Kathleen Wessel (Host): AI is redefining healthcare leadership, transforming AI from a tool of automation into a strategic partner, driving smarter decisions, operational resilience, and elevating patient care.

Hello and welcome to AHA Associates Bringing Value, a podcast from the American Hospital Association. In this podcast series, we speak with AHA Associate Program business partners, check in on their healthcare initiatives and learn how they support AHA hospital and health system members.

Join us to discover how agentic AI is transforming healthcare, and how AHA members can strategically evaluate, implement, and future-proof AI solutions to drive impactful results and avoid missteps. I'm Kathleen Wessel, Vice President of Business Management and Operations with the AHA. And today, I'm joined by Amy Chivavibul, AI researcher with Ailevate. Amy, welcome to the podcast.

Amy Chivavibul: Thank you so much for having me.

Host: I really want to start this discussion. We'll kind of go general question, and then maybe get a little bit of your background. But AI is everywhere, but what sets agentic AI apart is how it's helping AHA member hospitals lead smarter and more adaptive operations. Can you share some insights with us on that?

Amy Chivavibul: Yeah, absolutely. AI is everywhere. And it can mean a few different things to different people. So if we start at the very basics, most people think about ChatGPT, which is a great starting point. ChatGPT is based off of large language models, which are great at generating text, hence, the term generative AI, which you may have heard before. And agentic AI builds on the basis of generative AI. AI agents are able to understand context, to understand goals, and then actually execute actions.

So, just to give a concrete example, if we wanted to take this podcast transcript and turn it into a newsletter, this would be a generative AI task. But If we were to use an agentic system, an AI agent might crawl upon engagement data and marketing data to understand AHA's audience segments. It might crawl through its marketing content to understand the market team's goals for outreach and then. take action. AI agents would then be able to personalize and actually send out the newsletter to the appropriate audiences.

So when I think about agentic AI in healthcare, I would say we're still in very early stages. I think of tools like the smart sensors that are being used in operating rooms to track activity. I think of ambient listening AI, which is generating clinical notes for conversations between patients and clinicians.

And then, when I think of, on the administrative side, we're starting to see a lot of tools being implemented in areas like revenue cycle management, which is what we're focusing on at Ailevate. So, I think, at the very heart of it is that agentic AI is moving this sort of AI needle, as I like to call it, from passive prediction to active participation. And that's where the real value is going to be when we think about agentic AI in healthcare.

Host: Thank you for kind of walking us through some of those examples. I think that was really helpful for me to kind of think through the broader scope. I always get very narrow on that topic. I'm really curious about you. Can you share a little bit of your journey and how you got to Ailevate and kind of what's sparked your passion for AI in healthcare?

Amy Chivavibul: Of course. So, I actually started my career in user experience design, which if you weren't familiar with the field, is all about understanding digital interfaces and the products that we interact with on a day-to-day basis. So when I was in the UX space, I was really focused on how people understand and navigate technology.

And one of the main takeaways I got was that users lose trust in technology pretty quickly and pretty easily. And so, in 2023, which is when ChatGPT took off and large language models became very popular. I became super interested in how LLMs were generating outputs and how people were responding to them. Were they trusting the outputs? Were they convinced? How were they responding to this floating text that was coming up from this mysterious machine? And so, I really began to focus on AI research, which eventually led me to Ailevate where I actually am part of the research and development team where I build and test multi-agent systems that are able to reason at a very high level.

For example, we are working on our revenue recovery platform, which is helping hospitals automatically recover and correct denied claims. So when I think of how I came into healthcare and why it really spoke to me, it's one of the spaces where technology and humans really meet in a very distinct way. Every decision that gets made, every data point impacts someone's life. And especially coming from a user experience background, that also really spoke to me. And I think working in AI and healthcare is really also an opportunity to demonstrate that AI can be a force for good when it's implemented ethically.

Host: Yeah, I think, you've hit on a couple points where it does seem like in healthcare, the barriers are a lot higher than other industries, other markets. So, healthcare has been somewhat slower in adoption and lagging, and that element of trust is just huge. So, thank you for touching on those things.

When hospital leaders are exploring AI, what should they be looking for? And what are some of the common missteps that you can help identify for them to avoid?

Amy Chivavibul: Yes. There's so much hype in the AI space. And if I was a leader, I would think about three different questions, one being can agentic AI plug into our current systems? Like I said before, the power of agentic AI is being able to understand your data and execute actions. So if a potential solution can't hook into your EHR or your scheduling system, or your claims software, an AI agent can't do anything really. And so, it's completely siloed. And so, interoperability in this new world is going to be very, very important. So yes, does it plug into our current systems? Question number one.

Number two, I would consider, can we see what AI agents are doing and how they are reasoning? Transparency in the age of AI is very important. Many of us know that we shouldn't blindly trust AI outputs. And so, leaders should be asking vendors how AI agents are operating in different scenarios. Whether we can see what data they are using and then also demonstrating how humans can override actions. So, this level of transparency is imperative-- yes, absolutely-- for compliance, but also for trust when it comes to teams. At the end of the day, someone isn't going to use a tool that they don't trust. So, that was question number two.

Number three, I would reflect on how agentic AI might change ways of working. . So technology is going to shift roles in every industry, and I would encourage leaders to step into the shoes of the people who are potentially using these tools every day and reflect on whether it's going to make life easier or make life harder. So, what kind of responsibilities are shifting as agentic AI is embedded into different workflows? What kind of responsibilities, such as oversight are created? And which responsibilities are going to be outsourced? And are teams comfortable with those responsibilities being outsourced in the first place? These kind of conversations about responsibilities are super important when it comes to avoiding future resistance of technology.

Host: I think one of the current-- you've raised a couple of points. And even this week, within our organization, we've had a conversation about when you're thinking about the adoption of AI and how it relates to your current systems, like, how do you not replicate what was already a bad process? I think there are so many considerations when you're thinking about kind of how does it fit with your current workflows, how does it fit with your current teams? So, thank you for raising some of those elements. From your experience, like, if we play that out a little further, where does agentic AI deliver kind of the most impact for acute care settings? And what have you learned about how teams are adapting to that, given what we've just talked about?

Amy Chivavibul: Absolutely. I believe that agentic AI is going to make the most impact in tasks that are high volume, repetitive, but also require a sort of reasoning that pure automation can't handle when it comes to certain nuances.

As I mentioned before, I think healthcare is still in the very early stages of agentic AI, but we are starting to see early adopters push the boundaries on AI implementation. For example, several hospitals and physicians are starting to adopt decision-making tools like OpenEvidence, which is basically the medical ChatGPT. And it helps physicians generate evidence-based answers to their questions. And OpenEvidence is still very much in the generative AI phase. But if we were to think of it in the agent phase, this is all theoretical, but think of OpenEvidence being embedded in your medical charts where OpenEvidence is able to track a patient's chart over time and identify patterns and surface relevant evidence and recent medical studies that a physician might find interesting when thinking about a plan for that patient's care.

I also think agentic AI is ripe for administrative workflows, like prior authorization and revenue cycle management. The latter of which I'm focusing on at Ailevate, where we have AI agents that are reading 835s and 837s. and trying to understand why a claim was denied in the first place and leveraging complex rules like payer guidelines. So, there is a lot of opportunity in this world of agentic AI and healthcare. And I think the underlying theme Is change management and how do we deal with this change? Because it really is uncharted territory. And I think OpenEvidence is a great case study of how when a tool is easy to use and trustworthy and still allows room for human judgment is when people and teams adapt in and can thrive with an AI tool.

Host: I want to ask as we look ahead. How do you see the evolution of AI in healthcare operations? What should members be preparing for?

Amy Chivavibul: Absolutely. I see a very natural progression, wherein healthcare technology, we started in automation, which is very much rules-based tools. And now, we're starting to dabble in the generative AI phase with tools like OpenEvidence. And then, the next stage is agentic AI, which I've been touching on.

So then, what comes next? Like, what should leaders be preparing for. And I envision a future where we are seeing continuously learning multi-agent systems. So, that would look like a network of many specialized agents working on across all areas of a hospital's functions.

So, just to give some ideas of what that might look like, I'm imagining a specialized supply chain agent, for example, who is talking to the operating room tracking agent. And through that communication, they're able to understand that, "Hey, we don't have enough equipment for next month's surgical slate. So, we need to order X amount of equipment." And they go ahead and do that with some sort of user approval.

I also think of the staffing use case. So, we have an agent who is really knowledgeable about hospital staffing schedule. And another use case I think of is staffing. So, consider a staffing agent who is able to communicate with the scheduling agent. And they're actually able to rebalance nurse schedules in real time due to upcoming commitments and future bottlenecks that they can identify.

So, this is really what I see as the future of a hospital that's embedded with a multi-agent system. I really envision a world where AI can predict future needs for a hospital, but also learn from processes that worked and then also fell short. And when I think about what leaders be preparing for, in terms of do leaders prepare for this new world? And I mentioned this before, but interoperability is going to be paramount. Making sure that all of our systems can work as a whole is going to be very important for the future. Making sure that AI literacy is in place is going to be paramount. I think that is a through line of this entire podcast as well, is clarifying how AI is going to be used, where human judgment still matters, and in demonstrating that AI is going to be a thought partner and a helpful tool as opposed to a replacement.

So just to sum everything up, I believe the trajectory that we're really on is a path from automation to an adaptive intelligence that is going to be really embedded in to the fabric of healthcare.

Host: I love that kind of end-to-end picture. So, I thank you for walking through some of those examples. I hadn't really thought of the multi-system process. But of course, that makes perfect sense. Amy, thank you so much for joining me and joining the podcast today and sharing your insights with members. It's been always educational.

If you would love to learn more about Ailevate, please visit Ailevate at www.ailevate.com. If you would like to learn more about the AHA Associate Program, please visit us at sponsor.aha.org. This has been AHA Associates Bringing Value brought to you by the American Hospital Association. Thanks for listening.