Transcript 0:00 [on-hold music playing] Hello, everyone, and welcome to the Digital Leaders Show. 4:58 We're live every other Wednesday at noon Eastern, 9:00 Pacific, talking key issues and news at the intersection of business and technology in hopes of making you a better digital leader. 5:09 Today, we're discussing leveraging GenAI in healthcare with Optum Health's Head of AI and Intelligent Automation, Sumit Punjabi, plus other headlines for digital leaders. I'm Dan Goodstein. 5:19 I head up IRPA AI, the Institute for Robotic Process Automation and Artificial Intelligence. 5:24 We're a professional association, media, and peer-to-peer research network providing education, networking, and advice to over 130,000 members worldwide all around automation, enterprise AI, and digital transformation. 5:36 Check us out at irpaai.com. 5:38 My co-host today is Chris Serdak, Senior AI Advisor to IRPA AI, author, consultant, and former Chief White House Chief Transformation Office Advisor, Partner Digital Transformation at Gartner, and Automation AI Practice Lead at EY. 5:52 Welcome back, Chris. Hey, good to see you. It's been a while, been a while, uh, since you and I have spoken. Uh, the internet is broken And hence us being late this morning. We were, we were c- CrowdStriked? Yes. 6:06 Is that, is that- I'm gonna blame it on that... is that the new term? 6:07 That's the good news with something like this, is even when you do have a, a non-related technical issue, you get to blame the, the, the mass outage that's been going on. So I'm gonna, I'm gonna blame it on that. 6:17 I, I, I love the memes out there. It's like, "Oh, our, our data is in the cloud." It's like, no, it's not, it's on someone else's computer. [laughs] It's- Yeah. You know? It's one of my- It, it's funny you say that. 6:27 We could start with that topic. Today me and my, my dad called me up, uh, on Friday and, you know, he's trying to figure out what's going on. 6:34 He's, you know, big, you know, stock market guy, and he's like, "What, what is, what is actually going on?" He's like... Well, he goes, "Where is the cloud?" And I'm like, "I-" [laughs] "... how do you explain this?" 6:43 But, you know, it's, it's, it's always great to try to explain something like that to someone who's not in the industry 'cause it forces you to, like, really explain it in such a way where you have to make sure you understand it. 6:53 And so, yeah, right? It's, it's not this fictitious, magical, uh, mysterious thing, right? Well, and, and, um, it, it, my... It take... 7:04 I put on my systems engineer hat, and that is, you know, if something can go wrong, it will. It's just a matter of time. 7:10 And then the more, the more people that don't acknowledge that, that the, the, the fail-safe or whatever is out there, the more likely it is to be really bad when it happens. 7:18 So when every other machine in the world kind of, uh, goes down, um, you kind of notice. [laughs] So- But, but help me understand, Chris. 7:27 Help me understand- I mean, how do you think that these things actually work as well as they do, and this doesn't happen more often? That's the part that blows my mind. Well, I... That's... You beat me to it, right? 7:34 I was just gonna say, you know, I'm a technologist. I, I evangelize this stuff. I trust this stuff. 7:41 I argue with people on a daily basis about, you know, uh, innovating and, and, and why we need to, you know, evolve technology as a... But, like, how... 7:52 Yeah, everybody's so caught up, and, and rightfully so, everybody's so caught up in, like, the AI ethics and AI is gonna take over the world and, and, and all this stuff. 8:00 And, you know, here we have a good old-fashioned outage. Not even a cyber attack, right? A good old-fashioned outage. How does this happen in 2024? Um, a miscoding. It's, it's someone just screwed up in a line of code. 8:13 And I always joke Agile is just waterfall development without testing or documentation, and here it is right in front of us. 8:21 So, so without getting into the, into the nitty-gritty, like, what, what went wrong and, and how, how do we avoid these things? 8:29 Because you think, you think you're kind of safe with this stuff at this point, um, and then you find out, well, 85% of the world is using the same provider. Like, I, I'm thinking with all the... 8:40 And we can ask our, our guest today who's, you know, no, no stranger to risk. But I'm thinking if you're a risk or compliance guy out there, like, how do, how do you put all the eggs in one basket? 8:51 Well, and, and, and I make the same parallel to using, um, AI. You know, generative AI is a bias amplifier, so if there's any bias out there, it's gonna make it even worse. 9:02 And so all of us throwing all of our eggs in one basket like this is exactly the same thing. Everything is great until something goes wrong and then we're all Chernobyled, right? 9:12 The, the best scientists in the world can work on this, the best engineers in the world can work on it, and it's gonna work brilliantly until it doesn't. And that's just, like, the accumulate... It, it... 9:21 This is going back to my rocket days when we were trying to figure out what's the chance that we're gonna make it to Mars. Um, the errors accumulate until the error happens. 9:30 And, and so again, it's just, it's just a, a standard systems engineering thing. I mean- If you don't, if you, if you don't design for resiliency, you won't have any resiliency. Well, I... 9:41 You know, if, if any- if there's any compliance or even better yet, any, any regulators watching this, I, I, I know you're busy with AI, but th- th- this to me has to be addressed. 9:50 I know Department of Transportation's going after, like, Delta now, you know, to say, "How did this happen?" But it, the, the issue is bigger, right? 9:57 The issue is, you know, you know, uh, should, should companies, large companies, global banks, airlines, you know, medical companies, you know, uh, all be on the same platform? 10:09 I mean, they're so worried about breaking up Google and, and Apple and, and all these, and all these things, but meanwhile, you have one provider that, you know, single-handedly, probably one person at that provider, who single-handedly broke the world, um, just because, you know, the percentage of businesses that are using the same exact platform is just... 10:29 That's just crazy to me. S- so since we're talking healthcare today, I, I, I had a client many years ago who I won't name, but they were a blood fractionator. 10:37 So they collect blood and they separate out plasma and stuff, so they make therapeutics for hemophiliacs. And literally, if, I mean, if they don't get those injections every week, that person's gonna die. 10:47 And, uh, I was helping them deploy an SAP system, you know, ERP, that managed the inventory of their therapeutics. And they always had a year's supply on hand, which means it also takes a year to resupply, 11:01 and they only had a single instance running in production. And if that single instance went down, the FDA rules said that they had to throw away all of their inventory because they lost chain of custody, right? 11:11 So a, you had to throw away a billion dollars of inventory that would take a year to resupply, and all of your customers would die over the course of that year. 11:20 And rather than putting in a hot backup of their SAP system, they just bought a billion dollars worth of insurance because it was cheaper. 11:27 And so, so those kind of decisions are rampant, I think, throughout business and, and, um, you know, every once in a while, you lose the bet. 11:35 So while we're, while we're talking about risk, Meta has released its, uh, new Llama AI model. And, uh, the news here is not only is it supposedly very, very good, uh, it's free. They're gonna offer it for free. 11:49 So while everybody else is trying to sell and monetize these models, Meta's gonna offer it for free, which is, you know, uh, a- adding another chorus of people saying, you know, "Hey, uh, you know, how do we put guardrails on this thing?" 12:01 It was, it was democratized and kind of the Wild Wild West enough- Um, when it was free to end users, but corporations had to pay for it. Now, it, it, it may, it may be free to everybody who wants to, to play with it. 12:15 I, I, I guess I have mixed feelings about it. In one in, in one hand, because you have all of the weights of all the nodes that they've trained on, anybody can replicate, um, the outputs of, of Llama now. 12:27 But nobody know... There's no way of knowing why those weights are what they are based upon the training. So, so it's, in one regard it allows a little bit more openness and a little bit more oversight. 12:37 On the other regard, it's a big nothing burger because I still don't understand how you got to the weights that you have, uh, and necessarily what those weights mean. 12:45 So, so it does allow a bit more oversight, but I would argue not nearly enough. 12:50 The last headline I wanna hit real quick, and I know we're, we're running late so I wanna get to our, our guest, but, uh, uh, interesting, uh, uh, FTC issued, uh, civil subpoenas to eight companies in the financial services and consulting space, including Mastercard, JPMorgan, and McKinsey, saying that they're potentially targeting pricing products at Starbucks, McDonald's, Home Depot, Kirkland's, others, um, based on your data. 13:16 In other words- [laughs]... they figure out that you can, you have the, the propensity to spend more money, so they're actually gonna charge you, uh, more money. 13:25 So they're calling it surveillance pricing, kind of the cousin of dynamic pricing, which is what rideshare companies and airlines use for things like surge, uh, pricing. But, uh- How is this- Interesting, right? 13:36 How is this a surprise to anybody? I wrote about this 12 years ago, dude. Where have you all been? [laughs] This is a good, uh, good time to plug your book, Chris. Oh, my God. 13:44 Because I think people are shaking their heads now saying, "Now what?" I mean, we do behavioral analytics and surveil all the people that carry our smartphone all the time, and now we're worth $6 trillion. 13:55 Who, w- who could correlate those things? That's hilarious. But yes, I, I mean, I've been writing about increase, the increased 14:02 aggression almost of the FTC and some of their, their, um, enforcement activities, and it's gonna continue to go up. 14:09 So those people that have kind of slid along a little bit on their compliance activities and information privacy and things like, you know, cataloging your information and knowing who's accessing it and so forth, you're gonna have to really pick up the pace on those activities because those almost predatory analytic practices, um, are, are finally coming under a bit of a microscope here, so. 14:31 Well, and the, and the initial probe claims to be f- you know, focused more on equality issues, like are certain groups of people being targeted? 14:41 But I, I think the bigger question, and, and, you know, this may, uh, you know, 14:47 border up against politics, but, you know, uh, depending on what adminis- administration is in power, will they use this to, you know, start stopping, you know, some of these, uh, data collection policies? Well, by... 15:00 I mean, behavioral analytics is necessarily des- i- i- the whole point of it is to look for biases. 15:06 The whole point of it is to discriminate one group of people against another group of people, and then monetize that understanding. So it's inherently biased. It's inherently discriminatory by design. 15:19 Um, the challenge is, am I benefiting from being discriminated? And, and, uh, we used to call it the market of one, right? So every individual is a unique market. Um, so it's the catch me... 15:31 It's push me, pull you, right? Where as long as I benefit from being discriminated, it's okay. But then if I, if I don't benefit from being discriminated, then it's not okay. 15:41 It's the privacy paradox that we've been discussing, like, again, for, like, 15 years. Yeah. Yeah. Okay, great. Well let's, let's get to our main topic today, which is, uh, leveraging GenAI in healthcare. 15:54 We're, we're joined today, uh, by Optum's Health, Optum Health's Head of AI and Intelligent Automation, Sumit Punjabi. Sumit, welcome to the show. Muted. 16:06 [laughs] We've been cloud struck again. [laughs] Sumit's having some audio issues. 16:23 So I don't feel so bad, Dan, because I, I've had how many audio issues in our shows over the last six months? Yes, I know. It's usually, uh, your fault. I usually- Yeah, it's always a configuration thing... 16:32 I kinda blame you for spreading the, the audio bug. We got you now. Oh, thank God. [laughs] But you hear a, an echo? No. No, we're good, Sumit. 16:42 So, uh, uh, other, other than, uh, the local tech issues you're, you're having, tell us a little bit about, uh, yourself and your role for those watching who don't know you. Oh, sure. 16:54 Um, actually, before I begin, can I ask who from Texas has dialed in? You could do a thumbs up or do a heart or whatever emoji you guys like. Anyone from Texas? In the audience? Okay. 17:12 Anyone from Chicago? Chicago land area rather. All right. Okay. Anyways, sorry. I just wanted to see if our audience is engaged and participating. Um, but- Yeah, I think they can't, uh... 17:27 I think only if they're on YouTube, they can, they can comment live. Uh, so I may have some people popping in as we, as we go. No, I think you should... Uh, anyways, sorry. 17:37 So Sumit Punjabi, um, as you mentioned, Dan, automation leader with Optum. Um, I came through the Change Healthcare acquisition, um, which was back in 2012, but I led the automation concern, um... 17:54 I'm getting a lot of echo on my end. I can hear my end two times. We're okay on this side. We're okay on this side Okay, I'll, I'll just keep rolling. Um, 18:11 let the NTT, um, data's automation consu- consulting practice... Sorry, this is really distracting. It's like I can hear my house, uh, voice boom and everything. Let me try to, um, use a different headset maybe. 18:26 I apologize, guys. Um, Dan, technical difficulties continue. I'll hand it back to you, and I'll be back in a second. Okay, thanks. So, um, Chris, you were telling me a story before we got started. 18:41 Maybe that's a good setup, uh, for with Sumit when he gets back, right? Um, uh, you know, some of the work you've been doing, you did in healthcare early on, and this is... These are n- not new, 18:51 not new questions, new challenges, right? It's just kind of been accelerated in the last year and a half or so. 18:58 Well, I, I, going way back to [clears throat] my time at Accenture or HP, you know, a good 10, 15 years ago, we were always looking at ways of using machine learning and a- analytics to improve healthcare outcomes. 19:11 And I don't, I don't know if you... 19:12 There was always the famous one where IBM was using, uh, Watson in early generation machine learning for, um, reading, um, X-rays for things like cancer and, um, how do you, how do you pick up the instances of, like, breast cancer and so forth from, from doing the X-rays. 19:31 And it was actually quite effective. I recall the o- So it improved the outcomes there, 19:36 but what it tended to do, and this is exactly what we see with generative AI right now, is for people with a high degree of expertise, it makes them more effective in what they're doing. 19:47 For people with almost no expertise or kind of low-end expertise, it can make them more efficient. I'm, I'm sorry. It makes them more effective, and it makes the experts more efficient, okay? 19:59 And that's exactly what we see with generative A- AI right now. If you have below average capabilities in a certain thing that you're trying to do, generative AI can bring you up to average level of performance. 20:08 If you're a, an exceptional performer at what you do, it can make you more efficient. [laughs] It lets you get your job done a little bit better and a little bit faster. 20:18 Um, and so that was very much the case with doctors. The other thing I, I recall from back then is that when you showed the results from that analysis of, of X-rays to doctors that were experts, 20:30 um, they would, they were, they would agree with the first or second sort of diagnoses that came up, but they would be shocked by the fourth, fifth, sixth and seventh answers that came up. 20:40 It'd be stuff that they never thought of, and it's like the old TV series House, right? 20:44 House could always put these weird correlations together and come up with a diagnosis no one could ever think of, and those sorts of results tend to come out of machine learning and, and AI when you're applying them to questions, um, in, particularly in healthcare because there's so many different factors to consider before you come up with a diagnosis. 21:06 You're on mute, Sumit. Now can you guys hear me? Now we got you. But do you have an echo? [laughs] Little bit. Let's, let's, let's keep going, Sumit. Let me jump on the phone. Maybe that's, that's the thing. Yeah, sure. 21:17 That, that sounds good. Let's, uh, have you on phone. All right, how's now? There you go. You can hear me, but I can't hear you guys at all. [laughs] That's, that's been the challenge I've been having. 21:30 All of Texas has been having this problem actually for the last couple of days. Um, how about now? Can you guys hear me now? Yes. Yeah, we can hear you. I can't hear you. I don't know why. Send but not receive. Oh. 21:42 [clears throat] Here. Dan, if I, if I share my screen, would one of you guys be able to walk me through how to turn my audio on? 'Cause- No, we're li- we're live, Sumit. So let's, let's just, uh... 21:54 What I'll do is I'll, I'll chat you the questions over. How's that? Let's try. Ah. I'm gonna try this one more time. Try it till you fail, and fail and succeed. [laughs] Can you hear me now? 22:10 And, and again, so Dan, while we're working through this, I just started working with a company, Karna Health, uh, advising them, um, and they're trying to address the, uh, global pandemic almost of chronic kidney disease. 22:23 So the estimates are right now up to a billion people have some degree of chronic kidney disease, and 99% of people that have it don't know that they have it. 22:32 And so it's expected to be the number five killer in the world, uh, over I think the next 10 years. Um, and, and the preponderance of people that have this are sort of in developing countries, Africa, Asia, so forth. 22:46 And so, you know, the first challenge is identifying if they have it, and the second challenge is what do you do when you- Can you guys hear me now? Yes, we can hear you. Yeah, we can hear you. I can, I can hear you. 22:57 I can see you guys clearly. I have my Air- AirPods on. Ah, Texas is giving me a lot of trouble this morning. [laughs] Um... But Dan, what we're trying to do is we're gonna be using- What about now? 23:16 We got you with an echo. You got an echo? Oh, no. How about now? We're good. Oh, okay. [laughs] Sorry, guys. It's... Trust me, I have a house. We have Alexa. We have Siri. 23:34 We have Google, um, Meta AI on my phone, um, Gemini on my phone. I have it all. 23:44 Um, and then I have, um, Microsoft Copilot on work, and on Alienware I have ChatGPT-4. So- Y- you have the entire security state in your home. Wow. Yeah. [laughs] I work in tech. That's it. There's no way... 24:02 Okay, so sorry. I didn't introduce myself. Uh, Sumit Punjabi, head of AI for Optum. Um- I've been with them for two years. Uh, tomorrow is gonna be my last day effectively. 24:17 Um, so today I'm still part of Optum, sorry for any additional [audio glitches]. And then, um, I led the IPA consulting practice, uh, for NTT Data. Uh, I'm one of those Legacy Pro, um, acquisitions. 24:35 I was a principal consultant at Legacy Pro. Then EM- EMC acquisition happened, went to NTT, so a total nine-year stint. Um, before that I worked for KPMG for about three years. Um, 24:53 I'm from India, as you can tell. Um, I did my MBA from Indian School of Business, Hyderabad. Six Sigma Black. I have a lot on my cough, so Six Sigma Black Belt, um, and a bunch of stuff. But who am I? 25:08 I'm a soccer dad, a dad. Sorry, guys. I'm a soccer dad. Um, I love pets. I have a eight-year-old golden retriever. Um, I've had dogs all my w- uh, life. Um, essentially we... 25:25 I'm, I'm an Austinite. I'm a hardcore Austinite. I moved to Austin in 2011. Um, 25:32 and I hope to God that the next opportunity that makes me move would be much, much bigger than what I have, and that would be the only reason why I would move out of Austin, 'cause I have it really fun down here. 25:43 No, great, great. That's my introduction, guys. No, thank you. 25:45 So, um, so what do you think, you know, as, as a, as a technologist, but also as, as someone who's spent some time in highly regulated space like healthcare, what, what's your, what's your sense of just the general trends around gen AI and, and, and how it applies to the industry? 26:03 Uh, actually, before I begin, I want to comment on two things. One of the observations made was about the data leak that happened with the largest provider. Um, 26:17 I'm impacted by it too, but if you all have looked at the trend in the last two years, one of the larger oil and gas companies in Texas, 26:28 one of the largest healthcare companies in North America, and now an airlines company, all, all have got hacked. 26:35 So if you all are thinking it's all about coding and it's all about, uh, defining your architecture and all of that stuff, it's not. If you're thinking like that, then you're thinking wrong. I'll tell you why. 26:47 [laughs] Sorry, I... That was the first thing I wanted to comment on. Um, so I told you I use multiple devices. 26:57 Um, at work we have a multi-factor authentication which came in now, and public news, uh, not something that you don't already know. 27:05 Uh, MFA was one of, or multi-factor authentication was one of the major reasons why we had a breach. Uh, but among other things which I cannot disclose, MFA was a public news. So as a digital leader, I use 27:20 MFA on my laptop, the password of which is saved on my iPhone, for which I use my face to unlock. iPhone for work stuff and all my personal passwords are on my Google phone. 27:36 So I use Google Password Manager for my PlayStation, for my Alienware, for my VPN, all of that stuff. At work is what my employer provides, and I stack that on top of, uh, the Apple face authentication. 27:52 So if you all are thinking that cybersecurity incidents are all about architectures, y'all are 10 years behind. Trust me, guys. So that's [laughs] the first observation I wanted to make. Sorry. Over to you guys. 28:07 Chris, thoughts? I, yeah, I mean, the... 28:10 I, I also saw in the news, I think earlier today, that there was a major breach in the US military, one of the vendors to the US military, so compromised all kinds of stuff from DoD, et cetera. 28:22 And I, I, I think sort of what you're implying, Sumit, and I agree with this, is that you, you need to assume that you've been penetrated all the time, and then what are you gonna do about it? 28:32 Um, 'cause that's largely the environment we live in. The threat, I mean, they, they always said when I was doing cybersecurity, you know, the, the, the bad guys only need to be right once. 28:39 You need to be right all the time, so, um- So- We, we, we, we have so stripped away resilience in all of our systems in our desperate desire to reduce costs that, you know, we have none. Resilience, that is. 28:53 No, you're right, Chris. That's absolutely on the money. Um, along with resilience, one of the points that you made that I really liked, so when you talk about 29:02 cybersecurity, and I- I've seen your profile on LinkedIn, uh, you've been a place or two. [laughs] Um, you know, we have war rooms full of warriors who sit there and 29:16 test our systems, which is called ethical hacking, right? 29:20 Now, it's safe to assume companies like Apples and Googles and Amazons and UnitedHealth Group, and whoever you talk about, the top Fortune 50, without disclosing anything, have a team, have a focus on this, right? 29:33 That's why they have a chief strategy officer. That's why they have a chief digital officer. 29:38 Digital is all about cloud, AI, gen AI, but the strategy officer is the one who makes sure that the data is protected, is safe, is hosted in an environment as per the ethical policies or policies of the company, right? 29:53 And then it's protected from breaching, so making sure all your stage gates and ethical hacking is kept at bay. But let's say really smart people still get through that stage gate. 30:07 That's when it comes to the end user, right? Multi-factor authentication. So if you look at the front end, you have a whole army fighting this, right? And if y'all have to imagine, um- Russia-based actor claimed threat. 30:24 So y'all can understand the scope of this issue that we're talking about. Um, but I, I'll tell you this. Today I'm leaving, as y'all can see, I'm leaving UnitedHealth Group, but I'm still wearing the colors, right? 30:40 I'm still wearing the colors, so I'm not leaving disgruntled. I'm leaving in really, really good terms. Uh, y'all have to understand that 30:50 as of today, and again, public news, not nothing new, um, UHG has already spent about $1.6 plus million dollar... billion dollars in, in this recovery effort. That's more than a startup in Austin earns today. 31:05 An average startup can't even aspire to earn that kind of money that UnitedHealth Group fronted, right? So- The, the challenge, the challenge with any of these systemic failures is that when they... 31:16 You can't really estimate their cost, so when you're doing that sort of risk assessment ahead of time, it- it's almost a meaningless calculation. 31:23 And you're not gonna know until you actually hit the failure and, and it almost always is gonna be far greater than you anticipated in the, in the beginning. 31:30 So to your point about having a strategy person that's trying to lay this stuff out, it... You're, you're, you're building a castle on quicksand to a certain extent, right? 31:39 As you're building all these defenses to keep you protected and so forth, you're literally building it on quicksand because you can't quantify the risk until you experience the risk. Actually, you can. 31:48 That's right, until you've experienced the risk. I have. I am- [laughs]... an cybersecurity intelligent automation watchdog. That's been my role for the last year. So when I tell you UHG did not... What you... 32:03 Also, right, you're saying, did not leave a stone unturned. They did not. 'Cause y'all have to understand, if y'all go into my personal LinkedIn and, and, 32:14 uh, Facebook, y'all will see that I have been coached and mentored by Vietnam War veterans. I have 32:24 the Chicago Police Department who has coached and mentored and created some of the programs for the organization where I volunteer. 32:31 I'm not gonna name names, uh, but some of y'all can see photos in the background, I guess, in my office. But [laughs] um, I guess where I was going with this is 32:42 the magnitude of this cybersecurity incident is beyond our comprehension. That's what I'm trying to say, that every textbook that was written on IT infrastructure, every page 32:58 was outlined and reviewed for the incidence response. 33:04 So if you were in the IT field and a nerd from the '70s, everything you learned in the textbook, including how to unbolt and bolt stuff back on, build the racks back on, 33:17 turn them back on, put the connections back on, every single thing went through checks and balances. So trust me, guys, [laughs] when you say 80% of the data is managed by one provider, 33:31 there is a reason why it's being managed by one provider. Well, and, and if I may, that, that's the conundrum of doing this cloud stuff or whatever. 33:38 To your earlier point, an Apple, a Google or whatever, they're gonna have the, the best security people, the best tools, the best everything available. 33:48 They're, they're gonna be able to do far better in terms of the resources and tools they have a- accessible to them a- a- as any company not in the Fortune 100, right? 33:57 Any company not in the Fortune 100 will get better security infrastructure tools, techniques from going to the cloud. The problem is, is that when, when that inevitably fails, everyone is impacted. 34:09 And it's not a question of if it will fail, it's a question of when it will fail. It's just the accumulation of rounding error. 34:15 And, and if we never plan for those events, then necessarily the, the negative impact when it occurs is gonna be greater. It, I mean, it's sort of just math. Chris, you, um, you said it's eventually gonna fail. 34:29 I was gonna step away for two minutes 'cause I wanted to show you something I found while cleaning my closet today. I found my iPod Nano. [laughs] iPod Nano, and I have a charger. 34:41 When iPod Nano was launched, did you think it's gonna fail and this is gonna come out, or your phone's gonna come out? Like I said, we failed, but we failed with pride, and we put all our forces behind it. Like- Yeah... 34:56 there are people who, Chris, you know this. There are people who are in really, really high classified positions, right, whose healthcare is managed by the leading healthcare provider in the country. 35:10 So I tell you this, guys. I know there's a lot of jitters about the cybersecurity incident that happened. Trust me, I would resign if UHG wouldn't have done what they did today. 35:23 And I'm a proud UHG Optum alumni, I can tell you that. It's not... It's beyond our control. It's beyond our forces. I know we can sit here and talk till the cows come home, as my fellow Texans say, but 35:38 y'all haven't seen what we have seen. [laughs] We've had people volunteer to run and get pizzas for our employees 'cause that's how overworked they were. 35:48 People have worked on 4th of July, Memorial Days, holiday weekends, with people bringing in pizzas and chicken wings. And it was a festive environment. When you bring your troops together and you create a war room, 36:03 you're an ex-military professional, service professional. When you bring your athletes and your A team together, you're revolutionizing the industry. 36:13 Like, every gentleman and lady who's been part of this Change Healthcare incident is gonna make a difference in IT industry next 10 years. That's my prediction for you guys 36:24 So Sumit, uh, in the remaining time we have, I wanna make sure we get to the GenAI topic and, uh, you know, I'm curious as to, you know, what you, what you see, what you're kinda excited about, uh, uh, as it relates to GenAI, and then, you know, any, any, any s- any specific use cases you can, you can share for anybody in healthcare watching this. 36:43 Okay. So I will switch to my phone one more time, and if you can't hear me, go with it. Um, or one sec. Well, y'all were talking about an article about Llama. Llama existed months ago. Mm. It's on WhatsApp. 37:02 You know how I create my presentations for Optum? Midnight, when my kid goes to bed. WhatsApp Llama, I asked, all I asked, 37:12 USA healthcare democramatic, uh, demographics data, and I know you guys can't see this, I apologize. Age 65 plus years, 16%. So all this data, it's giving me demo... Let me see if I can move this guys. Nope. It's tough. 37:29 Anyways, um, demographic data. There you go. So here, there you go. I think I got an angle. So if you guys can zoom in, USA healthcare demographics data. [phone chimes] Look at this. 37:43 Midnight, after my kid went to bed, I didn't even have to Google. On my phone, on my personal device, managed by face authentication. I create business plans doing this for my living, I guess. 37:58 Then next question, Medicare Advantage budget cuts. No question mark, no prompting. WhatsApp Llama AI is giving me the answer. Again, I lost you guys. I don't know if I can... 38:11 I'm trying to find an angle where I can get this. Sorry. But that's, that's the demo I was trying to get to. Um, it shows you everything. 38:20 It talks about 2025 Medicaid Advantage payments, it talks about ins- impact on insurers, talks about recovery of overpayments and, and everything. Then I said, "What are the top five companies in USA?" 38:32 I didn't even say, "What are," I said, "Top five companies in USA." UnitedHealth Group, CVS, McKesson, Censora, and Cardinal Health. Everything on my phone after midnight, 38:51 WhatsApp Llama. The whole logic of working remote has been revolutionized by GenAI. Anything you can imagine, anything you can imagine, that's the prompt. Imagine 39:08 and put it, it'll work. That's how you train your data. So when you talk about how do you manage the data, right, and cloud hosting, 39:18 like I said, imagine is your prompt, and you literally type it out, "Show me Willie Nelson's guitar." Go to Gemini, imagine Willie Nelson's guitar, and see what it creates. 39:31 It'll give you a logic, a text, or if you say, "Show me an image," or something like that. Again, I'm not a geek. This is things that I practice and, and, and work on my own, right? 39:40 It'll generate an image of Willie Nelson's guitar, right? So I guess where I'm going with this is GenAI is gonna transform the way we think work. 39:51 If you're still thinking traditional IT infrastructure, you're probably missing the boat. Cloud hosting is an option, but again, you have to work with your chief strategy officer and your security officer to understand 40:09 where are your stage gates, those breaches. Those breaches could be defined on cloud, on-prem, or a hybrid, right? 40:21 Essentially, that's what a top strategy, chief strategy officer incident response team does, right? 40:28 So cloud is a way to go, sure, but it also depends on how you are as an organization and how your infrastructure is being set up. [clears throat] Sorry, I, I know I'm a talker. No, no, no. Well, interesting. 40:44 So Dan and I actually were on a call with a, a organization yesterday where they were, um, recogn- they, they were acknowledging, "Hey, we wanna use generative AI, but we're heavily regulated." 40:55 And in fact, new regulations in their jurisdiction prevented the use of PII, personally identifying information, in programming generative AI. 41:05 Which to me stri- it, is clearly a challenge in healthcare because the whole point of healthcare is, you know, I need to diagnose people, so PII is, is invariably part of it. I mean, how- Right... 41:15 have you been addressing any of those issues of- Yeah. So... sort of the privacy paradox? Let's think about how you can use GenAI, right? So I'm an employee of UNG and Optum, right? Today, 41:29 my internal search has been transformed by GenAI. Now, I don't have to click on the browser tabs, scroll down on the menu. I just go to the search bar and say, "Find my paycheck" or "Find me this knowledge article." 41:46 It runs through all your search results because it's churning. That model is working through all those tokens, all those vectors to get you a result, and it'll give you a very tailored result. 42:00 Then GenAI gives you a thumbs up or a thumbs down, right? So it, it did all that fancy geeky stuff for you and gave you the power to decide if I did good or not. 42:13 So the way I use GenAI, right, I showed you, I use GenAI every day on my phone. I, I should probably not admit this, not safe for work, but I've used it in my bathroom too. So [chuckles] that's how much I use GenAI. 42:25 But, um- The thing is, um, things are changing. The way we work is gonna change. 42:35 I have kid you guys not, I have created business plans formatted for Excel on WhatsApp Llama, copy-pasted it into Excel to test how the formatting works. 42:49 And then I'm using, I'm learning how to use ChatGPT to run and create a model, financial model. Now think, you hire smart people to do ROI analysis. 43:01 The smartest guy in the room will create a ChatGPT model to build an ROI analysis. He won't even think what are the factors in an ROI analysis. 43:10 He will go in and say, "ChatGPT, factors, cost, labor cost, infrastructure cost, hosting cost, license cost, other cost, six parameters, build me a model, ROI analysis model," and ChatGPT will run to work. 43:25 So that's the transformation power of ChatGPT. So, so this, we, we started the conversation earlier, and I've brought up this point extensively. 43:33 This is the polarity of this technology where you can either have people be more efficient or more effective, right? 43:42 The more effective people are those who have l- someone who wouldn't know how to do an ROI calculation or, you know, a business plan like that, they could do one. Yeah. 43:51 Um, someone who's good at it, such as yourself, can be much more ef- ef- efficient, [laughs] right? You can do it much more quickly. But the, but the- No, I'm just a gamer. 44:01 Well- So I like a problem when I try to solve it... 44:03 the, the challenge of the technology is, is this, however, ha- half the people that use it aren't gonna know the correct answer when, when ChatGPT or Llama is giving it to you. 44:13 And, and because of the feedback mechanism in, in generative AI, if w- when they accept the wrong answer because they don't know any better, they're actually training the LLM to be more deceptive and stupider. Yeah. 44:28 So, so- No, no, no. Hang on. Hang on... in, in, in healthcare, we can get two kind of results. Lean Six Sigma. Lean Six Sigma. Garbage in, garbage out. Yeah, yeah. Op- optimally. 44:35 Optimally Remember, Gen AI models are trained on every single available public source data, which means it's smarter than the three of us in the room, right? It is. Mm, I- And probably smarter than most of our users... 44:47 strongly disagree with that. You haven't tested it. No, I have. I will show you evidence- Yes... that it is smarter than me. And I, I think I'm a smart guy, don't get me wrong. [laughs] Yeah. 44:59 I don't know if s- s- 75 billion tweets about the Kardashian does not make it smarter. [laughs] She's a- she's making money. Her followers are making money. While we sit- You know-... 45:08 and talk about ChatGPT, they're actually making money for America out there. [laughs] So, hey, I'm on Instagram. I'm a boxer. I stream my boxing workouts on Instagram only to very close friends. 'Cause like I said, 45:20 I work with some really, really high-profile people in Chicagoland area. There's a reason why I don't get very active on LinkedIn and Facebooks and, and such, 'cause 45:32 I can't risk exposing some of the people that I work with. It's a huge, huge... It's not even cybersecurity, it's- But Sumit, but, but, but, but-... national security risk... but you, you work for a healthcare company. 45:44 So, so what about, what about the risks? And, and I'll, I'll use your, your, your, your boxing as a, as the analogy, right? Everyone's got a strategy till they get punched in the face, right? So what, so what do you do? 45:54 What do you, what do you do to, to get all these benefits... Um, listen, re- regardless who's smarter, right, there's obviously benefits. How do you weigh the benefits versus the risk, uh, to a, to a- So-... 46:06 to a, a publicly traded, highly regulated organization? I'll tell you two things. Two, two quick answers. If you get knocked down, you get up and punch harder. If you don't have that courage, you're not a leader. 46:20 Step aside. That's observation one, right? Second, your models are only as smart as you. Human in the loop. AI will never become transient and take over. 46:34 The day it becomes transient is gonna be 400 plus millions years ago. You know how I know? I read books called Survey of Hinduism. 46:45 Hindu histo- The calculation of the distance from Earth to Moon, that entire calculation has been written in Tr- Sanskrit in Hanuman Chalisa. 46:57 While you guys are sitting and using science to calculate it, an ancient text that people worship has that entire calculation written in Sanskrit. So guys, like I said, 47:10 this is a tipping point in all our professional careers. 47:14 Like, I am now counting and hoping and praying that I will bank some really good stuff from here, some good respect, 'cause good work experience, some good money, some good travel. 47:27 That's what I wanna do now, 'cause finally, I feel like I've figured out something that works for me. That's Gen AI. Sounds good. Sumit, thanks so much for joining us. Really appreciate it. 47:40 Good luck with the, uh, next gig. Hopefully, we'll have you on again soon. Thank you, guys, and I apologize we had some technical difficulties. Um, Texas has been hit really hard, especially Houston. 47:51 Half of Texas was down because of Spectrum, uh, outages. Believe it or not, the day Spectrum was down, I couldn't order stuff on Amazon. Oh my God. That's how widespread the impact was. 48:02 So consider- My wife, my wife would've been very depressed... [laughs] really. So anyways, thank you, guys. Thanks guys so much. I'll talk to you guys soon. Take care. Cheers. Chris, uh, little bit of a rollercoaster. 48:11 Sorry for the tech issues, everybody. But, uh, yeah, interesting, interesting conversation. I'm, I'm still, I'm still... 48:18 It's very interesting, you know, uh, obviously he's defending some of the things that are going on there. 48:23 But, um, I also, you know, I, I also wonder, you know, you talk about just policies and governance around, you know, how people are using Gen AI. It's, uh... 48:35 I, I agree with him that it's inevitable, uh, but it's definitely interesting. 48:39 Well, it, people, uh, because I'm, I'm finishing up th- my upcoming book on AI governance, and everyone's like, "Well, you know, what do we have to do differently?" 48:47 It's like we, we just have to do what we've always known we needed to do. We just needed to do it more. 48:53 And, and so all the, all the effort and, and costs and so forth associated with governance, uh, it's, all, all that we've done is amp up the risks associated with it, and so we need, we still need to do the right thing. 49:05 We just need to do it more so. Yeah. Yeah. Interesting. Well, thanks for joining, everybody. Thanks for sticking with us. We're back in two weeks, uh, Wednesday noon Eastern, 9:00 Pacific. 49:14 In the meantime, check out urpaai.com for more resources, or schedule a free health check on your automation or AI program. Chris, have a great week, buddy. You too. Thanks, everybody. 49:25 [outro music]