Here's what we'll cover
Here's what we'll cover
Here's what we'll cover
Check out our first blog of this series: Bringing patient transparency to the insurance process.
Side effects are one of the leading reasons patients stop GLP-1 treatment. That’s why mitigating side effects as they arise or managing them as they happen can make a significant difference in helping patients stay on the medication to achieve their health goals.
For example, gastrointestinal side effects such as nausea, bloating or stomach pain are common with GLP-1 use. If a patient experiences nausea and doesn’t know whether it’s normal or how to manage it, they might skip their next dose or stop treatment altogether. But if that same patient can easily report their side effects and get timely support from their provider or care team, they’ll be more likely to stick with treatment and achieve better outcomes. This is the experience we’re scaling at Ro.
Bringing speed to side effect management
Every day, Ro receives tens of thousands of messages from patients—questions, updates, concerns, and everything in between. Buried in that volume are messages detailing side effects that require timely intervention. Manually sifting through all of these messages and responding quickly just isn’t feasible with human resources alone. That’s why we set out to build a technology solution that could help our providers focus their time where it’s needed most.
We developed an in-house large language model (LLM)-based triage tool that uses AI to:
Automatically identify messages that are related to side effects
Classify the potential severity of the side effect(s) mentioned
Alert the right care team member to respond
As part of the Body Program membership, patients already have 24/7 access to message their provider and care team, but this tool is designed to ramp up the speed at which their side effect reports or questions are answered. Our custom LLM is trained to recognize message patterns that may signal side effects or urgent needs. The model analyzes incoming patient messages, labels potential side effects by severity (mild, moderate, severe, or urgent), and generates appropriate alerts for providers and care teams to review and action on.
Importantly, the tool doesn’t make clinical decisions—it gives providers and care teams the visibility they need to deliver timely, informed, patient-centric care.
Real-world results
When tested on our platform, the model achieved a 94% true-positive rate in detecting side effect–related messages. And when we studied response times before and after implementing the tool, the improvement was statistically significant:
Median response time fell from 115 minutes to just 33 minutes—a 71.3% improvement
Response times significantly improved across every severity level
Urgent messages were answered fastest – on average in less than 26 minutes
Our team published and presented these test results as a peer-reviewed abstract at Obesity Week 2025, demonstrating the potential of AI tools to help make care safer and more responsive for patients.
In a traditional care setting, managing side effects often means scheduling an appointment, waiting days to be seen, and hoping to get the right guidance once you get there. With ro.OS, patients can reach out as soon as symptoms arise and receive timely guidance the same day. And if they need additional support, such as anti-nausea medication, they can get relief much sooner.
“It’s exciting to see the clear evidence of impact when we apply scientific rigor when evaluating technological innovation. In this case, our LLM improved response times for patients. Because side effects are a common reason for medication discontinuation, faster follow-up and support are likely to help patients stay on treatment and achieve their health goals.” – Dr. Lauren Broffman, PhD, Senior Director, Scientific Affairs and Real World Evidence
For patients, this tool means greater confidence that a provider is there when they need one with counseling or intervention when it matters most. For providers, this tool means timely notifications, and reassurance that they’re triaging the patients that need them most. With the support of AI, providers can spend more time focused on patients and less time sifting through messages.
But the work here is far from done. The team will continuously review patient messages with human experts to refine the model’s accuracy. Coordinating these audits is a critical part of the feedback loop that will inform our iterative improvements to deliver a better experience for both patients and our providers.
“This tool is powerful because it speeds up care for patients when they need it most. And in designing it, we established infrastructure that we can use to more easily build additional AI-powered tools across our platform.” – Will Walmsley, Principal Machine Learning Engineer
Powered by ro.OS
Ro’s LLM-based AI triage tool is already proving its benefit for patients and providers alike, and is just one of many features made possible by ro.OS. With patient messaging, medical records, provider task routing, and care delivery workflows integrated into a single platform, tools like this can support a seamless experience by analyzing messages, surfacing alerts, and guiding patients or providers without friction.
