Nashville-based Reimagine Care, a virtual oncology care organization, recently announced a major evolution of Remi, its AI-powered assistant that supports cancer patients between visits and helps oncology care teams respond earlier to symptoms and treatment concerns.

The improved system effectively uses artificial intelligence (AI) and large language models (LLMs) to facilitate more natural, advanced conversations with patients, gather crucial clinical details about their symptoms, and recognize when escalation to human providers is necessary.

Healthcare Innovation spoke with Dan Nardi, CEO of Reimagine Care, to learn more.

Reimagine Care specializes in virtual oncology care. How does this work in practice?

Virtual oncology care in our world is being there as that point of contact for patients for those hours that exist between visits. When we say virtual care, it's a combination of technology and Remi, our virtual assistant via text. The benefit is that it can be there for you within seconds. It's interacting with you, asking questions, and helping to infer the right intent that you need in the moment. We always have it backstopped by an oncology-trained team. We have MA’s, RN’s, APP’s (Advanced Practice Providers) on staff 24/7, so we always have the right human in the loop at the moment. We let Remi do a lot of the heavy lifting.

Tell me about Remi. What features make this tool specific to oncology care?

What we do is build clinical pathways, specifically into Remi. They're based on NCCN guidelines, originally years ago, when we first started to build them. We've adapted and evolved them to the point where they're very specific to our use cases and Remi's, but they're based solely on oncology. I think that's also important to know Remi is Agentic AI. But there is a point where the AI takes a backseat, and then the clinical pathway takes over. That is what helps us prevent any hallucinations or any kind of concerns about AI delivering clinical care, because AI is not delivering the actual clinical care. AI is there to help interpret intent, understand responses, and get you into the right pathway. But once you're in the pathway, it's very structured, with the questions and follow-ups and how it works through that flow. The underpinning is all oncology-based because that's our population we're supporting.

Remi has some contextual information from Health Information Exchanges (HIEs) that allows us to work on a personalized approach for each patient.

The press release mentions Remi evolving. Could you explain?

We've moved away from Natural Language Processing, NLP, which were some of the earliest versions of agents, and we moved to true LLM (Large Language Model). It's actually a combination of agents. We have one running for enrollment; it's a separate agent that helps them answer questions about the program and their specific use cases. Remi is the one that is doing more of the day-to-day interactions, both proactive check-ins and reactive. It’s more empathetic, but also just more humanistic in its responses.

We're very clear anytime one of our team members takes it over. This new version is a lot more like natural like language and conversation.

What challenges did you want to address?

Engagement. We're always really focused on engagement. Are the patients enrolled in the program using it? Are they getting any benefit from it? Right. We still have very high engagement, 93 percent engagement within the first 30 days. We are constantly thinking about how we are building programs with the highest engagement rate, because it's proven that when patients are engaged in their care, outcomes are better. Even though we have very high engagement, we're always thinking about that.

One of the things that will help with the scalability of care delivery in the future is how much Remi can resolve on its own. Right about now, Remi is resolving about 50 percent of all the interactions that patients have with it. As we get it to maybe 55 percent or 60 percent, that reduces the human time needed. It allows us to support more patients at scale. And that's a big thing that healthcare in general needs to be addressing right now, because we don't have the caregivers, nurses, providers, and clinicians we need to support these patients.

What challenges did you face when implementing the AI assistant?

If you're not pushing the envelope, you're not going to run into issues. As we were scaling up so rapidly with this new version and having more and more interactions, we realized we couldn't just rely on human quality checking. We actually built another agent, so every single one of our patient interactions is now fully automated. We have an evaluation tool that runs against every interaction, and then it goes through and will determine which ones should be flagged for human follow-up. Now we have multiple agents out there doing these evaluations, and that's had a huge impact on us.

What has the feedback been?

The employees love it. It's a source of pride, this new version. Our existing clients really love the new features. We're now able to move forward and enable new languages much faster; in the previous version, it might take months to implement a new language. I think, honestly, it took us three or four months to do Spanish.

We're talking with the product team and thinking about new languages. There's a really cool opportunity, and we haven't done this yet, but we've started to play around with it, where you can have real-time translation back and forth.

We've built Remi and these programs for the patients, but everyone going through cancer has a caregiver, if not multiple. Their family members who help to support them have access to this. They feel comfortable.

What other future developments are you considering?

One of the things we're actively working on right now is building programs specifically for cell therapies like CAR-T (Chimeric Antigen Receptor T-cell) and bispecific therapies like BiTEs (Bispecific T-cell Engagers), as these are moving into the outpatient setting, which is where much of the oncology industry is headed.

What is your advice for healthcare leaders?

Push the boundaries with care. We have to be pushing the boundaries as comfortably as we can, as carefully as we can, because that's where real innovation comes. If we just play it safe, if we go with the status quo, we're not delivering the better experiences that patients need and expect. We need to bring healthcare into that on-demand world, meeting patients where they're at in the moment and supporting them how they want and need to be supported.