This Week in Startups - Next Unicorns: Building the future of healthcare communication with Abridge CEO Shiv Rao | E1784

Episode Date: July 26, 2023

This Week in Startups is brought to you by… Crowdbotics. Great ideas can change the world, and Crowdbotics is the fastest way to turn those ideas into code. Get a free scoping session for your next big app idea at crowdbotics.com/twist Vanta. Compliance and security shouldn't be a deal-breaker for startups to win new business. Vanta makes it easy for companies to get a SOC 2 report fast. TWiST listeners can get $1,000 off for a limited time at vanta.com/twist LinkedIn Marketing. To redeem a $100 LinkedIn ad credit and launch your first campaign, go to linkedin.com/thisweekinstartups * Today’s show: Abridge CEO Shiv Rao joins Jason to discuss how his AI platform saves doctors time by automating consultation note-taking(8:29), the future of the medical field(37:23), and much more! * Time stamps: (00:00) Abridge CEO and Co-Founder Dr. Shiv Rao joins Jason (3:22) Abridge's journey and tailwinds in the industry (6:43) The sudden advancement in AI tech (8:29) The importance of conversation in healthcare (12:00) Crowdbotics - Get a free scoping session for your next big app idea at crowdbotics.com/twist (13:29) Dr. Shiv demos Abridge (21:49) Reinforcement training and building trustworthy models (23:10) Use of this product in the field and feedback from clinics (26:53) Vanta - Get $1000 off your SOC 2 at https://vanta.com/twist (28:00) Time saved using Abridge (30:30) How patients interact with Abridge and data privacy (35:56) LinkedIn Marketing - Get a $100 LinkedIn ad credit at https://linkedin.com/thisweekinstartups (37:23) The future of the medical field (49:11) The pace of healthcare today * Check out Abridge: https://www.abridge.com/ Follow Dr. Shiv: https://twitter.com/shivdevrao * Read LAUNCH Fund 4 Deal Memo: https://www.launch.co/four Apply for Funding: https://www.launch.co/apply Buy ANGEL: https://www.angelthebook.com Great recent interviews: Steve Huffman, Brian Chesky, Aaron Levie, Sophia Amoruso, Reid Hoffman, Frank Slootman, Billy McFarland, PrayingForExits, Jenny Lefcourt Check out Jason’s suite of newsletters: https://substack.com/@calacanis * Follow Jason: Twitter: https://twitter.com/jason Instagram: https://www.instagram.com/jason LinkedIn: https://www.linkedin.com/in/jasoncalacanis * Follow TWiST: Substack: https://twistartups.substack.com Twitter: https://twitter.com/TWiStartups YouTube: https://www.youtube.com/thisweekin * Subscribe to the Founder University Podcast: https://www.founder.university/podcast

That's the thing that's blowing my mind more than anything else in 2023 is that healthcare is moving at the pace that it is with this technology. Is it really moving faster finally? It is moving so fast. Why did they go from not moving fast to moving fast? Is it just because it's so broken and they're so exhausted and they're finally ready to capitulate and use the technology? Yeah, we just pushed way past the point. This weekend startups is brought to you by CrowdBotics. Great ideas can change the world.

And CrowdBotics is the fastest way to turn those ideas into code. Get a free scoping session for your next big app idea at CrowdBotics.com slash twist. Vanta. Compliance and security shouldn't be a deal breaker for startups to win new business. Vanta makes it easy for companies to get a SOC2 report fast. Twist listeners can get $1,000 off for a limited time at vanta.com slash twist. And LinkedIn Marketing. To redeem a free $100 LinkedIn ad credit and launch your first campaign, go to LinkedIn.com slash next unicorns.

Hey, everybody, welcome back to the program. Today, another amazing episode of our The Next Unicorn series. You want to see all the startups we featured in this over the years this week in startups. com slash next unicorns. What is the idea behind the show? We find promising companies that we think could change the world. And just last week, I interviewed Andy Beck. He's building Path AI, if you remember that.

It helps make pathologists superhuman, right? They get a 3D model. They look through the microscope. And it makes it easier to diagnose things like cancer, maybe more accurately, will drive treatment and hopefully extend people's lives and suffer or lower suffering. Great, great mission. Today, another AI founder in the healthcare space. Dr. Shiv Rao is a cardiologist teacher, former corporate VC, don't hold it against them.

and the CEO co-founder of A Bridge. Now, A Bridge, you ask, what is that amazing AI platform? Well, it takes verbal conversations between doctors and patients and it converts them into what the industry calls, I believe, soap notes.

Soap notes stand for subjective, objective, assessment, and plan. Basically, concept is to save doctors and healthcare professionals a ton of time, maybe reduce the errors, and again, reduce suffering, increase lifespan, so that the human species can enjoy more of the limited precious time we all have on the planet.

Welcome to the show, Shiv. Thanks so much, Jason. How has the last five years been with the ascension of the show's secession? And everybody telling you, Shiv, like on the show. I love that character. That must be incredibly annoying. I love the show. I just sort of nod politely and wish they were likening me to the prison knife.

but, you know, probably it's all the same thing. I think that was the inspiration for her character, the nameship was to that. I mean, in the, what do they call it, prestige TV era. Yeah. From Sopranos to Scession, I gotta think it's in my top three of all time. Yeah, totally.

I mean, it's the Sopranos and Scession, and then I got to fight for my number three spot. Totally. Yeah, it's just incredible. So how long have you been working on the company? Yeah. And it's, for people who want to check out the website, it's abridge.com. Yep.

A, B, R-R-I-D-G. So we started a bridge in 2018. And in 2019, I'd say, like, it was all in. Like, we were going at it. And in 2018, when we started, I was juggling a couple different positions. I was sort of unwinding from that corporate VC role at a large health system. And we were also getting our research chops going inside the company, because we're an AI native company. So we're dealing with lots of data. We have a lot of AI researchers,

a lot of annotations that needed to happen when we first started the company in order to have a straight face coming out with a product that truly was powered by proprietary technology. So 2019 is where things got started, started to get very real. And fast forward, though, to 2023, I mean, it's insane. Like, I'd say the last six months for us, maybe even eight months for us has just been like game-changing. It's just been sort of like slow, slow, slow. And then suddenly it doesn't feel like tailwinds. It feels like a tornado. And there's two tailwinds going on for us right now. One tailwind is just the industry. Like the moment in healthcare right now is such that clinician burnout, staffing shortages, labor shortages are one of the first priorities for every single health

system executive out there. They need to solve that problem. It's mission critical. And, And the other tailwind is generative AI because everybody understands this technology now. And not only do they understand it, they use it in their daily lives. So they're open to it. They want to try it and they believe in it. And that's not where we were even last year. Like in November of 2020, we held a dinner for a bunch of VIPs in healthcare. And it was teaching them about generative AI.

And this was two months before Jack CPP came out. And so when it busted open, everyone called us like, oh, we get it now. like what you were talking about a couple months ago, we're ready, let's try this. And you're like what I've been working on for five years. Yes, it is interesting how something can just break the dam. And then everybody's like, they go from being like, I would never even ride in an electric car. And then they're like, you know, when can I get one? Right.

And what's the wait time? And how many can I buy? You know, we just, it's these revolutions are like, slow, slow, slow, slow, slow, slow, low and then boom, they just arrived. So one of the things I've noticed is the ability to just understand what people are saying and get it accurate has changed dramatically in just two or three years. I was one of those guys who kept trying drag and dictate 20 years ago. I remember attorneys 30 years ago having me install this stuff with headphones and training for two or three hours to get and it would have them save these words over and over again. And now, you know, I start

looking at, I was looking at the Zoom transcript. And you can watch the Zoom transcript in real time during a call. I didn't know you could do that. And I was interviewing Darmesh, the CEO, or the co-founder and CTO of PubSpot. And I was watching our discussion in a live transcript. And I was like, it's completely accurate now. Explain to the audience how it went from being, you got to train this thing for three hours to be a disaster.

Yeah. Or less of a disaster. Siri can't call your spouse or get the, you know, put the Beatles white album on for you. To all of a sudden, now it's doing transcripts perfectly. And it's not just doing them really well now. It's also doing them multilingual. Like two days ago, actually two weeks ago, a federally qualified health center. So a health system center that is caring for a lot of Medicaid patients,

patients from many different backgrounds, social and cultural. And they asked us, they actually challenged us if we'd be willing to let them come on a demo and allow them to actually run the demo, which I'd love to go through with you. Yeah, absolutely. We'll do it right now. Yeah, and speak in multiple languages. Wow. So they came in and spoke in Brazilian Portuguese and Haitian Creel, and we still, and English,

and we still created one English note at the end of this and structured all the data and did all the other sort of manual tasks that clinicians have to do and needless to say, we're still talking to them. Yeah, I mean, that's quite an eclectic collection of dialects there. No Farsi or Iroquois or you didn't want to go back to some Aztec, you know, ancient Egyptian language we've never actually heard before.

Yeah, I mean, it's, Pretty crazy. We're going to literally be able to take languages that no human has ever heard before and probably figure them out based on all other dialects. So let's do a little demo here. For those of you who are not watching, what the hell are you doing? Go to YouTube.com such this weekend and sign up and watch these incredible demos.

Or, well, sports cats, if you'll describe what's on the screen, if you're driving. Totally. Please do not watch YouTube when you're driving. Totally. Unless you're in cruise. If you're in cruise, in the backseat of a cruise or a waymoh, you can do that. So I can share a little bit like a couple lines of like high level context before to just set the stage for the demo. Everything that we're building at the company is really based on a thesis that when you think about it, health care is all about people and they're having conversations.

So they could be having conversations like this over video, telemedicine. They could be having conversations over the phone or call centers. And certainly most of health care happens in person. And we're Omnichannel. We have a Swiss Army knife approach. we can be a part of all those conversations wherever they are. And that's where our technology is going to structure and summarize them. And so when we think about these conversations, they're upstream of a lot of processes

in health care, a lot of potential. They're upstream of all the diagnostics we order as clinicians for patients, all the therapeutics we also prescribe. They're upstream of clinical trial recruitment, of coding for all things revenue cycle to keep the lights on and risk adjustment in the sort of value-based care world, which is a whole other paradigm of how you can create and capture value as a provider in the United States. And then they're also upstream of the most important things, experiences and outcomes.

And so when I think about experiences and outcomes, so you mentioned I'm a doctor. So I'm the founder and CEO, but I'm also a practicing cardiologist. I'll see a handful of patients every month in my free time. So I'll take the shift nobody wants on a weekend. And on Memorial Day weekend, I signed up to take cardiology call. So I was in the hospital and in some clinics, and I'd walk from room to room to room. And I get to use a bridge. That's a big part of why I do it.

And I just hit the button, I hit record, and then I'm capturing a conversation with a patient. And we might talk about, you know, the Pittsburgh Steelers for a while because there's a lot of Carnegie Mellon DNA in our company and we're distributed, but we got a lot of people here in Pittsburgh. Yep. And that's fine. We could talk about politics, you know, the NFL draft, the weather out. outside, anything we want in what... A bedside manner is always good to build rapport with the patient, right?

It's a doctor trained to do that, actually? Yeah, well, yeah, we are. Third year of medical school. They sort of try to teach you. This is how you can build trust with a patient, because at the end of the day, that's going to have the biggest impact, or one of the biggest impacts on their ability to understand and follow through, and be the healthiest versions of themselves. So we just have this normal organic conversation.

I hit stop. I swivel my chair. The note's ready. Everything's in the record. It's all good to go. And before a bridge, what I had to do is take chicken scratch on a piece of paper, like tall guy in the Mets Hat. And that means Saturday morning, I'm reading Tall Guy in the Mets Hat. And I have to remember who was that guy?

Like, who is that? And I'm spending hours perseverating over this and the hospital system is losing revenues because it's too late to bill at that point in time. And it's burning me out. And let alone accuracy and the fear of inaccuracy. Oh, my God. what if my chicken scratch, I remember it wrong, or I, you know, the memory is just such a fragile and unique thing in humans that, you know, you can juxtapose, right? And you don't even know you did it. Exactly. Yeah, exactly. All right, we all know the one thing that separates

great startups from the good ones is product velocity. What does it mean? Product velocity. Fancy term, right? Here you got your product and you have velocity. Speed. The speed in which your product improves. So can you ship updates? Can you you release new features, can you do bug fixes? Can you iterate on the interface? Can you solve problems for your customers? And can you do it quickly? Because you're not alone. You have competitors and your customers have choices. They may solve their problems by writing their own custom code, or they might use your solution. This is what startups are about. How fast can you get that product velocity going? And so, you know, how do you supercharge it? Everybody says, okay, yeah, we want to go

faster, but you've got to go faster intelligently. And CrowdBotics is going to help you do that. They're your CTO as a service. Basically, they provide you with the most optimal architecture to get your product to market as fast as possible. You'll have access to an on-demand product manager and developer talent, and they will help get your app into production 10 times faster than conventional development. CrowdBotics can work with your in-house dev team, or you can just have them work independently. And you own all the IP, you own all the source code. Let the folks at CrowdBotics supercharged or product velocity today, no more waiting. Get a free build plan at crowdbotics.com slash twist.

That's a 499 value just for the Twist listeners. You get that for free. That's C-R-O-W-D, B-O-T-I-C-S dot com slash twist for a free build plan. So I'm going to share my screen. This is how I used it myself on Memorial Day weekend. I think you can see my phone. Yeah, beautiful app. Awesome.

Yeah. So this is what I use. I'd just open this app up on my phone. I'd walk from room to room to room. And so if you're my patient, I'd say, hey, Jason, do you mind if I use this no-taking solution? It's going to help me focus on you instead of- Not mind at all, Doc.

Awesome. Awesome. Cool. So now hit the button. And now we can just sort of do some improv. So, hey, Jason, tell me about your shortness of breath. I heard you having some shortness of breath. Yeah, you know, it's just the last week or so.

I found coming up and down the stairs in the house, I just tried to feel shortness of breath, a little bit of dizziness when I got to the top. the stairs, but I sat down. I took 10 deep breaths and I was fine. But then when I woke up in the morning, I kind of tripped a little bit because I was so short of breath. And yeah, the dizziness came back. Okay, got it. And so have you noticed that the shortness of breath is worse with activity? Yes, definitely when I'm going up the stairs, it's much worse than going down the stairs. We're walking on a flat surface. Have you noticed any chest pain or chest pressure? No chest pain, but I did feel a little tingling in my shoulder.

on the side of my heart, just, you know, that's why I called. Okay, got it. All right, have you noticed any lightheaded? You did mention lightheadedness, right? Yeah, I did. Yeah, when I woke up in the morning and then when I got to the top of the stairs, those were the two instances of feeling lightheaded in the last two days. Okay.

You know, I'm looking at my medical record right now, and you have a history of diabetes and hyper, I'm just making this up. Yeah, yeah. And hypertension. And you know your diabetes is totally out of control, like your last. Yeah. I'm working on it, Doc. Yeah, you know, I saw you at McDonald's last week, and you had like six Big Macs in front of you.

Those weren't all for me, but yes, I did eat too. I have been overeating. I'm 40 pounds overweight, and the diabetes isn't good. I got to cut back. I need to go. Yeah. Well, I'm not here to confront you on that. I'm only bringing it up because, you know, diabetes, hypertension.

When you tell me you have shortness of breath, I'm worried it could be related to your heart. Because you have these risk factors, so I think we should take this seriously. So I want you to come in this. afternoon. Let's get an EKG and an echo cardiogram. Okay, those are quick tests. That's going to give us a sense of how your heart's functioning. Okay, great. Are they covered or do I have to come out of pocket for those? No, they're covered. They're covered. They're easy.

And I can take them both today or do I have to schedule them? No, let's get them. I can do it today. I don't want to waste too much time. Yeah. Yeah. And then, you know, let's talk about your diabetes. You know, like, I'm going to ask our nutritionist to call you next weekend. I want them to start you on a low-salt, vegan diabetic diet. Okay, I think that's going to help. Can I just do the OZepic or Wagovi? I keep hearing about those.

Yeah. I think I'm going to... That one you might need to pay a lot of money for. So I'm going to ask our nutritionist to call you next week and discuss all these new weight loss medications. All right. And then, you know, also today, you know, in terms of your lightheadedness, I think that's probably because we have you on too many medications, actually, for your blood pressure.

So I think your lightheadedness is because of what we call orthostatic hypotension. When you stand up or you go up the stairs, you're feeling lightheaded. So I want you to stop taking your metoprolol, all right? Okay. Stop taking my metaprolol. Yeah. Yeah. Last thing.

You know, I'm sorry to confront you on this. You're a really famous person. You're traveling in the country. You're doing all sorts of stuff, the world. I'm totally willing to take us off. It's cool. Oh, sweet. And you're under a lot of stress, too.

Yeah, that's true. It's true. Last time we chatted, you were drinking like five bottles of wine every night. Good wine. Oh, yeah, it's a good wine, yeah. I mean, I was sharing them. Yeah, okay, so where are you at now? I mean, I have two to six glasses of wine a night.

Okay, all right. It's a range, you know, depends. Weekends, they're a little heavy. All right, it's a little bit better maybe than where we were before. So in terms of your alcohol habit, I'm going to ask our behavioral health therapist to call you next week and just sort of talk through your stress and see if there are other things that we could get you going with that could help. All right?

All right. Yeah. Well, you try doing six podcasts a week and we'll see how stressed out you are. I get it. I get it. Yeah. All right. Last thing.

I want you to start taking, you know, for your high cholesterol, I want you to start taking a torvastatin 40 milligrams a day. Okay. Okay. Torvostat. Yeah. I got you. Don't worry.

We're going to give you something that helps you much. But yeah. Torvistat. Liportisat. Okay. Yeah. A torvastatin also called libator. Oh yeah, I've heard of that.

Yeah. All right, good. I think we have a good plan. Anything else you want to talk about? Nope. Just want to get that Ozampic or Wagovi if you can get it covered under my insurance. That'd be awesome. Okay.

We'll talk about it. All right. So we have that conversation. Now, what the technology does, I can sort of describe in detail and then we'll look at all the output. That was a good role play. You and I are good. We got a future. We should be doing like the, when they do a spin-off of secession and they go back in time to the

childhoods. You and I could. we can take the show on the road. Yeah. Exactly. Awesome. So that means I can kind of call you in to some of our demos maybe. Absolutely. Swoop in.

Yeah, no problem. So, you know, when I was seeing patients on Memorial Day weekend, I'd walk from room to room to room. And then I'd leave and I'd realize, like, I have to do a lot of clerical work that serves three masters. I have to write a note that other people on the clerical team are going to read. And they need to understand what's in my head. Like, why did I order an EKG?

an echocardiogram. But then I also need to structure data. I need to pull out diagnoses. I need to map them to these specific codes that make money for the health system. And I'm the worst at that by the way. I do such a bad job that people chase me for weeks and months trying to figure out what I was thinking because I always do the most nonspecific thing, which is not good. But then the third master constituent is the most important. It's the patient and the family member. And they're going to go to their portal and they're going to see my note. And in my case, they're going to see a term sometimes like trans catheter erotic voluoplasty. And who knows what that means.

And so what a bridge does is it helps with all three constituents. So here's the note. And I'm going to show this in a web form factor because it's just going to give us a lot more real estate to look at. And then we can kind of geek out a little bit on how it works and now the tech pulled it off. So here is this web app. And I'm not exposing you or anything. anybody to PHI. These are all demos. And here's our conversation. So we can kind of look at the output. So patient named Jason presents with shortness of breath and dizziness for the past week.

Shortness and breath is worse with activity, particularly when going up the stairs, not experienced chest pain, but has tingling in the shoulder on the right, on the side of that heart. A lightheadedness upon waking up in the morning and reaching the top of the stairs has a history of diabetes and hypertension, overweight by 40 pounds, admits to overeating and consuming alcohol regularly two to six glasses of wine per night, more on the weekends, amatopalalal and has high cholesterol. So try to summarize your story, right? And so that must be some incredible prompt to do that.

It's got to go through that whole four-minute conversation and then summarize all the facts that were in the whole dialogue. It's a whole bunch of machine learning models we can get into that live below, beside, on top of other larger models to basically make it work. And you can see below in the assessment and plan, it's organizing all the problems we talked about, like shortness of breath and dizziness and diabetes and weight management and orthostatic hypertension and alcohol and high cholesterol. These are all problems that's summarized and that's so important because that's how you make money too as a clinician. You need to, and that's how you communicate most effectively with other doctors. You need to organize your thoughts.

And we're organizing those thoughts the way Medicare wants to see it. That's sort of a gold standard. How would the government want this? but it's sort of summarizing a little bit your story and the plan, EKG, Echo, diabetes, nutritionist to start you on that diet and we're going to have a, have them talking about Ozempic and Weegovi, and then orthoacetic hypotension, you know, discontinue the metoprolol, alcohol, behavioral health therapist is going to help you. Here's a really, really important feature, and here's the Lipitor.

If I was like, wait, did I mention Lipitor? I don't remember telling Jason about Lipitor. I could highlight any part of this note. And you see what happens. Yeah. the summary. And it takes us to that portion of the transcript. Yeah.

And then gives you that thumbs up and thumbs down. Was this, what did I say there? Was this evidence relevant? And so we get this, it's like RLHF. Like we're getting feedback now from our users. And this is a model that was trained with.

Reinforcement learning. Yep. Exactly. This is reinforcement learning. This is a model trained with millions of dollars of data that we were creating in 2018 and 2019 and also like 2020 to be able to build something that can actually build trust with these generative models because sometimes you don't know. Did that happen or not?

And if you don't trust the model, you end up looking at everything. So instead, you could be, you know, you could highlight a word. You could highlight a phrase. You could highlight a whole paragraph. But this is. Because we had a debate about how much I was drinking and it needs to go, you got to make sure that it wasn't your assessment of my drinking or mine or, you know, maybe some average of the two, whatever.

Exactly. Yeah. And that's documented correctly. Yeah. Yeah. exactly. And let's just say, I'm like, wait, did we, I don't really remember that part. I want to listen to it again. You could even listen to it. And then, you know, also today, you know, in terms of your lightheadedness, I think that's, that's probably because we have you on two minutes. So hopefully you could hear that. But you can go all the way from the summary to the transcript to the ground truth, which is the audio. How long have doctors been using this in the field? Where are you at? Yeah, probably like about six to eight months now. Wow. And how many, like for the, for the, the person who's the tip of the spear, the doctor who's got the most, who's not you, how many have they done? And what's their reaction been over the last six months?

Yeah. Once a clinician starts using this, we are like over 90% of clinicians stick with it. They use it for every single one of their patients. That's how we define it. Every time in clinic, do they use it? It's like a... Oh, so your batting rate is 100%. Yeah, exactly. What was the most positive feedback and what was the most negative or, let's just say, scared, cautious feedback? Yeah. Because this is, you know, any recording of a healthcare interaction and given the litigiousness of the United States, I can see people framing this multiple ways, hey, this is a way to document things so that you don't get sued. And if you do get sued, you have the evidence.

I can say other people. You know, I've been on different boards over the decades where they say, hey, we should record the board calls. And every law firm's like, don't record the board calls because now we got a transcript and don't transcribe them and write the shortest board notes possible. So it's concise so that we don't have, if there is a shareholder or any kind of lawsuit, they don't weaponize it against us. So let's walk through the reactions and then this very acute issue. Yeah. If you use one of your industry terms around liability, go up, down, sideways, left, right? Yeah, yeah, absolutely.

So I can answer those and I'll just I'll wrap up this demo really quickly, too, on the other side of that. So University of Kansas Health System, they have like over 2,000 clinicians there across any number of different hospitals and clinics. The chief medical information officer is Dr. Greg Ader. He's an E&T doctor. So this technology works across specialties, by the way. It's not purpose built for one. We've been training and annotating to be able to build for all. and Greg Gator uses it his first time in clinic, and he told us afterwards his immediate reaction

was that this is the most important thing to healthcare since the stethoscope is what he wrote for us, and we can't wait to put that on our website. But that kind of feedback is rolling in on not just a weekly basis, it's happening every single day with clinicians. So I truly believe this is the biggest, most profound opportunity for generative AI. You can't swing a stick in healthcare and not hit a billion dollar opportunity to change something that's so important to every single one of us. And as long as you believe in the thesis that health care is about people and they're having conversations, that it's not about machines, that it's about people at the end of the day. And it's local too. It's happening all over the place.

Fellow medicine, in person, call centers. Then you can believe in all this value that, not. not just we're demoing today, but what's downstream? Like, for example, what's downstream here is coding? Like, we can pull out all the structured data and give you the evidence to support the code. I see. It says there I see D codes. What does that stand for?

Yeah, so that's like this industry, um, sort of nomenclature around a type of code that you can bill for, that you can get comments. Those are the billing codes. Yeah, we hear about that all the time. Exactly. Yeah. So these are billing codes. These HCCs are different type of billing code.

But the key piece here is that we're, pulling out the billing code, but we're also giving you the evidence that you can, again, go all the way back to the ground truth to use it and defend. So it's like revenue insurance in a way. It's like it's creating a very trustworthy, not just transcript, but these insights that it's pulling. If you're a SaaS or services company that stores customer data in the cloud, then you need to be, uh, SOC2 compliant. You knew that from a third party. And you need that third party to close big deals. And if you want to get compliant, easier. And, faster you need to use. V-A-N-T-A. V-A-V-A-V-A-V-A-V-A-V-A-V-A-V-A-V-A-V-A-V-T-A-V-T-C-A-V-V-A-V-T-C-A-V-V-A-V-V-T-V-P-T-V-R-V-V-A-V-P-R-B-A-P-R-B-P-R-H-P-P-R-HIPA-P-P-R, and more. You can't afford to lose out

-on-M-M-A-R-R-H-P-R-P-P-R-P-R-H-P-P-R. You can't afford to lose out on major customers. We all know that. Listen, it's a hard year. Last year was hard. You can't lose those major customers because you don't have your compliance dialed in. Just work with Vanta. Get your compliance automated and tight and tight is right. Lock down those big deals. Here's the best part. Vanta's going to give you $1,000 off. That's $1,000 off at vanta.com slash twist. That's vanta.com slash twist for $1,000 off. Your stock, too. How much time does this save per interaction on average? Right now, doctors are telling us we're saving them two to three hours a day in a clinic.

Yeah. If you purchase... If they work a 10-hour shift, or 12-hour shift, I mean, we're talking about a 25% lift. Yeah, yeah, exactly, exactly. That's weird. You know, it's so interesting when I had Aaron from Box on,

Reed Hoffman on, and founder of Airbnb on, Ryan Cheskey, I asked them like what percentage of time they were saving. I asked their mesh the same thing. They all said like 20, 30, 40%, depending on the role. It feels like this current wave of AI innovation out of the gate seems 30% of people's time. That's kind of mind-blowing.

If you think about on an entire economy, now we're seeing it not just an artist making, you know, wedding invitations or, you know, SEO writers or people writing blog posts or somebody doing web research or travel research or travel planning. I mean, we're seeing it now with doctors. And as we do this, these are people whose time is even more and more valuable. And then you mentioned earlier, there's a shortage. Yep. You get all your doctors to use this and use it well. In year one, you get a free doctor for every four. Right?

Every three doctors, you get the fourth free. Yeah. Yeah. Just think about what this could save in healthcare. Totally. And then less mistakes. And the mistakes are costly. So that's like, and obviously there's cost in money, but there's also cost in death or missing a diagnosis and suffering, right?

Which are really hard to even quantify. Although I guess people do quantify them. Yeah. And there's an article that was published in the Journal of General Internal Medicine just some months ago that estimated that doctors actually require 27 hours in a day just to do all the work that we put on their plates. Yeah. So we need this. I mean, I think one of the most profound bits of feedback that we're getting from users is that they just feel unburdened. They feel like they're operating at the top of their license.

They can focus on what they want to do, which is be creative and figure out what is the differential diagnosis. What are all the possibilities here? What should we try? And I think we're also being clear that this is also an extension of every doctor and nurse's best intention to be there for their patient, even when they're not in front of them. This is the, I guess, the website, you're doing mobile web or an app? This is a mobile app now that's yours. So it's like React or something in a wrapper or something, yeah. Exactly.

So it says your conversation is ready for review, explore your summary, review changes to your medication. So this is my view of our interaction. Fantastic. Exactly. So I get that note. I make my quick edits. On average, it takes like about 20 seconds, if that for people to quickly make it perfect.

And then I say, send a record, but send a patient. and now this is what you get. And it's the first step of our machine learning pipeline where we're not just recognizing the words. We're selectively pulling out just the medical moments, not the part about politics, not the part about the NFL draft. Your kids or whatever. Exactly.

So if you went home and you could filter by the starred moments, which are like the most important moments, like the next steps. But if you went home and you're like... The starred moments are done by AI? So you're saying like little snippets from our conversation, or are these done by the doctor? By AI. Yeah.

Got it. Can I play them or are they just transcript? like this. All right. So do I get the entire conversation or just? You can get the excerpts, but it's kind of up to the enterprise when we do these enterprise deals. So I guess the question then is we have to adjust because, you know,

things that can be leaked will be leaked. You could have a patient who a phone could get hacked and somebody's calls could be leaked and their private conversation with their doctor about something very private, depression, you know, whatever, fertility, STD. whatever, you know, something obesity. And so how do you think about that? Yeah, we have to jump through all sorts of hoops, you know, in relation to HIPAA, which is, you know, the main, one of the main constructs in healthcare around data privacy

and security and how really like this should belong to the patient. And so we have to jump through those hoops. We have to jump through hoops around where is this data stored and how is it stored, how long is it kept? And what is the data? Is it just transcript? Is it summary? Does it include audio?

But this piece of it is, I think, a kumbaya across everyone, which is what we also do for you, for the patient, is we pull out the key pieces, parts of the conversation. We decode it. We define everything at a fourth-grade reading level. So you're explaining to people what an EKG, an echocardiogram, diabetic diet, OZempic. So there's a glossary attached to all this. So when I felt in this conversation with my doctor that they were speaking French,

totally. Now I can kind of catch up. And so that whole intimidation and like not questioning your doctor, etc. And now I would suppose, you know, I might want to communicate with the doctor and double click on one of these and have a threaded discussion. So and annotate this in some way. Yeah. So I guess that's probably on the roadmap, yep. Absolutely. And I think one of the ways we approximate for that right now

is this where you turn the app off. Maybe in a week you get a notification like this. And it says, hey, like, Jason, we think your doctor talked about OZemPEC. And we're pulling out just the key moments and reminding you, like keeping it top of mind, like an endocrinologist to talk about OZempic. I told you to take a Torvastatin. You could play that digest if we've enabled audio and the enterprise is okay. We've jumped through all those hoops. If you say no, then you give us annotations. You tell us where we went wrong.

And then that improves our model as well. So we get that sort of that data network effect on both sides. not just the clinicians, but also the patients. Well, you know, I think it's, this is a double opt-in kind of thing. You did the courtesy of, hey, would it be okay for me to use this if I'm an oldster and I don't want this? Or I'm a paranoid person or, as you mentioned, a celebrity. And I don't want there being a record of any of this because you're criticizing my alcohol consumption. I think you were trying to do this with Chamath.

But he's the one who's drinking far too much wine, I think. even as good as it is. So interesting. That's, yeah, and this is so many societal best practices in thinking about this. Young people, I mean, they're recording themselves doing all kinds of

crazy stupid stuff, so doing something intelligent like this, yeah, they're going to be fine. And then the people who are part of the paradigm who don't want the stuff recorded, well, paradigms don't die, people do. So they will be part of the paradigm that doesn't want to record this, but my lord, can you imagine

if I had this since I was going to the doctor since I was, you know, whatever, eight years old, and I had my entire history. And then the AI could look at it and say, you know, the shortness of breath thing, he had asthma when he was a kid, and he never brought this up. And he had an inhaler.

And by the way, like, you know, maybe we should screen for something. Maybe he's got some lung deficiency. The AI is going to find stuff in here. That's going to be incredible. I mean, it's extraordinary. So how do you charge for this? You charge per interaction or like $25 bucks in interaction

or you just charge a flat rate or per doctor? Yeah, we try to make it as simple as possible per doctor per year or per clinician per year. And we go deep on the other side of this note. We integrate it all into the right places in the medical record because there's discrete fields. There's, you know, part of this is not just the generative piece. It's all the copy and pasting you might have to do if we weren't integrating. Like we're going to make sure we already do with these large health system customers, but across all the EMRs, we'll make sure to get.

get all the data in the right places. When you're selling to B2B buyers, you really want to get your pitch in front of the decision maker, the person who gets to sign the check. Because these upper level execs, they're the ones who make the purchasing decision. Everybody can have an opinion on the team. Of course, it's 2023. But there's always somebody where the buck stops. And that buck stops on their desk and doesn't get into your bank account.

These high level folks are hard to find. They're hard to target on social media platforms. but LinkedIn is the social network for business and they have 930 million members ready to do business with you and that includes the 180 million senior level decision makers. Plus, don't tell anybody there's also 10 million C level executives there. That's a ton. Purchasing power. LinkedIn ads is built specifically for B2B marketers. No other platform in the world can offer these eyeballs and you can target them obviously by their location, the size of their organization, they're vertical and their title. When you think about business, I want you to just think about LinkedIn, LinkedIn

equals business, business equals LinkedIn. It's that simple folks. When you present them with an opportunity, they will, of course, be in the mindset to receive that because they're not posting pictures of their food for mentally on vacation. Make B2B marketing everything it can be and get a $100 credit towards your next campaign by going to LinkedIn.com slash next unicorn to claim your credit. That's LinkedIn.com slash next unicorn. Terms and conditions apply because LinkedIn is so generous to the this week in startup's audience. All right, I got to ask you the obvious question. Yeah.

And you're a cardiologist, correct? Yep. Don't take it the wrong way. Yeah. But in 10 years, if you, in 100 interactions, like the average, 500 interactions, how many of them would, you know, 10 cardiologists out of 10 cardiologists give the same basic feedback for? It's a great question.

I think what your, cardiology is an interesting field because it is, there's a lot of dollars in cardiology. So there's a lot of research studies. There's a lot of algorithms that we all follow almost too robotically sometimes. Heuristics. Yeah. Yeah. If you do this, do that. If you do this, do that.

So everybody's going to look at this just like the writers who are on strike right now and say, aren't you just building this in order to replace the doctors or maybe replace the first third of interactions? And so I'm looking at and I'm going, you know what? I can't afford health care. I'd rather just pay you 50 bucks to talk to your AI and just, or, hey, I'm a, I'm a country in Africa that has no medical system. And the idea of getting a meeting with a cardiologist is for this 10 million person population impossible. It's a frontier market. Or it's in the south here in the U.S.

There's somebody on the Appalachian in the mountains in Appalachia, and they don't have. have any health insurance. Your AI can beat nothing already. You would agree? Take a second to think this through because it's going to be in the episode.

But if it was your sibling and they had a choice no doctor or talk to your AI, what would you tell them to do? That's a good question. Our AI just to sort of recenter and not make sure not to overpitch

what we're delivering to the market right now, we are not delivering a chatbot experience where you can kind of talk to a AI doctor. Of course. You're transcribing it. You're categorizing. We get that in five years when you have 100,000 interactions. We both know where this is headed. In five years, we got 100,000 of these in here.

AI continues to grow at a modest pace in terms of its fidelity. Would you rather your sibling not go to a cardiologist at 50 years old when they had symptoms or talk to the AI in five years? Yeah, I mean, absolutely the latter. And I think that there's a responsible way to do it, though. And I think those are the solutions that are going to win. Like in healthcare especially, it feels like the ultimate sort of moat is around trust. And trustworthy genera AI, trustworthy AI is probably some combination of credibility. Like, have you shown that this works, like that it can be sort of kind of safe for maybe at least the front line kind of issues that people have on a daily basis?

is, is it reliable? You know, once you hit it once with the same question, is it the same answer or is it a different answer every time? You know, that's one of those issues that we see with some of these models. And is it transparent. And I think what we as a company are trying to do is build those three pillars into the DNA of everything that we do. So we've published tons of peer-reviewed computer science papers about how do you point a transformer

at health care use cases. We do that in the clinical research world now as well. we validate and we get as rigorous as we can. We build our own models with our own proprietary data sets. So our own speech recognition system internally we call ears, the multilingual system that we fine-tuned with healthcare data to get words like Ozempic and we go V. But then on the LLM side,

we have our own in-house LLM called ELMS. And we also use the power tools that are out there, but we always make sure to build those models like I demonstrated that give you that transparency. Like, where did this come from? How did it get generated? So I think whatever that time horizon is in 10 years, 20 years, five years, can AI start to be an incredibly good triage nurse or clinician in an urgent care and deal with a lot of the issues that all of us have? Absolutely.

And I think that's awesome. So that's like fantastic. That's what we need. I mean right now, just think about the internet. There's whatever billion or so people not on the internet yet. If they got Starlink and they get access to Wikipedia and a bunch of health websites, that's going to really help emergency care or whatever. And then you start thinking about the AI impact.

You know, there's going to be somebody in the field who has some symptoms and they don't realize that's indigestion versus a heart attack or a stroke. Lives are going to be saved. So I'm super excited about lowering the costs and increasing the accessibility of this. I think that's like this is a, this is this incredible fulcrum that you found, which is there's an acute problem right now. Yeah, save time for doctors who are quitting because they're nobody,

they're overworked doing 27 hours and a 12-hour shift. They're underappreciated. They quit constantly. And they're exhausted. And they're underpaid. And they're sued too often. Now I take this great fulcrum. You make them superhuman.

And then the AI is going to really be able. able to take two-thirds of, you know, interactions maybe off their plate and they can focus on the most important ones. Totally. That must happen already. Like, the idea of even taking your temperature at home or your blood pressure at home or even knowing how many hours you slept last night, that's something that didn't exist, whatever, 50 years ago, right?

Yeah, exactly. And my patients, as a cardiologist, sending me Apple Watch strips. Oh, do they do that? That was a change. Yeah. Yeah. accurate enough to be of value? Absolutely.

It's really good at detecting atrial fibrillation. So there are patients who've been diagnosed with AFib because of the Apple Watch. And then we kind of pull them through. Yeah. Is that the proper path or is it like chest based more accurate? Well, the 12 lead, you know, the chest base is going to give you a lot more detail on what's going on. But the Apple Watch is validated to do a good job on detecting that diagnosis, that issue. I wonder if they're going to get that glucose monitor work.

There's that rumor that they're going to have a watch that will be able to tell your glucose level or whatever. Yeah. I mean, is that possible? Is that real? Blood pressure, too? Blood pressure would be a game changer for me. How would they get blood pressure through a watch? I mean, I know you can buy like the weathings makes a smart blood pressure cut and that will go to the cloud. Yeah. Yeah.

Is that even possibly anything with a watch? I mean, there are people who are trying to approximate blood pressure even from EKG strips. So with enough data, you know, and. And maybe you can start to validate some sort of instrument that can create a proxy for a blood pressure at the very least. And maybe that in and of itself can be valuable. Well, congratulations. You started on this years ago and the world caught up. And now everybody's going to probably want to throw money at you after four years of you begging them to support your vision and them saying it's too early.

And now they're too late. And all these investors are going to have to play catch up now and pay a little bit of a higher price for not having the vision to back you earlier. Congratulations. You're hiring right now? I take it. We are. Yeah. So let's get some positions filled here. Go to abridge.com. I guess you got a careers page up there. I'm guessing. We do. Yeah. So go to the careers page. What do you most need? What do you most acutely need? What roles and what's the culture and type of person you want to have at the company? Yeah, absolutely. Absolutely. I mean, it's a refrain that we use all the time. We all hear all the time. But this company is absolutely mission driven. If you want to come in and see, impact today on a rolling basis feedback from all sorts of clinicians and patients and administrators across some of the biggest health systems in the world. And you're training a model and you ship it in a matter of minutes and you get that feedback. Like, that's the kind of culture we have. We're building responsibly, but we're building incredibly quickly. And we're getting feedback loops that are just like instant. Yeah, you've got to check these things, right? You have people's,

yeah. You got to check them. But we have the instrumentation in place where we can benchmark very quickly quantitatively and decide to ship. So I think one of the best compliments we got, compliments we got from an early customer was a bridge on Monday was amazing. And it was better on Tuesday and even better on Wednesday, Thursday, Friday. And that's the speed that we're going at right now, which in healthcare is a totally new thing. It's all happening. And if you're not using chat GPT4 every day, you're nuts.

If you're listening to my voice, buy everybody on your team, $20 a month for chat GPT4. you can go pay for Po or start using BARD, but turn all the experiments on and then have your teams using it every day and see if they can get 20, 30% faster at their jobs and make your company more profitable or get to profitability. This is an incredible revolution. It is not undersold. It is not oversold.

It is undersold. You and I who are in the trenches with this, you can just see how this is going to compound I just had Tim Urban on. I don't know if you've ever seen Tim Urban's favorite famous chart. It's like progress and then boom. Yeah. Yeah. Yeah.

And I'm making a line that is just a very slow, gradual growth line. Hopefully producer Brian has it handy on the show. And it's like puts a little stick figure on the line. And then the line goes hockey stick straight up, like more than a hockey stick, like a skyscraper. And it's like, this thing is going to become in, I mean, it's passing the Turing test already. so I and vertically I is always better but there it is human progress in time we are on the cusp of something bizarre and uh yeah I mean doctors are going to be learning stuff they just the the idea that you're a doctor and you have to do trial and error and yeah you know triangulate around what's going on like you said you're starting to get Apple watch data now imagine like you've got a thousand patients in your network as some insurer yeah and you know you're start sending messages to your clinicians, like, this person's heart rate and this,

and you both get an alert, you're both dumped into a chat room, and it's like, and the ambulance driver is on the way to their location. Like, that's the world we're going to live in, where, like, all of a sudden you're going to get a phone call from your doctor. And it's like, hey, Dr. Shiv. And it's like, yeah, the ambulance is on the way. And you're like, uh, well, we saw you're having this happen with your heart.

We just want to be safe. And, uh, yeah, they'll have the, uh, whatever that drug is they put under your tongue if you're having a stroke. Yeah. Is that thing they do? Yeah, they give you nitro glycerin and aspirin and like a whole lot of... Yeah, you know, with the nitro and the ambulance is two minutes away. Here it is like an Uber coming to.

I mean, it's going to be a crazy world. The fact is like people have gone flying off their mountain bikes alone. And if you have the latest iPhone and it has like the crash detection. Yep. And like somebody got knocked out. They were knocked out unconscious and somebody came and saved their life because of their Apple watching a phone. I mean, this is going to be, I was an EMT for a little bit. And like, just what's going to happen in emergency medicine combined with this?

I used to have to sit there. I was, I did dispatch too for an ambulance. And they would be whaling down the BQE in Brooklyn yelling at me the vitals. I'm writing the vitals down. I'm on the phone with Victor Memorial Hospital reading or, you know, Momadies and I'm reading the patient's blood pressure to the nurse at the emergency room while they're screaming to me. And it's like, this stuff is going to be like all. Totally.

And imagine if you had a bridge, like even for those calls and you're just getting the summary already and you can communicate that to the hospital. Absolutely. That's the thing that's blowing my mind more than anything else in 20203 is that healthcare is moving at the pace that it is with this technology. Is it really moving faster finally? It is moving so fast. Why did they go from not moving fast to moving fast? Is it just because it's so broken and they're so exhausted and they're finally ready to capitulate and use it? technology? Yeah, we just pushed way past the point. And on the other side of the pandemic,

clinicians just collapsed. And it was enough was enough. We had to do something. We have to do something. And so it's desperation. We need tools. And these tools are working. I think that's the other thing about generative AI. It's we're coming in and we're just putting, we're just saying, use the product. You will see. Like, the value is there. This is not presentation. It is practice. We have thousands of doctors using it today. Yeah. So, yeah. Amazing. The proof is there. Dr. Schiff, I'm glad you're in the world.

You can follow Shiv, D-E-V-R-A-O on Twitter. A Bridge is a company. You go to their jobs career page. Continued success. I'm just glad you're out there in the world doing this as God's work. If you're an atheist, it's humanity's work. Okay, pick whatever you like. But you're going to save a lot of lives.

And for that on behalf of the audience, humanity, as the host of this week and start out to think. Thank you, Jason. my unique physician. Where are you based? Thank you so much. Where are you based? I'm in Pittsburgh. Oh, yeah?

Oh, wow. Very nice. Cardi, melon. I mean, what a university, huh? I mean, the talent coming out of there is just extraordinary. I went there a lifetime ago and then became a cardiologist, then became the corporate VC, and now I'm just trying to put it all together here. You can buy an amazing home there.

Yes, you can. You can buy like a mansion there for 300 grand. It's crazy. You can have an incredible. incredible lifestyle here. You can, we've got, we don't might,

we might not have like 10 of everything, but we have like at least a couple of all the things you probably need. We have a vegan Polish restaurant that is like one of the best in the world. I was on CNBC and I knew somebody who was like at the endowment of Carnegie Mellon. He was a very proud person from Pittsburgh. And they were like, do you think like Amazon HQ2 is going to be in Pittsburgh or this place or you think it's to be my head?

I'm like, young people don't want to be in Pittsburgh. They're going to want to be in New York. Miami, Austin, San Diego, Los Angeles, and maybe D.C. So I'm going to go with like Austin, New York are going to be placed. And it was D.C. New York. Yeah. And he was like, what the heck?

Like, why are you doing? I'm sorry, bro. Sorry, I'm going based on a stereotype. My bad. But then he's like setting me pictures of like this incredible, beautiful bucolic countrysides and houses. It's gorgeous. It's awesome.

It's gorgeous, yeah. It has all the vibes here. And great companies here. like a sister company of ours, Duolingo. Of course, yeah. Another Union Square Ventures company. Oh, your Union Square? Yeah.

Yeah. Yeah, we're USV and Bessemer and and we'll announce more. Congrats. Well, shout out to USV for another winner. And we'll see you all next time in the sweet start. Bye-bye.