SPEAKER_00 0:00

You're listening to I See What You Did There, a Bausch and Laum podcast offering concise conversations on the people, trends, and topics defining modern eye health.

SPEAKER_02 0:16

Welcome to another episode of I See What You Did There. My name is TJ Crawford, and I am the Chief Communications Officer at Bausch and Laam. AI is moving into healthcare very quickly. And the most interesting question might not be what it can do. The most interesting question might be what it should do and what it shouldn't do. I'm joined today by two people who are tackling that very question. Munjoul Shah, who is the CEO and co-founder of Hippocratic AI, and Manisha Nera Simhan, who is the president of digital health services at Baoshanlam. Thank you both for joining me today. Thank you for having me. We're gonna well, I I'm gonna I'm gonna start with you, Munjal, and this is perhaps a an uncomfortable question.

SPEAKER_03 1:05

What should we never let AI do in healthcare?

SPEAKER_01 1:12

You know, I think that I know the answer to what we should not let it do today. And I think, you know, never is a mighty long time. And so um I think one's accuracy in prognosticating the future is hard. But I'll tell you what we know today, which is that we decided when we started Hippocratic AI that the best way to deliver safety was to deliver scope. So we said we're not gonna let it diagnose and we're not gonna let it prescribe. I don't think AI is safe enough to do that. A lot of people are like, oh, but I use it all the time, ChatGPT. I'm like, you're using it to get a second opinion. You're using it to do almost a pre-triage to decide whether you're even going into the doctor. That's not the same as like, I'm like, would you really have it diagnose your child, order a medication, send it to your house, and you'd put it in their mouth? I'm like, well, you know, not unless it's uh, you know, an antibiotic or a painkiller only. Like, I doubt you're gonna do that with anything even remotely complex. And so I think that's what we did. We found other places we don't think it should do. We don't think it's ready for kids under two because they're very complex patients. We don't think it's ready for high-risk pregnancies. We don't think it's ready for hospice. As empathetic as AI is, the situation of hospice is so emotionally complicated that even humans struggle. I mean, hospice nurses burn out at a very high rate because it is so emotionally tolling. And so um, I think that we found some of the boundaries and we've set them. You know, I there's this debate today in AI where they're like, you know, I saw this a couple weeks ago where it was like, I don't know, AI is so crazy, it might not be safe. And, you know, we need the government to tell us how to make it safe, and we need other people to stop moving forward, otherwise, we won't stop moving forward. I'm like, wait a minute. I'm like, if you really believe your product's not safe, don't make it. Like, and so I think that we took this very strong personal responsibility model that just says, actually, we spent 18 months not shipping our product because it wasn't safe enough. And so I think that it takes an organization that has the self-discipline to say, hey, I I I don't need anybody to tell me or to force me. I'm just gonna do it because it's the right thing. And I think we found some of the boundaries.

SPEAKER_02 3:33

Manisha, as someone who is building this capability inside of a company, what what's your own line? What what's something that that where you say, no, this still needs to be a human?

unknown 3:46

Yeah.

SPEAKER_00 3:46

Two things that come to mind, TJ, are trust and safety, right? So in all our conversations that we've had, and look, I'm only speaking in the realm of ophthalmology because those are our use cases. We are an eye care company. Um when you think about doctors and the fact that they train for years and years, and that trust that they earn with their patients is so sacred, right? When you think about a vulnerable patient who's sitting in front of a doctor, they've got something wrong with their eye, um, they're worried about their vision, right? It is only a doctor who can look them in their eyes or hold their hand and tell them, this is gonna be okay. I'm gonna be able to treat you, right? And that trust is so important. And I don't want us to be in a position where we do anything that would jeopardize that in any way, shape, or form. So I think trust and safety are two really, really important things as we think about how we deploy it and how we use it, especially when we start to get engaged in a doctor-patient relationship.

SPEAKER_02 4:44

Yep. So let's let's flip that question around now. I I had to start with the negative. Where do you look at healthcare today and think why is a a a highly trained human still spending time on this? Manisha, you want to start?

SPEAKER_00 5:02

Yeah, I'll start. I think like that's a great question. I hope that there will be some point in the near future when we look back upon this time and think to ourselves, remember the time when the most trained person in the building was spending their time doing these admin tasks that have now been given over to AI. Um, and that's that is really the goal. And I think we'll get there. Um I think, look, for us, we're starting, our goal is really to help. Um, we're first of all, we're building something that is product agnostic here, right? We want something that's going to really help deliver the best care possible. That's what doctors want to do. And the challenge we gave ourselves was how could we help doctors deliver the best possible care to their patients, which is what the doctor wants to do. Um, there's a lot of things today in eye care, in ophthalmology, really, frankly, across all of healthcare, uh, where it's a lot of repetitive tasks, it's a lot of admin, it's a lot of education, right? And I think AI is particularly suited, and what Hippocratic is building is particularly suited to the education component, right? And we're starting with patient education, that is a beach head here. But I think that there's a lot of value when you're a patient and you have a question at two in the morning, do I need to take this drop twice a day or four times a day? I can't remember, right? And look, this is well documented. Patients don't really remember um all the information that's communicated to them, they recall even less. I think if you're if you're you know in a good situation, they might be able to recall 25% accurately. You're giving them a lot of dense information. So how can we help have AI communicate that on demand whenever the patient needs it? Um I think that would be great if we can do that. And we're very optimistic.

SPEAKER_02 6:50

Yep.

SPEAKER_01 6:50

Manjo, anything to add? I think that the the main concept that Manisha was describing is really what we call clinical abundance. That, you know, for all of human history, we've actually designed healthcare assuming scarcity. The word triage assumes scarcity. Our prior auth protocols assume scarcity. You know, I had a sty on the top of my eye, and I remember I went in to see my um my ophthalmologist, and she was like, you probably need me to drain it, but I can't drain it right now. I said, Well, why not? She's like, Well, it looks, you know, you've had it long enough. Obviously, it isn't going away. Well, the protocol is you have to come back in a week. You know, I it has to be the second appointment. Otherwise, I won't get the prior Os because they, you know, they don't want to be doing it unnecessary. But if you think about what why was that developed? Was that for the convenience of the patient? Was that for the time savings of the physician to be able to treat more patients faster and get more throughput? No, it was because the insurance protocol designed it, assumes scarcity. And so I think that there, for the first time in human history, we can deliver a level of clinical um supply that will allow us to do things very differently. And, you know, many of our cadets, I'm like, why why do I get an eye exam only once a year? Does my prescription never change in six months? Sometimes it might. And but again, it's the scarcity assumption that we have in there and the financial assumption, but but all of this is going to change, uh, I think. And that's that's I think the most exciting part about what's coming.

SPEAKER_02 8:29

Yeah, and and and it's a perfect leading to my my next question. Healthcare has been promised that technology would make people's lives uh easier before, right? And and instead, in a lot of cases, that's led to more screens, more clicks, more administrative tasks. What uh what do you believe this wave, this AI-fueled wave, how do you believe that would be different?

SPEAKER_01 8:58

I mean, I I would start and just say, um, have you ever had AI work as good as ChatGPT did? I mean, I think we all were like, oh my God, what just happened here? Yeah. Um, this is not your father's Oatsmobile, right? Like if you remember that commercial. Like this is a different kind of AI that came out. And we've gone from AI having the IQ of 60 to having an IQ of 120. We've gone from it being barely conversant, and if you don't say the right exact word to your Alexa, it turns on all the lights in your house instead of turns off all the lights in your house. And now you're saying complex sentences, you're sharing your greatest fears, you're sharing all kinds of things. And it's not a few people doing this. Like 20% of all the queries on Chat GPT are healthcare. And there's a billion monthly active users. That's 200 million people, maybe more, right? Uh, people. And so, anyways, there's there's a very different level of technology we've had. And it can finally allow for what I call full delegation, where you actually can give the AI the task and let it do it. Whereas all those other examples you cited were always a tool to work with the human, which burdened the human with having to interact with it and therefore didn't deliver it. So I think there's a paradigm difference, and I think there's an intelligence difference. Although I think most of this is coming from the intelligence difference.

SPEAKER_02 10:34

Yep. So, and Manisha, you touched on this earlier. I I wanted to get to trust, right? So we're talking about something very different than asking Chat GPT to help with with dinner plans, right? If you're if you're wrong in healthcare, it it matters. So how do we go about, and and we're going to get into the partnership next, but big picture, how do we go about earning that trust from physicians and from patients to put AI into that relationship?

SPEAKER_00 11:08

Yeah, TJ, we've all had situations where you've asked AI a question and it's responded with great confidence and given you a fact that it's completely made up, right? And we're slowly sort of overcoming that to a certain extent. I think from a trust perspective, the way I see it, like anything else, you have to build it over time, right? And so you kind of really have to, and look, I think that's why it's very important for us also to work alongside physicians and surgeons and build solutions with them rather than for them, right? And say, well, I think this is your problem, and here's a solution I've created for you. But our approach is we want to spend time with you and understand your pain points, right? What bothers you, what burdens your day? What do you wish you could just wave a magic wand and you know would magically disappear? And let us help take some of that off. So look, I think trust has to be built over time. Um, you kind of have to start with the first step. And um, you know, for many reasons, that was why we decided to start with patient education, um, right, and build that trust. I think once you have built that relationship and that trust with the physicians, with the surgeons, with the patients, then it becomes easier, right? Uh you're familiar, you're trustworthy, and then you can then take on more.

SPEAKER_02 12:28

You did a perfect job of uh leading into the collaboration between Bashan Lam and Hippocratic AI. So you're you're working together, or we are working together uh on something called Orphia. Um what problem are we setting out to solve initially?

SPEAKER_00 12:50

Yeah, so Orphea, think of it as a, I call it a clinical infrastructure platform, but at a very fundamental level, the problem we want to solve is how do we help doctors deliver care more efficiently? Right. Look, we're really good at diagnosing most diseases, right? Most diseases today, several of them have pretty good treatment options, right? Can be reliably treated. So I don't think you need help AI to help you with the diagnosis or the treatment. But it's all these other things around care, right? Patient education. When you think about chronic diseases, for example, how do you make sure a patient stays in therapy? How do you make sure they receive therapy to begin with? And they stay on therapy, they understand why they need to stay on therapy, how do you check in? How do you monitor? All those things today are done by a human, right? And it's a human being who's making that same phone call 15, 20, 30 times a day. It's a human being who's explaining this is how a monofocal lens is different from a trifocal lens over and over and over again, right? But what if you live in this world where a patient could be educated? You know, they never the patient never felt that there's not enough time. And going back to Manjal's um thought of abundance, the patient feels the doctor's practice has all the time in the world for me. They can answer my questions. They will repeatedly answer my questions. I don't have to feel bad that I'm taking time, right? I can go into this conversation with my doctor better prepared, better informed, so that I have more agency and I can participate in that decision making. I think that is going to be really uh very well received by providers, by surgeons, by physicians, uh, by caregivers, um, anybody in a clinical setting.

SPEAKER_02 14:38

So, so give me an example. And I know this is early innings here, but give me an example uh of Orphea. We're starting it with cataract patients. Tell me why they're, but also give me an example of how a patient would actually utilize this offering.

SPEAKER_00 14:54

Yeah. Uh so we're starting with cataract patients, TJ. Cataract is the most commonly performed surgery, right? Uh globally. There's over 4 million cataract surgeries in the US every year. Uh, there's a lot of dense information that a patient has to receive, they have to process. Um, the cataract lens technology has improved so much over the last five to seven years. Um, it's a lot of information for a patient to take in. Um what is the surgery? What happens to my eye? How long does it take to recover? You're giving me all these options. I have to think about it because it's going to impact how I'm going to see post-surgery and going forward. And there's also some economic implications to the patient, right? This costs X, that costs more. How do I think about these things? What's the value? So there's a lot of information to try to convey and communicate. And the reason we pegged that, we said, okay, this is a well-defined problem. We know that a patient needs to understand all these things or will be better served if they understand these things before they come in to have that conversation and um that visit and the consultation. Um, but look, our ambition and aspiration is really all of ophthalmology, right? So this is just the beach head. I think once we show that we're successful here, um, you know, we've got multiple business cases uh as we think about it and use cases.

SPEAKER_02 16:16

And and maybe, Majal, you you can jump in here to help me understand what the what the patient will actually do uh in order to make this process more seamless.

SPEAKER_01 16:27

I mean, if you think about today, um before you go in for your cataract surgery, right? You're gonna have a pre-op call where you know the ophthalmologist's um nurse is gonna call you and walk you through um the process. You're gonna have a lot of questions. Um, in that process, also there's some education about the different lens options you want, as Manisha said. Patients need a lot more time and energy than really their staff can ever spend. I mean, you're about to take out something from my eye. Like, I got questions, right? You know, and I'm nervous. I mean, you know, any of us have ever had LASIK done. I remember having a ton of questions on LASIK. And then I remember coming out, and actually, in my case, I was lucky, my LASIK surgeon was actually my next door neighbor. Um, but I had so many questions. I walked over to his house that night and I was like, is this normal? Is this okay? If it's burning, I don't know what's going on. And imagine being able to have that on-demand access. And then if you're having an issue, it'll escalate and transfer the call directly. And so I think there's a lot of it's a phone call. Or if they have an app, it can be a you know, a voice. What we found is there's no better way to communicate information and persuade and calm and show empathy than voice. There's something very magical in voice that's just not there in text. So it's a phone call or what I call a VoIP call inside of an app, you know, where it'll just pop up and talk to you. Um, but um either way, it's voice. And that's where the empathy comes. And it's explaining to them what's going on. It's walking them through the questions. It's it's really having the logic to say, oh, you said you're having that issue. Well, are you also having this issue and that issue? Oh, actually, then I need to transfer the call right now to a human to be able to escalate. Um, maybe you are having a post-surgery complication. And so I think I think that's part of it. The other part is do patients always get their surgeries on time? Like cataracts are a degrading condition with a surgical endpoint. How many people go in, see a doc? He says, Oh, you don't need it yet, but you know, if it gets keeps getting disleveled, then come is somebody, is some ophthalmologist calling them every single month and saying, ma'am, tell me about your vision. How's it doing? You know, is it time? Basically trying to determine if it's time. And, you know, if you notice with cataracts, a lot of people afterwards are like, oh my God, I should have done that sooner. I don't know why I waited so long. They're rarely like, yeah, I got it too soon. And so, you know, what a what a novel concept that we can now do what I call longitudinal vigilance. So there's some simple kind of periop ideas that come in and buy time and help our clinicians. But then there's also just vigilance on a variety of different ophthalmology issues that the AI just has the capacity to check in with patients. And the current healthcare model is not, you know, we check in on how you're doing. I mean, we do a little of that, but nobody has the staff to do the level of that that should be done. And but now the AI can do that vigilance, can do that check-in. And so I think, again, there's very simple and straightforward ideas that are kind of within what's being done today. And then there's a set of ideas that you wouldn't have ever done unless you had a large capacity of clinical staff to be able to check in on every last patient you saw who was even earlier in their cataract journey. Um and that's that's I think I think that's where we get to start thinking differently.

SPEAKER_02 20:04

So as this product, similar products are rolled out, more patients, more practices are taking advantage of them. Does does AI ultimately become something that patients interact with and recognize, or does the best AI eventually just kind of disappear into the background and become part of normal patient care that people expect?

SPEAKER_01 20:33

I got an AI the other day and the Taco Bell aligned to order. I was going there for a friend in my defense. Um but yeah, yeah. You know, he just had to have a taco. But um look, I think this is gonna get so ubiquitous. I mean, name a product that a billion people use regularly now that launched only two and a half years ago.

SPEAKER_02 21:01

I'm sorry. That that keeps improving itself on a daily basis.

SPEAKER_01 21:05

On a daily basis, right? And so a weekly basis for sure. And um I think we're all gonna get, I think the banks are rolling it out, the insurance companies are rolling it out. Um, you know, people are very nervous at the moment. I mean, well, patients accept it. There's no question patients will accept it. They're already accepting it. They're accepting it in every other part of their life. And um, as an augmentation of the clinician, they will. But again, I think this is the genius of you know, Manisha and the Baushan Lom team, which is they figured out two elements. One, how to do it in conjunction with the clinicians from day one and how to build this with their uh guidance and support and clinical protocols that they want um enhanced and ones they don't want enhanced. And so I think, you know, I still think that um that is, you know, really a smart idea that they did with that. And I think the second part was that, you know. They recognized that step one was to focus on this clinical pathway of uh, you know, cataracts, build all the use cases to complete that journey and augment the supply, the staffing that an ophthalmologist would have to do. And so starting with the right beach head and I and working together with really um the clinicians in a way that they trust and a way that enhances them is the only way to build these things. Um but I think on the patient side, honestly, the patients are gonna be ready and they're ready, ready, and they're using it in every part of their life today already.

SPEAKER_02 22:41

This is the this is the lightning round before we wrap here. Uh I'm gonna ask you both the same question. Uh would love a very short answer. It never works out this way. Um, but maybe we'll change that up today. Uh first question.

SPEAKER_03 22:58

What is one thing about AI in healthcare that is overhyped? Go ahead, Manisha.

SPEAKER_00 23:10

I don't know. There's one thing. There's, I think, a handful of things perhaps that are overhyped. Look, I'll tell you one thing that I don't see happening. Um you're never gonna replace the doctor. I firmly, at least not in the near-term future, right? Uh you can take away a lot from what they do, but you know, if you need cataract surgery, you're gonna need a surgeon, right? And you're gonna rely on them and their uh wisdom and their judgment and their training, their experience. I think that's gonna be that is going to be hard to replace at least the near term.

SPEAKER_02 23:44

So getting rid of doctors, overhyped.

unknown 23:47

Yeah.

SPEAKER_01 23:48

Yes. I was gonna say, I was gonna say that one, but I said the other one is actually this ambient dictation is not actually overhyped from how well it's working and improving the lives of physicians. But um many folks have told me that they have not seen the ROI. Um and so I think that's uh that's an area that we've seen. Um, you know, they just they've not, it's not resulted in seeing more patients, so to speak. It has resulted in less pajama time, um, which is great. Um, but you know, ideally the person buying the product wants to see both.

SPEAKER_02 24:25

If we were keeping score, you'd be up one on Manisha for a shorter answer, although they were both verbose. So we're we're gonna try to go for one or two sentences. One thing, or we're gonna flip it again here.

SPEAKER_03 24:38

One thing about AI and healthcare that is under hype.

SPEAKER_01 24:44

I'm gonna say what I've been saying all along, abundance. We keep looking at labor replacement, thinking that's where AI is gonna make an impact. It's gonna make an impact doing the things we don't even do today.

SPEAKER_03 24:59

Manisha? I agree with Munjal. That wow, you know what?

SPEAKER_00 25:04

I'm trying to be.

SPEAKER_02 25:05

Yeah, you actually win that round because that that that was one sentence. Well done. All right, so uh the the last question, and I'll let you guys go. Ten years from now, what is one thing that you hope AI has not changed about healthcare?

SPEAKER_00 25:27

I would say the human connection is so important, right? You're seeking help because and you're vulnerable and there is something to be said about that human interaction. It would be very tragic if we lost that.

SPEAKER_03 25:49

I think that um we want to ensure that we've been able to keep kind of the truth seeking.

SPEAKER_01 26:04

You know, meaning today we trust our clinicians because on balance, they bring us the evidence-based medicine. And I think that we want to ensure the AIs don't confuse the issue or don't bring us misinformation. And I think that's very important. Um, it took us a hundred years to get out of snake oil and get into evidence-based medicine. And I think we do not want to backslide on that. And there is a little bit of risk with AI of that if we don't constrain it right, test it right, train it right. And I think we've all figured out how to do it, but um again, it took us so long to get to this evidence-based protocol. It's the last thing I want to give up.

SPEAKER_02 26:52

Great answer. Manisha, Munjal, thank you both for uh joining us. And uh yeah, great conversation. Thank you for having me.