What it really takes to bring AI into the hospital room: Avo’s physician-first case for AI-powered clinical decisions

0:03And so this impossibility was really what drove me and my co-founders way back when we were starting AVO to start it. It was to solve our own problems of this overwhelming nature of what we need to do in order to really do our work well.

0:27Yeah, here's Saperstein, co-founder and CEO of AVO. Great to have you on at Startup Health. It is awesome to be here, Logan. We've had the opportunity of getting to talk multiple times over the course of building this company. And so it's exciting to step in now here in 2026 and hear where the company's at. You're at an important inflection point. You raised a Series A earlier this year. Congratulations. But that is the end of one journey and just the beginning of the next big one. So start off by giving us a snapshot of what AVO is right now for those who are uninitiated and kind of what the Series A enables you to do.

1:09So at AVO, we have a vision of making the world a place where better care decisions are made. And what that really means is that I, as a doctor, right, I'm a hospitalist and an informatician. I have a task in front of me, which is to make sure that the patient in front of me is going to do as well as they possibly can. And to do that, I need to make sure that I'm making the best care decisions possible. And in our current system, that has become almost impossible because there's so much that I need to do to make that happen. I have to take into account all the patient data that exists in the electronic health record.

1:51I have to take into account all the guidelines of the things that I need to do, whether they're from established bodies of guidelines or what the health system is telling me to do. And then I also need to consider all the other stuff that I got to do, which is, hey, doc, you have to discharge this other patient. Hey, doc, you have to fix your note because of this. And this is a query. Hey, doc, you got to do this also. so much to do. And so this impossibility was really what drove me and my co-founders way back when we were starting AVO to start it. It was to solve our own problems of this overwhelming nature of what we need to do in order to really do our work well.

2:32And this is something that led to us having burnout. Hey, you can't accomplish this well. And then how do we actually accomplish for patients the top quality that we can give them? How do we make the best care decisions? As we've progressed now, and most recently, as you said, raising our Series A in January, we are taking this from a level of, okay, solving individual clinical workflows with a founder-led approach to really a true enterprise approach and also solving system-level priorities. And what that means is, number one, rapid adoption. How do you make sure that all of your clinicians that you want to be using the software or using it.

3:12Number two, ensuring that there are system level goals that are accomplished. So how do we have, of course, efficiency for the clinicians, but also quality of care that we can measure, how do you, that you can improve outcomes. And number three, are there ways that we can also improve compliance and ways that we can also improve the financial status or revenue for the system? And so we have been doing a lot more work over there too. So tell us what the actual product experience is like for the clinician. I know it's clinical decision support assisted by AI. I think a lot of people who are getting more familiar with LLMs could envision what that would be like because they get advice from ChatGPT or Claude.

3:54So they kind of understand that. Help us understand what the bedside experience is like right now. And then as a second, you know, as a tag along to that, why aren't they just getting that through their current medical record, like their Epic, why do you need a sort of an add-on service? Take my workflow as an example. I'm a hospitalist. Okay. So I have 20 patients that I'm going to be caring for today. Let's zoom into one of those patients. I need to figure out what's happening with them over the day. Most importantly, what happened over the last 24 hours and especially overnight that I may have missed as I'm coming in the next day?

4:33What am I doing with them? And importantly, am I discharging them or not as part of that decision process? And then of course, I need to communicate with them and communicate with everybody around in the system to make sure that the other clinicians are going to do their part and that the other ancillary staff will be able to help for whatever needs to happen. Epic is able to guide me as to perhaps a summary of what happened with this patient. But it's not going to guide me as to, well, what should I do next, right? For me, if I have a patient who has heart failure and their heart failure is currently leading them to gaining weight, well, I'm not going to get that in a summary.

5:16I need to figure out, well, what is my next step? And if my next step is to reduce the amount of fluids that the patient has, but yet they simultaneously have sepsis where I would normally give fluids, well, that's pretty complex. And I might be going out to an external decision support to try to ask what should I do, whether that's chatbot style, or if that is a, you know, amalgamation of guidelines style. I want that inside of my workflow. I want my system to be able to have all the patient context, and then to guide me, here's what you should do. Here is suggestions of what your next steps might be.

5:53I live in an agentic world. Like I have an agent drafting my email responses every morning so that I can review them, make whatever changes I want and send it. So I expect things to work the same way in medicine and they don't unless you use AVO. And so we sit inside of the EHR as an intelligence layer that can take in all the patient data, run it through clinical decision support. and we currently partner with Dynamed to be able to run these questions, get concrete answers, and use that as your next step of the workflow. And for me, the next step of my workflow is producing a note that incorporates this decision process.

6:34And what we're able to do is in writing this note, there's also the idea of capturing what I'm doing and the complexity of the patient in that note writing. So I don't just say, I'm going to give a little bit of fluids. I say, There's a balance between the heart failure and the sepsis, and therefore, it is a strong consideration of giving this amount of fluids or not and how I'm going to watch it because of this complexity. Well, that complexity should be captured because that capturing of complexity allows for anybody else that's reading it to understand what's going on, but also allows for health systems to bill appropriately for what I did.

7:13And so it allows for true driving of quality capture at the point of care through the process of evidence-based driving of what I'm going to do. So I know you started this company from the perspective of being a practicing physician, solving a problem that you were having. And so it's easy to understand the angle why this would be attractive to the physician and also to the patient who's getting this more accurate care. Now let's shift our focus to selling into enterprise, into a hospital system. What are the things that they care about and how is that shifting your focus, shifting the product?

7:53So health systems are spending a lot of money on post-care delivery revenue capture. Like how can we recover value after care? and there's a shift that's happening to try to capture value during the clinical encounter. Like how can we actually capture all the things that are happening while the patient is there in front of me? And the reason for that is multifold. One is if you're trying to figure out what happened afterwards and then you're going to send me, the doc, a query of what's happening afterwards, there's the possibility that I won't answer it, which will be revenue leakage. there's also the possibility that I will change it, but it won't actually impact care the way that you expect.

8:41Whereas if you're able to capture it at a point that is truly at the point of care, you can not only influence the ability to actually deliver quality better, but you're not going to have revenue leakage because it's all caught when the patient is there and at the point of care. So systems are trying to capture that. Are there flags going up for you in terms of concerns about the push towards revenue capture, driving clinical decision support? Are there ways you're thinking about that interplay and ways in which the enterprise buyer could kind of impact the system in a way that's maybe not physician first?

9:19I'd say these two things have always been at odds. It's always, are you going to be doctor first, where it's, okay, I just want to get my notes in a way that can be understandable. As an example, I want to have efficiency. I care only about what's happening with the patient. Or is it going to be quality, as quality means to the system, which ironically is revenue capture, right? When they say quality, that's what they mean. And so these two are at odds, because when you think about revenue capture, well, I want all the details in your notes as much as possible so I can capture everything from the complexity of the patient so that I can get more for their stay, as an example.

9:59And so these two have long been at odds. And the approach that we're taking is we are driving through the lens of making the best care decision, right? So when I'm trying to figure out what to do with the patient, I want to have clinical decision support as an intelligence layer that will help guide me. What should I do? And as it's guiding me of what I should do, well, I want that documented in a way that is accurate and reflective of what I'm actually doing. And that leads to me having the appropriate documentation that will satisfy the requirements for revenue, compliance, and quality instead of you needing to chase me later.

10:35So it actually is coming from a physician-first lens, but in a way that also makes the compliance and quality guys happy. And to give you an example, let's say I have a patient with heart failure. If I write heart failure versus heart failure with reduced ejection fraction, I will get paid more for the latter. If I order an echocardiogram that shows that the ejection fraction is reduced, well, that's influencing my clinical decisions knowing that. Appropriately documenting it is something that should be present. The way that most docs are asked to document that is through a click button. Does the patient actually have reduced ejection fracture?

11:18Click here to confirm. That's crazy. I already got the echo. I saw the echo. I want that in my notes. Same idea with, let's say, DVT prophylaxis. We try to prevent clots for patients that are in the hospital. And so I give heparin or enoxaparin as a drug to be able to prevent these clots. If I give that order, that means that I'm doing it. Yet I am being asked at the bottom of my notes, are you giving DVT prophylaxis? Choose which one you're giving. This should be automatic in my notes. It should be already there. Those are some examples of the way that we can have this revenue capture lens in a way that's really from driving the quality of what I'm doing just by picking up from what I'm doing inside of my notes.

12:02Over the last five or six years, this market has heated up. How hot is this agentic clinical decision support market today compared to where it was when you started? How competitive? What kind of energy is there? And how does that shape your direction? So everyone claims to be doing everything in the current AI world, right? And so from an initial standpoint of competition, there's always the fear of, well, can anybody do anything with AI? And then you start to realize when you are deep in it, when I go back to work as a per diem hospitalist, holy smokes, there's so much that needs to be done here.

12:43No one's solving my problem because no one gets it. No one gets it. And the thing is that there's really three different parts that are going on here. I need to be able to understand the patient context well. I need to utilize that in a way that's helpful for me, the doc or the clinician. I need to combine that with the decision support of, well, what should I actually do? Not in a way that's best practice alerts that I'm going to ignore, but a way that's helpful for me. And then I need to incorporate the revenue side because that's the system priorities. Otherwise, this is going to be an extra additive layer that's just going to bother me.

13:18This is putting things right into the place where it needs to be. Nobody really gets that unless you experience it and live it. And so all of these companies are like, oh, we can use AI to prototype something new that'll work here. Like, sure, but it's not really going to solve the problem the same way. Final question. I want to shift gears a little bit. We talk about building a startup. There's sort of a truism that what got you here isn't always what's going to get you there. And having just raised this series A, I want to hear from you kind of how you've changed your thinking, how you've evolved as a CEO of the team as you think about building, getting the right people on the bus.

13:56Where's your head at right now? And maybe what are some of the things that you have thought about over the last few months? So we are continuing to build our band. I heard two really great analogies. One is that every company, as they move through the startup journey, is really building a band. Think a jazz band as an example. And as you're adding on a additional players that are able to play their respective instruments amazingly. You have to figure out what level of expertise are they bringing in? How do they play and jive with the others? And then what's the sound that you want to be able to produce at each part that can work really well?

14:33The second analogy is that I am standing on the bank of a river and as logs are floating by, I need to hop onto the specific logs. I'm not hopping onto all of them. I need to hop onto one that can support us, take us forward, sustain us, and maybe continue to hop log to log. And the thing is that as we hop log to log, we take those band members with us. It's really both analogies at once. And so the seasonality is like, we're going from a founder-led, product-driven, solving our own personal pain point company to now one that is an enterprise focused with almost 50 employees with the different specialties and expertise that are being brought into the company.

15:18And so through this journey and the seasonality, as we're hopping log to log together, right, to get onto larger logs that are continuing to move fast downstream, that are working well together, there is a change in the way that things are done. Like the leadership styles, the way and the format of how our mission has expanded the ability to deliver. And as these changes happen, I always have this self-reflection of like, wow, we are really in this and in this moment. And I want the intentionality of, okay, this is how we're driving it. Almost like, can I put a steering wheel on that log or maybe a tail fin to help guide it where we want it to go in the river as it's going along.

16:09In this latest iteration now, post-Series A, as we've continued to expand the team and continue to bring on expertise, it's always this reflection of like, okay, we're in the woods, we're going on, and now attempting to steer with that tail fin so that the whole band can continue to play music well together. I love it. I love it. You've got a little daughter at home. Once she gets a little bit older, how do you want her to be thinking about the work that you're doing today? I would say take it from bands and logs floating down the river to this is what is actually needed in clinical medicine. And this is at the forefront of bringing AI safely into clinical medicine in a way that can truly impact care for better care decisions to be made.

16:54And I'm hoping that she'll be proud of what we're able to do. It is a lot of effort and a lot of time to be able to pull this off. I'm very grateful to my daughter that when I come home at 11 o 'clock, she's up there waiting for me. When I have to leave the house at four o 'clock, she wakes up with me and she's only two. But to be able to say goodbye. So Liel, if you're listening to this later, thank you for that. But it is, it's, I am hoping that she'll be proud of what we're able to do here to really pull off a world where we can improve medicine. Beautiful. Yair, thanks for the update. It's been exciting to watch you grow from just a few co-founders at Columbia to the Series A, 50 people, new enterprise clients.

17:34And it's just going to be exciting to see what the next couple of years brings. So I appreciate the update. Thank you. All right. Take care. Be well.

17:45Yair Saperstein, MD, MPH: And finally, an invitation to all of the extraordinary health transformers, whether you are a startup, someone building the future of health or whether you are an investor, funder, or buyer of the solutions out there in the marketplace, this is an invitation to you to join the Startup Health member community. Learn more at startuphelp.com. Look forward to hopefully seeing you soon.