. Platforms like RadiantGraph AI Voice Studio are able to be tailored to a wide range of healthcare journeys, including chronic disease management, member education, adherence support, and care navigation.
Conducted over seven weeks with nearly 8,800 members across three distinct patient journey stages in Q4 2025, the deployment saved approximately 75 hours of staff time per 1,000 contacts. The member population has increased since then while results continue to hold. Since voice AI agents handled first-order support, the member support specialists at Oshi health experienced a 70% efficiency improvement. This approach can enable organizations to reach substantially more patients without increasing staffing levels.
According to Gill, staff numbers stayed exactly the same as what the team would expect to take on this outreach. They were able to take on more meaningful work. The group that was most impacted was the care coordination team, who was able to do fewer repetitive tasks and spend their time on calls including complex care issues, inbound calls requiring triage, and improving care outcomes. Reducing busywork helped staff operate at a higher level of their clinical license.
"Healthcare has long known that personalized outreach from human care managers improves patient engagement, but scaling those efforts has been difficult and expensive, especially with specialties that require more complex answers and care. This study provides evidence that advanced agents can help close that gap by reaching more patients, increasing appointment completion, and doing so safely alongside existing care teams."
RJ Ellis, PhD, lead study author and Principal AI Scientist at RadiantGraph.
"One of the biggest challenges in specialty care is helping patients take that next step from interest to action...We saw first-hand how effective tailored AI voice outreach can be in engaging patients, when it makes sense at a personal level, while supporting earlier access to care without adding burden to clinical teams.
Sameer Berry, MD, Co-founder and CMO at Oshi Health, and eminent GI expert
Study members who received an AI voice call were 3.42 times more likely to complete their first appointment than those who did not (p < .001). Members who had already created an account but had not yet scheduled an appointment showed a 24.7-fold increase in appointment completion rates (p < .001), suggesting a powerful point in the member journey where a timely, personalized outreach can move someone from intent to action. AI agents achieved a 23.99% answer rate, well above reported industry benchmarks of 2%-20%.
Gill weighed in on the importance of targeted outreach to patients during the care journey. She mentions that the gap between "account created" and "scheduling appointment" is a stalled intent moment. Patients may have made the decision to seek care, but something kept them from converting, whether it be a super busy schedule, uncertainty about next steps, or simply forgetting. Their work targeted outreach over call volume. The biggest lift showed up in the stalled-intent patient journey segment, not across the board. Clinics chasing "more calls" as the goal might miss this. Gill emphasized that the win is in knowing where in the journey AI adds the most value.
"What excites me most is that this isn't AI replacing the relationship between a patient and their care team - it's AI clearing away the friction so that relationship gets more of the team's time and attention, not less. Healthcare should hold a higher bar for that than almost any other industry, because trust here isn't just a nice-to-have."
Morgan Gill, Oshi Health
Potential future research areas including EHR integration, multilingual support, and larger-scale studies across additional clinical settings. The study focused on transparency, security, and targeted interactions to improve outcomes. Their recommendation for specialty clinics considering partnering with AI agents? "Start with defined, low risk conversation flows. The agents in this study could hold naturalistic dialogue but stayed within strict conversation flows and clinical guardrails. This combination of natural and bounded interactions is the single most replicable lesson from our work."
Read the paper, "Scaling Healthcare Engagement with AI Voice Agents: A Real-World Evaluation Study," in Lecture Notes in Artificial Intelligence (LNAI) 16749, pp. 1–5, 2026 https://doi.org/10.1007/978-3-032-30813-9_67
About RadiantGraph
RadiantGraph is the AI engine powering proactive engagement and better outcomes in healthcare. RadiantGraph platform, including its Voice AI agents, support a variety of use-cases in pharmacy operations, Medicare/IFP onboarding, preventative care outreach, and benefit utilization. The company has helped clients drive +70% team efficiency and have a positive brand impact that is 10x of human agents. Headquartered in San Francisco, RadiantGraph is HITRUST-certified, SOC 2 Type 2 compliant, and HIPAA-compliant. The company is led by Anmol Madan, an Massachusetts Institute of Technology PhD and author of the forthcoming book, The Machines Will See You Now (Johns Hopkins University Press). Learn more at radiantgraph.com
About Oshi Health
Oshi Health is a virtual-first specialty clinic dedicated to delivering comprehensive, evidence-based care for people living with gastrointestinal conditions. By combining clinical expertise with technology-enabled engagement, Oshi Health helps members access timely, personalized care regardless of geography. Oshi Health is headquartered in New York, NY. Learn more at oshihealth.com.