Welcome to The Edge of Healthcare, your premier destination for insightful discussions and actionable insights. In each episode, we dive deep into conversations with industry leaders, exploring the dynamic landscape of healthcare. From overcoming hurdles to embracing breakthroughs, join us as we discover firsthand the strategies and experiences of healthcare trailblazers. Whether it’s payer and health system leaders or innovative solutions, we’re here to empower you with knowledge that drives real change in the industry. Don’t just listen—be part of the transformation.

About This Episode

The future of voice AI is here, promising more efficient and intuitive user experiences, especially within healthcare.

In this episode, Israel Krush, CEO at Hyro, dives into the essence of why our current digital helpers often miss the mark and how Hyro is part of a burgeoning revolution that’s reshaping our expectations of technology. Martin and Israel explore not just the frustration that comes with customer service encounters, but also the incredible potential and the complexities behind making conversational AI that can truly understand and respond to us in natural language. Israel talks about Responsible AI in practice, delve into the upcoming regulations, and consider how Hyro is addressing the pressing need for efficiency and improved patient experiences in healthcare call centers.

So, whether you’re a healthcare professional, tech enthusiast, or just intrigued by the seamless blend of AI and everyday life, you won’t want to miss this conversation.

Read the transcript below and subscribe to The Edge of Healthcare on YouTube.

Martin Cody: Welcome to the Edge of Healthcare where the pulse of innovation meets the heartbeat of leadership. I’m Martin Cody, your guide through riveting conversations with the trailblazers of healthcare. Tune in to gain exclusive access to strategies, experiences, and groundbreaking solutions from influential payer and health system leaders. This isn’t just a podcast, it’s your VIP ticket to the minds shaping the future of healthcare right now. Buckle up, subscribe, and get ready to ride to the edge of healthcare, where lessons from leaders are ready for you to use today.

Martin Cody: All right. Hello again, everybody, and welcome to another episode of The Edge of Healthcare lessons from leaders to use today. My name is Martin Cody, Senior Vice President of Sales and Marketing for Madaket Health. And your host, as we get to interview some of the leaders from healthcare today and learn how they got to where they are, where they’re going, and some nuggets that we can use today to fulfill ourselves, advance our careers, and make certain we avoid any mistakes. Today I’m joined by Israel Krush from Hyro AI Israel. It is fantastic to meet you and I’m looking forward to jumping in. Thanks so much for joining.

Israel Krush: Thank you for inviting me, Martin.

Martin Cody: My pleasure. And the first thing I want to call out to the listeners and viewers is your current location. Where are you talking from today?

Israel Krush: Yeah, I’m talking today from Tel Aviv, Israel.

Martin Cody: Tel Aviv, Israel. And is that where the company is headquartered?

Israel Krush: No. So actually, the company is headquartered in New York City. So, I started the company in New York. Will probably talk about that. However, as I realized that I needed to build the R&D center, so the product and engineering and do the initial product iteration of a company, and I moved back to Israel, that’s where a lot of my network is. But I really like it when people ask me, where am I based? I’m saying on an airplane, so wherever I’m needed.

Martin Cody: So, I think that’s a good philosophy, because if you if anyone listening has some entrepreneurial desires, you will spend a lot of time traveling, even in the remote world that we now live in today, which has accelerated in the last 4 or 5 years because of the pandemic. But you will spend a lot of time at an airport, so keep that in mind. So, let’s talk a little bit about Hyro. And it’s an AI company in artificial intelligence, language learning models, machine learning the stuff easily by a factor of ten. The hottest subject not only in healthcare but across every single industry seemingly that exists. So, how did you get into AI?

Israel Krush: My background I’m a computer scientist, so I was a software engineer. I was a product manager, and I studied majored in computer science and statistics. Before, it was called machine learning. And I did some of that really purely because I’m a geek, so I like statistics. And when I think in retrospect about the various roles that I’ve done, most of them in the Israeli tech ecosystem, the common thread is dealing with human-computer interaction or HCI. So, it can be today with large language models and natural language interfaces, Tyro but I also done it in Computer Vision Company that lets you try on clothes virtually. A company called The Kid That Was acquired by Walmart. That’s really like the common thread of everything that I’ve done. And honestly, before gen AI or large language models were a hot topic. Maybe that’s part of like why we’ve even started Hyro. Rom, my co-founder, and I met in New York City and I did my MBA at Cornell University. He did his master’s in computer science. And as two Israelis that got into the US, we’re talking about 2017.

Martin Cody: Okay.

Israel Krush: We got exposed to the voice space, the entrance of Alexa and Google Home, and all of these nice devices. Long story short, we got excited about the opportunity. Purchased a few devices, tried them on, got disappointed but then basically got with two hypotheses that got us to start Hyro. One is that natural language interfaces are going to be the dominant interface when it comes to human-computer interactions. HCI, like where I’ve been all my life. And two is that for large organizations such as healthcare systems with lots of data from various data sources, it’s going to be very hard to deploy and maintain these natural language interfaces. And so I think that these two hypotheses really made us double down on what is known today as conversational AI, so how do we create chatbots or voice AI assistants so AI assistants that understand human language.

Martin Cody: I want to go back a little bit to the Cornell days. And even before that when you talk about HCI. So, in lay terms, what does HCI mean? Human-computer interaction.

Israel Krush: How we as humans interact with technology, and let’s think about the mouse or the keyboard. We know that technology can’t understand human language. They understand bits, so zeros and ones. We’ve created a bunch of graphical user interfaces and some devices to control these graphical user interfaces in order to communicate with technology. Let’s take healthcare as an example. Healthcare is actually a bit behind they wanted to sell. If I want to schedule an appointment with a physician today for most health systems, I need to call them. So, that’s my interaction with technology or with a call center. Lately, you are able to schedule some of these appointments online with forms that you need to fill out so you can register as a patient and even sometimes schedule with a PC without seeing them in advance as a new patient, which is advanced today, but it’s still a graphical user, right?

Martin Cody: So, then the GUI, the graphical user interface is a subset of HCI. It’s a component of it.

Israel Krush: Correct.

Martin Cody: Okay. And then from there, you talked about being a geek, a math person, and loving statistics. Those are three areas where you and I completely differ. I’m not a geek. I loathe math and statistics was not a fun course. But at Cornell, you met your co-founder and both of you are getting yourselves exposed to some of the early language HCI devices. And you were disappointed. Were you disappointed in the GUI? Were you disappointed in the capabilities? What were you disappointed?

Israel Krush: I think that one realization when it works, it’s magical. I’ve used the Waze, which is GPS software to navigate, and you’ll notice that there is a mic button. So, you have a GUI. You can type the address that you want to go to, but you can also type, click the mic button, and just say the address. The first time that I did this, it took me time because back then, I wasn’t used to doing things like this. I wasn’t ready to interact with technology using my voice. And he said, I don’t know, 415 Madison Avenue, it got it and immediately started navigating. That was a magical moment for me. And since then, I’ve never used the keyboard in Waze. I always typed the mic button because I wanted to start driving. So, like a hands-free type of interaction while in the car, that’s just one example that got my attention to why voice is so powerful. Now, the problem with voice is that you don’t have the GUI, you don’t have the graphical element most of the time. So, think about the voice AI assistant in a call center. You can’t fill out the form. You can’t look at a screen. You’re just speaking with an AI assistant. And what got us disappointed is that, one, it didn’t understand us more often than not. And two, the latency. So, the time it took the AI to actually reply with something back was taking too long. And three, the use cases were very simplistic and by simplistic use cases. Play this song. Set up an alarm clock. What’s the name of George W. Bush’s father? I don’t know. Generic questions. If I wanted to schedule an appointment with a physician with Alexa, I couldn’t do that.

Martin Cody: Okay, so that I can understand that frustration. Now, there is one thing we do have in common. I am today perpetually frustrated with chatbots, with the assistants, so to speak, and I am one of those individuals, whether I’m calling American Airlines, whether I’m calling a bank, whether I’m calling anybody. Correct. All I do is say agent as fast as I can, because you get, first of all, the organizations and the businesses, much to the consumer’s chagrin, the first 45 seconds of welcome to such and such organization is an infomercial. It says if you’d like faster assistance, please go to the web. Most of us have already been to the web and didn’t get the answer that we were looking for, so the chatbot is already behind. But you’re correct. I am saying the agent hitting zero. Then you get put back into the queue and then you come back to the same thing. So, why is it that chatbots today, and the virtual voice agents, if you will, are so far behind seemingly the user experience?

Israel Krush: So, we really are in the midst of a revolution in that regard and the sentiment that you have towards chatbots and voice AI assistants, that’s the wide sentiment I’d say that most consumers have today, including myself. I’m a Delta guy and I call Delta. I want the agent, and I don’t want to geek out too much about the technology. But really, the latest buzz around GPT, right? So, chat GPT basically took everything that we, geeks, like in the world of natural language. No. And so like the advancement is just a completely different level. And there is a good reason behind that. It’s not good marketing. It’s because we are in the midst of an AI revolution.

Martin Cody: What do you mean by that?

Israel Krush: Like this inflection point where we can really communicate with technology using our voice, using our native language, and these chatbots and voice AI assistants actually understand what we want to do and being able to help us out.

Martin Cody: Okay. Learning, if you will, because of the billions of computations that are going on behind the scenes that, oh, this person’s asking this, they don’t necessarily want to see their mileage account on how many miles they have on the air. They want to actually do something a little bit more complex, and then be able to solve some of the issues for you without much effort on your part.

Israel Krush: Correct. It’s like this combination of a couple of technologies getting things right. It started with like cloud and computing, being able to compute lots of data and being able to build this very large language model for a reason. They are called large because we had machine learning models for like.

Martin Cody: 50 plus years.

Israel Krush: But yeah, so we got to this point where it actually works and it works well. There’s a lot to do. In terms of scheduling an appointment with a physician, it’s not only about understanding that’s what you want to do, but you actually need to be integrated into the workflow, into the electronic health record, Dr. Epic, or whatever you’re using. So, that part as well. But from an understanding perspective, we’re really like in a good place today. And I said, I’m not going to gig, but a lot of the frustration from the call in the previous era of chatbots and voice AI assistants was because as a technology vendor or partner, I needed to think about all of the various intents that users might have and predefine them and create, like this very rigid conversational trees of if user says this, then reply with that. And that’s not how we humans talk or think. Our thinking isn’t linear. I started answering you, thinking that I’m going to go this direction, and I ended like in this direction. And now with the large language models, this changes, this context switches are possible.

Martin Cody: I think you’re right. I’m looking forward to the revolution, if you will. I want to switch gears a little bit and talk about Hyro. Now, as an entrepreneur, the first thing you recognize is capital needs from that standpoint. And we’re coming up on about a year now since you had a B round where you raised $20 million. I’d be curious for anybody out there listening that is also thinking of raising some capital. What were your top two takeaways from that experience of raising capital?

Israel Krush: Persistence prevails. That’s one of the core values here at Hyro. And that’s like really special to me. Like I’ve insisted on that people didn’t like the word persistence but it’s really about persistence. It’s understanding that like a lot of different hard things that you want to achieve in life, most of us don’t get it right in our first shot, not in the second one, not in the third one. So, when you think about fundraising, you should think about it as a sales process. And with that, you need to create a process. You need to understand where you are in this process, and you need to have a lot of different prospects. So, one of the cliches is it’s a numbers game. That’s the advice that I got when I just started the startup. And I understand that this is the reality of it, because different investors look at different stuff, obviously, like you have the commonalities and you need to understand how your growth rate looks like and your market and your competition. So, there are a lot of like boilerplate and templates. These are the areas that you need to focus on. But I think that entrepreneurs or people who want to start a company or to raise capital need to understand that this is like a sales process, and you need to be very vigilant about that, and you need to meet a lot of investors to make it work. It’s not like the first investor that you’re going to meet is going to hand you a check. And that persistence prevails in that regard.

Martin Cody: And I think there’s a core theme there as it relates to overcoming adversity, because you’re going to, especially in the fundraising rounds, you’re going to get a lot of no’s, like you said, and you’re not going to you’re not going to have someone send you or give you a check on the first meeting. So, how do you, Israel, what are some of your core go-to habits and fundamentals to overcome adversity, and you’re going to get knocked down and how do you get up again?

Israel Krush: Yeah, I think that one, it’s a mindset that you need to develop. And this mindset says it’s not a rejection or a bad thing as long as I’ve learned something. And how do I take this no and not become frustrated by the fact that someone told me no, but like, why did he say no? What can I learn from this no? And more often than not, I see that my most successful meetings, are meetings where people ask me questions and I write down every question that they ask me. And then I think about. So, for example, if I’m using like a pitch deck thinking to myself. Well, how do I iterate on this pitch deck? So, the next investor that I meet don’t question this hypothesis that I have, or don’t ask me a question about this slide, because the story that I’ve created makes a lot of sense, and a lot of the times it’s really about the narrative. You need the stats, but also the narrative. So, I’d say it’s a mindset. Don’t get excited by the no or don’t get offended by the no and thinking, oh, he’s stupid. He don’t want to invest in my business, but instead take it as a learning opportunity and really iterate over the product. So, the next time would be different.


Martin Cody: That may come easier to you as a data person and a stats person, because you’re right, the no is just a piece of data. And so now how can I iterate on that piece of data so that the next time I get a yes or a maybe or something like that makes perfect sense. Now in the description of your organization, you have the phrase responsible AI. What is responsible AI?


Israel Krush: I am so happy that you asked this question because I’m getting itchy from buzzwords. So, like when we say responsible AI or Gen AI, what does it mean? We’ve learned how to deploy and maintain conversational AI interfaces. So, these chatbots and voice AI assistants for industries such as healthcare, which are highly regulated, are very different than deploying them, for example, on an e-commerce platform. And there are a bunch of reasons for that. One is it’s regulated. So, you need to comply, you need to take compliance into account. And compliance is a big pillar of what we call a responsible AI. And when we say compliance, we don’t only think about HIPAA compliance and GDPR and SOC2 and all of these compliance standards that we know today but also understand that regulation of AI is coming faster than people think. You know, there was the Biden executive order and the EU already started pushing for it. And I think one of the states, Atlanta, maybe is now passing like a first bill in terms of identifying yourself as an AI assistant and answering an AI. Yes, like in a clear way. So, I think that you need to prepare also for future regulation of AI. You need to make sure that whatever work you do today, you’re not going to need to change it in six months time because of new regulation. So, that’s the compliance piece. Now besides, you need to have a lot of explainability. And explainability is something that is, maybe not that straightforward, because large language models, like other machine learning models, are black boxes. So, in the case of a chatbot, the patient asks something and it gets answers.

Israel Krush: And what happens inside sometimes is a black box. Black box basically. How do you give explainability to the patients, to the physicians, to the health systems? That’s the difference between e-commerce and healthcare in that case, that the cost of mistakes is so big and so different. So, if you’re looking for a blue shirt and I’m giving you a black shirt, so what, like maybe you would be disappointed with the chatbot. But what happened? But if you’re looking for a specific physician or reschedule an appointment in a time that wasn’t really available, that’s a big issue. So explainability is the second pillar of what we call responsible AI. And finally, control. I’d say something that is at least I believe in, but is a bit controversial and Gen AI, generative AI, everyone is talking about that. And the generative piece is not always a feature.

Martin Cody: Correct.

Israel Krush: Sometimes it’s a bug. In an emergency situation when you might feel very unwell. And we want the system or the agent or whoever talks with you to send you to the ER, you don’t want the AI to generate and hallucinate and make up stuff. You want the same answer each and every time. So, you want a controlled response. You might want to control some of the data sources that are being trained to train the model, the prioritization between these data sources, and understanding the bias of the model. So, control is another big aspect that is very important for regulated industries such as healthcare. So that’s the third pillar. So, just in sum, we’re talking about compliance, explainability and control. And you need these three elements for what we call responsible AI.

Martin Cody: No, it makes perfect sense. One of the areas that is interesting because you talk about call centers and efficiency from the healthcare system or the hospital. And it is something in my day-to-day conversations with helping payers with helping health systems, large provider organizations, everyone, almost 100% of the time is indicating that their call center volume has spiked their provider. Experience or provider satisfaction is declining. Their members or their patient satisfaction with their organization is also declining, and that can impact revenue. It’s an area that we address through a lot of the financial understanding of the health system to help mitigate revenue leakage, and to help enhance provider satisfaction. But a lot of it is related to the call center, and increased calls to the call center are usually symptoms of larger problems. How does Hyro AI help the health system in the call center efficiency?

Israel Krush: Yeah, absolutely. I think that we think about it as a multi-layer type of approach, and that’s based on hundreds of conversations that we had with CIOs and VPs of patient access and patient engagement, and so on. First, like call centers being one of the biggest cost centers for these large health organizations. They’re coming from, how do we become more operationally efficient? And efficiency is a big thing. But to your point, they don’t want the patients to yell at the agent immediately, so it’s not the expense. Until now, it was at the expense of the patient experience or member experience. Now, it’s not; we want to become more operationally efficient, but we also want the patient and member experience to increase. So, 24/7, being able to schedule appointments, for example, for many channels via text, mobile, web, call center, and so on and so forth. So, the multi-layered approach here is one, how do we make patients or members don’t call the call center and use their favorite digital channels. So, creating these digital front doors chatbots that actually work on websites and mobile applications and SMS that they can use now would deflect some of the conversations. But then people will still call your call center, and then after they call the call center, how can we direct them back to the digital channels? Maybe you called because you want your password to be recited in your patient portal.

Israel Krush: Believe it or not, that’s one of the highest call volumes that health systems get, both from patients and employees calling about password reset. This is not something that you should do on the phone. This is exactly how we send you a link via SMS to the page where you can reset the password. We might actually hold to the line and guide you through it as an AI, but you don’t need to do it over the phone. So, that’s the erecting them back. That’s the second layer. And finally people will still sometimes, depending on their issue, want to complete the action end to end via using their voice. And that’s where our voice AI assistants kick in. So today, you can actually schedule appointments and do patient registration, verify appointments, canceling but also prescription refills over the phone end-to-end. Talking with a voice AI assistant. They might be not as good as humans today, but you know what? They’re going to answer to you in three seconds. So, no wait time, 24/7, so you can call at midnight, and they’re never going to be frustrated or angry at you, and they will try to help you accomplish your task. So, these are the three ways.

Martin Cody: That’s a very good clarity on that because we’re talking about a similar problem from different angles because most of the frustration that we experience at Medicare as it relates to increased call center volume is because the provider directory, the physician directory, is inaccurate. It’s telling me that Israel is accepting new patients on the website when, in fact, Israel is not accepting new patients. When I called to book an appointment with you, I was frustrated because you’re not accepting new patients. So, that’s an easy clerical administrative fix. But to your point, the staff is burnt out. Staffing shortages are perpetual in healthcare. And so, working together, we have to figure out ways to make it easier for the health system to do these sorts of things because this is fundamental blocking and tackling. This isn’t open-heart surgery type of stuff. So, I love the way you’re approaching it and what you’re trying to do. We just completed Q1 of 2024. So, when you get to the end of 2024, and you look back, how will you gauge whether or not this year was successful?

Israel Krush: Knock on wood; I can tell you right now that this year is going to be successful. And I’ll tell you why. I think that in 2023, the market was pretty much closed, like the general market. Forget for a second healthcare. Healthcare was even closer. The only thing that gave us some hope is this AI buzz and talks about how we can be so much more efficient about it, and that it’s going to revive the economy. The experiences that we’re going to get are going to be better. Companies would be so much more efficient. But in 2023, it was really about this is interesting. Okay, I want to play with it a bit. And it was more of an experimentation year if you want. In 2024, what we see in healthcare, but generally speaking, also, is that these large organizations have made a decision to use Gen AI, the conversational AI. It’s not like they’re going to use it in a clinical setting, like immediately or patient-facing setting. It is like healthcare, the crawl, walk, run approach, right? But they’re definitely going to implement it. And they put some of the processes in place to make it happen. One theme that we’re seeing is the emergence of AI governance committees. So, health systems are creating AI governance committees from the provider organization, the patient organization, and the financial clinical staff to evaluate AI and think about how can we start deploying that. And so to me, I’d say back to what we talked about, the call center. I feel that 2024 is going to be a successful year if, one, we’re able to help these health systems increase their operational efficiencies. But in parallel two, we’re creating better patient journeys and better experiences. So, it’s not a trade-off here, but actually, it is sure.

Martin Cody: It’s not an either, or it’s a both. And I completely agree with you. Healthcare seems to have a crawl-walk-run mentality. And I would caution healthcare. We’re spending way too much time in the crawl area of this. We need to start walking and running. All right. I’m going to switch gears to a favorite segment where we’re going to do word association. So I’m going to mention a word or two words, and you tell me what pops into your head. It could either be a sentence, it could be a word. Whatever it is that pops into your head. Ready?

Israel Krush: Yes.

Martin Cody: Staffing shortages.

Israel Krush: Big problem.

Martin Cody: Big problem. Okay. Interoperability.

Israel Krush: Amazing opportunity.

Martin Cody: See, that’s an entrepreneurial mindset right there. Digital health.

Israel Krush: On the rise.

Martin Cody: What does digital health mean to you in a sentence or two?

Israel Krush: I think that it means money becomes digital, right? Most of us don’t carry cash anymore. Not even credit cards. Like it’s all in my mobile phone or with system numbers on a website, I think that when it comes to our health, not necessarily how do we get care? But when it comes to our health, seeing the data, being able to track it better, being able to approach it better, and also being able to navigate the healthcare ecosystem better in a digital way, not only digital is really a big part of that.

Martin Cody: All right. And our last one, Maccabi Tel Aviv F.C.

Israel Krush: I don’t know, probably the best Israeli sports team. But as Israelis, we’re not that good at sports.

Martin Cody: So, he said it, not me. We’ll leave it there. Israel, it’s been an absolute pleasure getting to know you and getting to know Hyro AI. How do folks get a hold of you and investigate the company a little more?

Israel Krush: Yes, absolutely. So, first off, feel free to go to our website. It’s Hyro.AI, so easy to remember. And also feel free to reach out to me. I’m still in a position. I’ll try to answer all of your emails. So, it’s Israel@Hyro.ai, and if you’re a healthcare system or health plan and you want to learn more, you can see some of the stuff on the website and request a demo, and some of our folks are going to be in touch.

Martin Cody: Awesome. Fantastic. Thank you so much for investing a little time with us. And remember, entrepreneurs and for anybody in healthcare, it’s persistence. We have to continue to get knocked down and get up again. And let’s finally move out of that crawl phase and start walking towards some of these innovations and embracing them more. Thank you so much for the time. Great to meet you. I look forward to talking with you more.

Israel Krush: Thank you, Martin, for a great conversation.

Martin Cody: Talk to you soon.

Martin Cody: Thanks for diving into the edge of healthcare with us today. I hope these insights will fuel your journey in healthcare leadership. For more details, show notes, and ways to stay plugged into the conversation, head over to MadaketHealth.com. Until next time, stay ahead of the curve with the Edge of Healthcare, where lessons from leaders are always within reach. Take care of yourselves, and keep pushing the boundaries of healthcare innovation.

Listen to the Latest Episodes