GetixHealth Podcast

Charting the Future: Insights from Chief Information Officers in Healthcare

Alex Oey · August 21, 2026

Watch on YouTube Listen on Spotify Apple Podcasts

Transcript — follow along with the video

Matthew Chambers(0:00) I think we might need pastoral care.
Eric Reid(0:02) I think we do.
Matthew Chambers(0:03) Yeah, I'm Matthew Chambers. (0:05) I'm the chief information officer for Baylor, Scott & White, and I apologize, I've just got a priority one text here, so I'm on call 24-7. (0:14) So at Baylor, Scott & White, we're, I suspect y'all know, we have the largest not-for-profit in the state of Texas, about $13 billion in revenue, and I think 80 hospitals, something like that now, and thousands and thousands, tens of thousands of employees.(0:31) We basically come south of Austin to north of Denton now, I think, all along 35, (0:40) and I actually started at the Scott & White side a long time ago there in Temple, and then (0:45) took over when we merged in 13, and overseeing a lot of change in our time, including, (0:52) you know, a major initiative several years ago to outsource a lot of our IT, or I'm sorry, (0:58) shared services migration, a lot of our IT, and continue to try to do good things and support (1:05) the idea of no margin, no mission.
Eric Reid Good morning, all. (1:09) I'm Eric Reed, vice president and chief technology officer at Crystis Health. (1:14) Crystis is based in Dallas, has facilities throughout Texas, Louisiana, New Mexico, and we also have a Latin presence in Colombia, Chile, and Mexico.(1:27) I've been in healthcare my entire career, came out of college, was grounded to purpose. (1:31) I find that doing good things in my role helps people that are in our facilities in a position of need. (1:37) If you're in a Crystis facility, it's not because something great's going on in your life.(1:41) You probably have something that you need our help with, so grounded to purpose.
David Stuart(1:48) All right, I'm going to start us off with our first topic here. (1:52) I'm just going to talk about the elephant in the room and talk about how the Change Healthcare Breach, and most recently, the Ascension Breach, has impacted your day-to-day operations and business.
William "Bill" Phillips(2:05) I guess I'll go first with that. (2:07) If you remember the headlines when Change Health happened, the first thing that was stated, a headline pops up in the news, pharmacies nationwide are down, and then it continued. (2:19) Then it rolled into, okay, it's a nation-state hacker.(2:22) It's nation-state, now we're in trouble, it's a foreign country. (2:26) Then that changed. (2:28) Then you started seeing announcements, okay, AHA's involved, THA's involved, CMS is involved, FBI is involved.(2:36) This thing just did this in a very rapid rate. (2:40) The first thing we do, and unfortunately, one of the weaknesses when organizations or yours and ours vendors are hacked, is there's no real direct, immediate notification to us, the customer. (2:53) You have to find this out in one way or another.(2:56) We've had other vendors that someone has called because a wife worked there and said, hey, I'm just letting you know inside, this company you use, we're under attack right now. (3:06) There's no good way. (3:07) We subscribe to services that help us find this with vendors, and we find it a lot.(3:12) The first thing we did is, obviously, we cut email ties, we sever every connection we have. (3:18) Then I have a canned letter that goes out with my signature that our legal department has drafted. (3:24) It basically says, we're aware of your situation, but we're going to stay severed, and it's your responsibility to give me regular updates.(3:32) It's also your responsibility to let me know when it's all clear and safe to go back in the water. (3:38) Then we also had to deal with the CIOs, and I think we all had the same problem. (3:43) Oh, it's the Change Health side.(3:45) Don't worry about the Optum side. (3:47) That's all good. (3:48) Yeah, right, right?(3:49) They're connected companies. (3:50) No, we kept everything severed. (3:53) Of course, impact on claims.(3:55) We're also, we're depending on what connections our third parties use, like SureScripts as an example. (4:02) We're calling them, where are you at with this? (4:05) Even when the, you know, what's safe to go back in the water came out, if they didn't clear, we didn't clear.(4:10) Same with Epic(4:11) Everyone was kind of playing the game. (4:13) What do you think we should do?(4:15) No should, it's what we will do. (4:17) So we stayed severed, and of course, I think every day I would get from the CFO to say, hey, when do we open it? (4:23) When do we open it?(4:24) And you know, we can't predict that, right, until, we're not going to open the floodgates until we feel everything's safe. (4:31) And then you look at what came out, and it was a black hat hacker group that is targeting healthcare. (4:38) Then you heard of Extension.(4:39) This just happened. (4:42) If your CIO or your leadership doesn't have good relationships with the FBI, you need to get one. (4:49) I found out about Extension before it was out.(4:53) I got a call from our local FBI. (4:55) You know, it's always, you see FBI come across your phone. (4:59) Oh yeah, it's, you know, is it fake(5:01) No, hey Bill, I'm checking on you. (5:04) You know, you can't look around. (5:05) Why are you checking on me?(5:07) Is the black car out front waiting for me? (5:09) What'd I do? (5:11) And they were like, no, look, this is what happened.(5:14) There's a massive hack on a very large, and they gave me, you know, how many hospitals, how many clinics, and it didn't take me long to figure out who it was. (5:21) And they said, this isn't out yet, but you guys need to watch it. (5:25) So then we went to our next heightened alert.(5:28) We called our vendors, put us on high alert, made sure we were scanning. (5:34) And that same day, later on the day when that happened, I got a call from our local federally qualified health center. (5:41) They were hacked that same day.(5:42) They're on paper. (5:44) I called a leadership meeting and said, you know, this is, we've got more and more hacks coming out. (5:50) And if you saw the news last week, there's another group that just announced, and now they're targeting healthcare again.(5:57) So this isn't going away.
Matthew Chambers(6:01) Yeah, what Bill said, I think he covered it there. (6:05) But in all seriousness, I think this is, cybersecurity is probably the most important thing we're dealing with now as CIOs, unfortunately. (6:17) Excuse me.(6:18) I think about what we do is play in offense and play in defense. (6:22) So, playing offense is implementing new technologies. (6:25) We'll talk about AI and doing great things to help our employees and our customers and patients, but playing defense is trying to prevent bad things from happening like this.(6:34) And it feels to me like we're talking about defense all the time now, and it's kind of sucking the oxygen out of the room. (6:41) We were pretty fortunate. (6:43) We weren't too badly impacted by change, but kind of ran through the same fire drills that Bill described.(6:49) Same thing with Ascension. (6:51) We immediately sever all connections. (6:54) But what it's really done is, we've already had cybersecurity presented in our audit compliance committee at the board of trustees.(7:04) I hired a chief information security, you'll hear us say CISO if you don't know, it's chief information security officer. (7:10) And I hired one two years ago, convinced her to come in. (7:14) And the question you get from CISOs most of the time is, well, do you have organizational support?(7:19) I was like, we've got more support than I want. (7:22) I mean, it's all we talk about in audit compliance, and it's number one on our enterprise risk management plan. (7:31) And I feel like it's probably the same for most everybody.(7:34) It's what we're talking about now. (7:36) And you just can't be too prepared. (7:38) What I've realized as we're doing more and more tabletop exercises, everybody's doing these drills to prepare.(7:46) And what I'm realizing is our business continuity plans are really more focused on historical types of issues. (7:55) In our disaster response plan, where we've got two data centers prepared to fail over. (8:02) They work great, but what they were really intended for was natural disasters.(8:06) If a plane crashes into one of our data centers, we can fail over to the other in an hour and be up and running. (8:12) But if a breach or an attack or an encryption or what have you shuts down your whole network, most of today's healthcare systems are not really architected for that. (8:25) And so if you're responsible for helping your CIO or your organization fund IT, you're going to have to take a hard look at how much you're spending on it.(8:36) And unfortunately, it's going to increase because it's like an arms race with the threat actors out there, unfortunately.

David Stuart(8:46) And difficult to find that balance between the business operations and securing the business operations.
Matthew Chambers(8:52) It is. (8:52) It slows down everything. (8:54) We've got to get this contract signed.(8:57) Okay, well, you brought me somebody that has four employees. (8:59) They're like a garage band. (9:01) And I don't know that I trust them to have seven million patient records.(9:08) So yeah, it does. (9:09) It really does require a lot of thought and assessment. (9:13) Some things I'd tease out of what was said already is there are, in the healthcare industry, some de facto leaders.(9:26) Clinical, Epic, and or Cerner. (9:29) So if everybody's on Epic, that's great as long as Epic's having a good day. (9:35) Switch that back to, okay, change(9:37) Without realizing it, we're all relying upon, many are relying upon one or two third parties. (9:45) That sounds great. (9:46) You've consolidated, you've got processes, you've driven down costs, all the things that we as leaders would want to do.(9:54) How many remember to check that box that says, oh, by the way, if change healthcare is offline, what are we going to do to survive as a business operations team? (10:04) Not too many people, you just don't think about those things. (10:08) You expect a third party cloud base to be always available, always.(10:14) It's just the flip side of that cloud offering is you expect certain things to come with it. (10:21) As long as those assumptions hold true, you're good. (10:24) Well, what if that rainy day occurs?(10:26) What do you do? (10:27) So that's business continuity. (10:29) For us, that has been an eye opener, business continuity.(10:35) We were able to acknowledge the risks, same thing, FBI conversation, everybody, you know, Chime and other organizations got the information out right away. (10:44) Out of an abundance of caution, you sever all the connections. (10:48) In our case, sometimes too many connections, turn everything off, wait a minute, wait a minute, dial it back, get it right.(10:55) That takes a little bit of time, but you have to, you know, if the wolf's at your door, you have to do something, right? (11:01) We were able to change our providers when it comes to payment processing, claims processing. (11:10) And for us, that was not too big a deal(11:12) We have things internally architected, so we could make a change in about four days. (11:18) So it didn't have the biggest impact on us, but it certainly did on all of our patients, our communities, us as employees. (11:25) I personally went to fill a prescription at CVS.(11:28) CVS says, I'm really not sure who you are. (11:32) I'm exaggerating slightly, but essentially they couldn't, they said, can you come back? (11:37) And I said, no problem.(11:39) So things like that are real world. (11:41) Change is the one that's got all the attention. (11:43) Ascension is the new one.(11:45) What I would tease out of that from like a CIO, CTO perspective is the footprint is the same. (11:53) You have interactions that are outside of your four walls with third parties. (11:59) You all have points of presence.(12:03) You're on the internet. (12:05) You say, interact with us. (12:07) You have IS support teams that use portals.(12:10) You might use Citrix. (12:11) You may not. (12:12) You might have all sorts of technical stuff.(12:15) You have people working from home and interacting, coming through your firewalls and such. (12:21) And while it's not simple, it's not exactly complex. (12:27) It's a lot of hard work.(12:29) It's a lot of things to keep track of, but it's essentially in many respects, blocking and tackling. (12:35) If you're going to provide something external, you must keep it patched. (12:38) And you can't have wishful thinking.(12:40) You can't just say, well, I'm pretty sure that team's keeping it patched. (12:43) You have to say, no, actually you need to keep it patched. (12:46) And you have to prove that you keep it patched and have like a maniacal focus on security.(12:52) There is a natural tension between business operations, productivity, and security. (12:58) Of course there is. (12:59) That's just the way of the world.(13:00) Healthcare may be the last industry to be targeted, but it's going to be the one that might have the most bang for the buck. (13:07) Think of the liability associated with patient information. (13:11) Financials is one thing, no doubt.(13:13) Patient information, that's the real deal. (13:17) So the advice I would impart would be, be highly intentional. (13:21) If you're putting anything on the web, if you're interacting with anyone via third party, whether it's explicitly via an API, via PN, and so on, be highly intentional and prove that you are secure.(13:35) Limit the number of browsers that you support. (13:38) Limit the number of remote control products that you support. (13:41) Have some that are officially supported and the rest you should be blocking, and so on.(13:47) Have a auto patch program in place. (13:49) Do the things that will reasonably keep your business enabling, providing care, in our case healthcare. (13:58) Servicing customers securely.(14:02) Don't overlook that little detail. (14:04) I'll wrap back to what I said. (14:06) You must assume that something will happen at some point.(14:09) It was alluded to. (14:11) We need to get it right all day, every day. (14:16) Hackers need to get it right one time.(14:18) One time. (14:20) I want to double click on something you said there about consolidation of vendor support. (14:28) Because actually, after the dust settled, that was the first thing that our CFO, I don't know if she's here, huge respect for her, she came back and said, are we too consolidated?(14:39) Yes. (14:39) That's a big part of business continuity is in any third party vendor space, do we just have one vendor? (14:48) It's funny, I was actually up at Epic at a conference and one of the senior executives there, I think it was the week of Ascension, and they said, no, it's change.(15:01) It was the week of Ascension, it was two days before Ascension and two weeks after change. (15:05) One of the senior executives there at Epic was like, what happens if it's us? (15:08) That's right.(15:10) They're just as concerned as the rest of us. (15:13) In essence, what you inadvertently create are what are called single points of failure. (15:18) If you have a single point of failure, that's okay.(15:21) If you recognize it as a single point of failure and you have contingency plans, if that single point is unavailable. (15:28) Back to that intentional word, just be very, very intentional about your business operations.
David Stuart(15:35) I often get calls on a Saturday or Sunday from my CEO where he's watched something on 60 Minutes, especially during this change. (15:44) He's always asking me, David, can this happen to us? (15:47) My answer is always, yes.(15:50) How do we protect ourselves from it? (15:53) Continuous education, continuous layers of security. (15:57) We can't just stop.(15:58) We've got to keep moving with that. (16:01) Did any of these result in breaches for your facilities?
William "Bill" Phillips(16:06) I tell you, I unfortunately lived through one in 2020. (16:10) If I had my slide, you'd see I caught the perfect storm. (16:14) We were implementing Epic, we had a pandemic, and two weeks after our go live, we had a cyber attack.(16:22) Not a good time. (16:24) How did the cyber attack happen? (16:27) human, right?(16:29) Stupidity. (16:31) What basically happened is our partner, one of their physicians, we tracked the hack and we tracked it down to a user ID. (16:41) We found out it was a physician.(16:43) I don't know if this physician is a good guy or a bad guy, but I need to know what he knows. (16:49) We called him in with the dean of the school next door. (16:52) It was like an FBI interview.(16:54) I'm asking what he's been up to, where he's from. (16:56) I got to see his resume, his background. (16:59) I knew he traveled overseas(17:01) I don't know if this guy is where he's at. (17:03) Ultimately, find out that he was hacked at his house. (17:08) What he had on his home computer was all of his user IDs, passwords, bank accounts.(17:14) University Health, here's my user ID and my password, and I'm a doctor. (17:21) What happened, the low-level hacker got that, took it to the dark web, sold it to the highest bidder, and one of the worst hacking groups had a hold of us. (17:31) We had 8,000 servers infected.(17:34) I was getting ready to turn on multi-factor authentication, so that gave me all the excuse to pull the trigger and make that happen. (17:43) Long story short, we ultimately got out of it with no loss, no encryption, no data exfiltration. (17:49) We were one step ahead of them.(17:52) We put enough down that when they detonated, it failed. (17:55) Microsoft told me, you're not getting out of this. (17:57) Cisco told me, you're not getting out of it.(17:59) Everyone we had on a retainer told us, you're not getting out of it. (18:03) We were very fortunate to get out of it. (18:06) What I mentioned in my slide, I show that time, and I have a picture of me sliding down a cheese grater, because that's what life was like.
Matthew Chambers(18:16) Oh my gosh, that's hilarious. (18:18) Yeah, no, you were fortunate.
William "Bill" Phillips(18:21) extremely.
Matthew Chambers(18:22) The next time we have an issue, I'm calling you for advice. (18:26) The biggest impact we've had the past couple years is third-party vendors. (18:35) It feels like common sense if you're getting beat up by this every day, but the education thing is important, because somebody will bring you something and say, I want to give, literally, I did a count, I don't remember how many, 17 million patient records.(18:49) We want to give these people 17 million patient records so they can dig through our data and see what they can find. (18:54) I was like, I know what they're going to find. (18:55) You're going to find it on the dark web, and we're going to be absolutely in a bad situation.(19:02) We've had data exfiltration from third parties, but then we end up, they actually own the breach. (19:11) They have to report it, but we have to report it to our customers as well. (19:15) You try to make it clear, but at the same time, a patient might look at that and say, okay, well, you gave my data to somebody who didn't protect it, so what are you doing to help me?(19:25) But yeah, to me, that's very impactful as well. (19:29) For us, publicly available, we did have a minor breach about two years ago. (19:35) We had internal controls that detected that something was going on.(19:38) An unusual attempt to elevate a privilege detected it, got on, shut it down in a few minutes, but it did occur, and it traced back to, hold on, we had something on the internet that was almost completely patched, but not quite, and ironically, we had a change management ticket in to apply that patch, and so it was a timing thing. (19:58) So I've already talked about patching. (20:00) Other things to think about are that dark web and what does that mean?(20:07) So anybody, whether it's a credit card or bitcoin, they can access most things, things you wouldn't think are available, they are. (20:18) And despite the, most people think a yellow sticky pad is the best possible place to put your passwords, usually under your keyboard, that's physically safe, I'm joking, and so on. (20:29) It's safer than online nowadays.(20:34) Multi-factor authentication, you must do it. (20:38) Long, complex passwords, 16 character minimum, complex, uppercase, lowercase, special characters and numbers, you must do it. (20:46) You can take the sting out of that by doing things like Windows Microsoft hello, use a pin for that physical local device, you can strike a balance around being secure and yet not being so secure that you can't get your job done.(21:01) I try to use an analogy when I talk to security people, it's like, well we all have perimeter security, even in our personal lives, right? (21:10) You've got gates, you lock your doors, do you lock your bathroom doors? (21:15) Do you lock your bedroom door?(21:18) Do you lock your pool door? (21:20) Pool door is closed, is it locked? (21:22) I mean, you can take it as far as you want, you know, but you have to think about that.(21:27) The other aspect I would ask everybody to think about is what's called social engineering. (21:33) Someone would call your service desk and say, hey, I'm physician ABC, I need to reset my password, you know, you probably have some protocol for that, hopefully it's automated and you don't have humans in line because humans can be tricked. (21:49) Isn't that what happened in that casino in Las Vegas?(21:51) Yes, humans can be tricked. (21:53) We are in a service industry broadly, people want to be helpful, right? (21:59) So someone calls you to get help, the person says, you bet, how can I help you?(22:03) Well, I'd like to change my password. (22:04) Okay, doctor ABC, what's your piece of information you're supposed to know? (22:10) Hemming and hawing, I don't want to give you that piece of information, so can you just reset my password?(22:16) And don't forget, I'm a physician and you're holding me up, right? (22:19) So now the pressure's on and if you're a service desk person sweating bullets because you don't want to lose your job, you know, this is the way that social engineering works. (22:29) You have to be cognizant of that, coach, prepare people, give them an out and say, I know you're a physician, I can't help you right now, I'll help you a different way, get you to this person.(22:42) I'll get someone to come visit you and, you know, implicitly prove that you're a doctor ABC but under the guise that we want to give you a white glove experience, something like that. (22:52) It's very easy for a person to talk their way in the door if you're not careful.
David Stuart(22:58) And that's usually in my executive briefings, you know, when asked what's the number one vulnerability? (23:03) It's always people.
Eric Reid(23:04) Oh yes.
David Stuart(23:04) And so it's just a continuous education of that, even if it's painful to continue to educate. (23:11) But in our organization, we've gotten to the point now where everyone is so scared to click on a link that they're sending it to our InfoSec email to say, is this legitimate? (23:23) And that's okay.(23:25) Every single one of them, we answer, we point to, here's how you can tell. (23:29) There's never any of you should have known this or you should have learned this. (23:33) It's that continuous education.
Matthew Chambers(23:35) You get an email from your leader who probably you want to make happy in some way. (23:41) Hey, I sent you the spreadsheet. (23:42) I think it's right.(23:43) I need, could you double check that for me real quick? (23:46) Sure. (23:47) I'll double check that for you.(23:48) Okay. (23:48) Do you just download a virus? (23:49) My favorite is when I get the text, hey, it's your CEO.(23:52) I need some gift cards for it.
Eric Reid(23:54) Yes.
Matthew Chambers(23:57) If I'm not busy that day, I'll be like, sure. (24:00) Where are you? (24:01) How many do you want?
Eric Reid(24:02) Yeah.
Matthew Chambers(24:04) I'll stop at Walmart and grab a bundle. (24:08) The one, you know, the clicking on the emails thing, that's absolutely true. (24:11) Like around Christmas, you know, it starts around Christmas every year of, you know, here's your package.(24:17) And first of all, I'd advise you don't use your work email for your FedEx packages and your Amazon packages and all that. (24:24) I mean, it's just, because now we filter. (24:27) I mean, that's one of the things that we've got.
Eric Reid(24:28) Yeah.
Matthew Chambers(24:29) I mean, we, anything that's suspicious, we'll block it. (24:32) So like, well, I didn't get that email. (24:34) And I'm like, yeah, it was, we didn't, we didn't like it.(24:36) So we killed it.
William "Bill" Phillips(24:40) And you know, the one thing we didn't bring up, which is my pet peeve is medical devices. (24:45) Oh gosh. (24:46) So the average, just so you'll know, the average medical device has 6.4 vulnerabilities per device, per device. (24:55) And that means they're outdated OSs. (25:00) You can't patch them and there's no policy regulation or anything to help us. (25:07) The FDA puts out guidelines and they're putting out process for newly created, but there's billions upon billions of billions and billions of pieces of medical equipment.(25:16) And if you ask me what keeps him up in the cyber world at night, that does. (25:21) And it's very difficult to protect. (25:23) And I mean, I recently went through one because we get beat on right.(25:28) Constantly. (25:29) Hey, we just bought this ultra or, you know, we got this, we had one and we got an ultrasound. (25:34) We, we cardiology and we got to get this online and connected built like today.(25:39) We looked at it, hadn't been patched in three years. (25:42) It had 3,000 vulnerabilities. (25:45) And we said, no(25:46) And we worked with the vendor and ultimately we put a personal firewall in front of it, but there's nothing to drive any of these vendors to really help with that or take some ownership. (25:58) I mean, I was on the phone last week alone with three major vendors, all three products are very weak in their cyber controls. (26:06) We make everybody fill out a cyber questionnaire.(26:10) If that doesn't pass, it takes my signature to buy anything that looks, smells like IT or medical equipment, and we won't do it. (26:17) But what happens a lot is the sales folks fill out, oh yes, we meet all these cyber requirements. (26:22) Then we go to install after we bought the product and the technical people come into play.(26:27) Oh no, we don't do those things. (26:28) What? (26:30) And so now guess what?(26:32) We have to get our attorneys involved and we start the battle and it's never ending. (26:36) But I wanted to bring up medical equipment because it is, it is ugly.
Matthew Chambers(26:40) I can add to that and I completely agree. (26:43) And this is not new. (26:45) In my experience, this has been forever since medical devices.
William "Bill" Phillips(26:49) 2017 was the first documented cyber attack through a piece of medical equipment. (26:55) You bet. (26:55) But they've been on networks, they're required.
Matthew Chambers(26:58) So again, that natural tension between enabling, in this case, clinical care versus doing it in a way that's secure. (27:06) And if you say, well, that's great. (27:07) You're not getting our network because you're using Windows XP.(27:12) Okay, so is IT helping or is IT a roadblock? (27:15) Is the CIO an enabler or is it a problem? (27:20) A defender.(27:21) A defender. (27:21) We must provide care, no doubt. (27:23) We must provide care in a way that does not harm the business.(27:27) I posit in several avenues that we all have one line job descriptions and probably everybody in here has a one line job description. (27:36) It's called protect the business. (27:38) So back to medical devices, are we protecting the business or are we inadvertently taking on some risk?
David Stuart(27:45) In our business, we call it the three Cs and everything's based off the three Cs. (27:51) The client, our colleagues, and the company. (27:55) And if you're not protecting all three, then that's not a path that we should take.(28:00) So that's how we use that to create balance. (28:05) So I'll move us along to another buzzword or acronym, which again has to do with what we just talked about, where our data is going to move us into AI. (28:18) And we're being pushed to deploy more AI.(28:22) We should have AI, but we need to step back and think about how does the AI get smarter, which is part of where the data's coming from. (28:31) So Eric, I'll start with you. (28:34) How have you embraced and deployed AI so far?
Matthew Chambers(28:40) You bet. (28:41) I could spend all day every day doing nothing other than just reading about all of the AI things that are going on in the world. (28:49) Show of hands, based on what you already know, is AI a friend or a foe?(28:55) Is it a friend or is it a foe? (29:00) A few foes, mostly friends. (29:04) In my opinion, that's the real deal.(29:07) AI is certainly here. (29:10) It's not optional. (29:11) It's not something you get to opt in or opt out of broadly.(29:15) Everybody is putting AI into their products. (29:18) You buy a refrigerator, it's AI-enabled. (29:21) You use your mapping tools, you use browsers, you have EMRs, they've all got AI-enabled.(29:28) You have the option of buying things like Microsoft Copilot. (29:32) You have Office and Teams and things like that, AI-enabled. (29:36) You have what are called large language model providers, like a Google, where you can build products.(29:42) So you can either inherit it from a provider, you can use APIs to interact, and you can go build your own. (29:50) I believe that we're entering a world of have and have-nots(29:55) You either have AI and you leverage it, or you don't, and you might fall behind or you will be out-competed by your competitors.(30:06) What that means for us is that we're trying to cautiously acknowledge all of the above. (30:13) We provide patient care. (30:15) We do not use AI to provide patient care, to be very specific.(30:19) Our providers do not use AI to generate clinical recommendations. (30:26) That's far different than, let's say, our physicians that are reading a research paper, or our legal team that is assessing contracts, or all employees who are just trying to be more productive and maybe write better emails to their boss if they're somewhat, you know, an IT person. (30:42) Maybe you don't write the most eloquent email.(30:44) You can have something popped out in an instant and probably makes your brand improved and save you a little time. (30:50) The data aspect is a big deal. (30:54) If you know how AI is constructed, it has essentially consumed all information worldwide.(31:01) Anything publicly available has been scanned, cataloged, pattern-matched, and so on. (31:09) So when you interact with it, you get a likely response. (31:15) It's certainly artificial.(31:17) It's not intelligent. (31:18) Let's not use silly marketing terms. (31:21) But it is highly competent.(31:23) It does what we think it should do. (31:26) It happens to interact with us as if it were a human being. (31:30) So we get somewhat lulled into thinking that it's intelligent, but really it's just a tool and a capability.(31:36) What we're doing, (31:37) just to wrap up my opening here, is we're working hard to figure out how do we enable it, (31:42) but also extend all the things that we would want for any capability from a legal and compliance (31:46) sense, protecting Krista's IP, for instance, making sure that what we do use is done thoughtfully, (31:53) safely, intentionally, and any information that is either allowed to go external is done as we (32:00) would want for any information going external.(32:03) AI doesn't release us from the world of compliance, regulation, risk, and so on. (32:14) Yeah, excuse me. (32:15) I agree with almost everything Eric said there.(32:19) Some days I wonder if it's friend or foe, and I think it can be both, honestly. (32:26) I look at it like any other technology that comes out and just kind of absorbs all the press, right? (32:31) Because think about it.(32:32) For one thing, a lot of technology is fueled by very, very intelligent, hardworking people who are trying to define what our existence looks like going forward. (32:45) A lot of technology is defined by vendors who are looking for something to sell, and then it gets picked up on by all the stuff on the internet and all the content providers who are looking for clickbait. (32:58) We go through these hype cycles in technology, and sometimes you'll get a question, what are you doing with X?(33:07) Like, what are you going to do with blockchain? (33:08) Me? (33:09) Nothing.(33:10) I'm going to wait until all the hype dies down, and I don't even invest in crypto, so I'm not going to touch it. (33:16) I've got nothing for it, and it's got nothing for me. (33:20) We don't see a practical application for it in healthcare.(33:22) I'm sure there may be somebody in here who can sell me a lovely blockchain app that'll fix all my problems, but we haven't met yet. (33:31) But when it comes to AI, this is one of those that you watch it, and when you sit there and read and study and then see some of the practical applications, you're like, holy crap, this is going to change technology more than anything in the past 30 years. (33:47) And I'm not an alarmist, I'm not a big hype guy, but I will say this to me feels like it's going to be a bigger change than the internet was when we first started computing, certainly bigger than client-server to date myself.(34:04) But like Eric said, you've got to use it practically, ethically. (34:12) What we're doing in our organization is we're actually, I was just reading it this morning, we're drafting policies and guidelines around the ethical and safe use of artificial intelligence, LLM, large language model. (34:28) How do we do it to ensure it's non-discriminatory?(34:31) How do we do it to ensure it's in the best efforts to care for our patients, customers effectively? (34:39) But I think there's so much untapped potential. (34:43) It's just amazing.(34:44) And if you haven't done it yourself, it's just kind of a funny story, I was telling this the other day. (34:55) So if any of y'all are comic book nerds like me, comic movie nerds, over the past couple of years, you've seen Marvel use quantum physics, they explain everything. (35:05) It's quantum physics.(35:06) And I was like, what is quantum physics? (35:08) And I tried reading, I was like, I don't get it. (35:10) So I go to chat GPT, and one of the interesting things you can do, I did it at home, my CSAT wasn't on our network.(35:18) One of the things you can do is, it's called prompt engineering, where you can give it a set of specific instructions on how you want it to answer you. (35:29) And it's important to note, Eric's right, the versions are changing, but the version I was using at the time, I think it was 3.5, it had all the information available up until 2021. (35:40) So there's a date cut off, if you ask it, what happened last Thursday, I didn't know.(35:45) So anyway, if you want to know about prompt engineering, ask chat GPT, tell me about prompt engineering, and it'll tell you, it's amazing. (35:54) But I'd say, explain to me what quantum physics is. (35:59) And it came back, and it was too many words, and too complicated.(36:02) So I say, in 150 words or less, at a 12th grade level, reading level, and you can specify this, it's kind of cool, explain quantum physics. (36:13) In 100 words or less, at a sixth grade reading level, explain quantum physics. (36:21) And finally, I was like, in bullets, at a second grade level.(36:30) And I'm just, I don't know anything about physics, is what I've discovered. (36:33) As chat GPT was like, give up. (36:40) But other than that, it's pretty cool.
William "Bill" Phillips(36:44) I kind of look at it a little bit different. (36:46) I kind of see it as the good, the bad, and the ugly. (36:49) Yeah, that's fair(36:51) Good movie reference too. (36:52) Right? (36:52) I just need the music.(36:54) It should be a movie. (36:55) If you look at it from a cyber standpoint of view, it's all three, the good, the bad, the ugly. (37:00) Right?(37:00) The good guys are using AI to help us with tools to block the bad guys. (37:05) But the bad guys are using AI to make their fish and everything else look seriously enough that you'll open it. (37:14) From a patient care aspect, we've embraced it.(37:18) We're putting together strategies. (37:20) But if I asked 15 of y'all in the room, what's AI? (37:24) I'm probably going to get 15 definitions.(37:27) There's weak AI, there's strong AI, there's super intelligent AI, there's self-awareness AI. (37:35) So AI is just like all kinds of things. (37:38) But I think we have to embrace it.(37:40) We don't have a choice in our reality. (37:43) Almost every system upgrade we get from a third party vendor, guess what? (37:47) Embedded AI.(37:48) Our radiologists are using it to read memograms. (37:52) It came with our upgrade. (37:54) It's there.(37:55) We're in Epic's pilot for AI. (37:58) We've started doing patient letters in AI with Epic. (38:03) So it's here.(38:04) It's coming. (38:05) You cannot stop it. (38:07) What we need to do is figure out policy strategies, what's next, how much are we going to adapt, adopt, and accept of it.(38:15) So I think this will be ever evolving. (38:18) I think the next two years will really kind of set the stage where AI is going to end up. (38:23) But I think it's also phenomenal in what it's doing in healthcare.(38:27) I talked to a company in Europe and they blew me away and they sent me brochures and how they were doing it. (38:34) They actually have an AI bra. (38:37) And what this does is this is looking at breast tissue for decay.(38:42) And it's detecting breast cancer so far in advance. (38:49) And you don't have to go get a mammo, but it's AI and that's good AI. (38:54) And so it's going to help in the long run.(38:57) You're going to see it's already happening a lot in imaging. (38:59) I already told you we're doing it in mammo. (39:01) You're going to see it more and more in imaging is very prevalent.(39:05) But we're also looking at a strategy for patient facing AI. (39:09) So how can it make the lives of our patients a little bit easier on whether it's making an appointment, canceling an appointment, getting directions to our organization. (39:19) So you really have to embrace it because it's here and you can't stop the freight train.
Matthew Chambers(39:23) Yeah, we're actually on the, I never say leading edge because I don't want to be bleeding edge, but I say we're a fast follower when it comes to adoption of some of these things. (39:33) So we do the first thing we did with the Epic in-basket responses. (39:37) So it automatically generates.(39:39) And if y'all don't know what that is, basically, if you message your doc, um, it can automatically generate a response. (39:45) It doesn't go out automatically. (39:46) The, the provider has to look at it and edit it and say, yeah, this looks good or whatever.(39:50) And what I've heard from them most firsthand is they're like, yeah, it's great. (39:53) It's it's more thorough typically because I mean, you're in a hurry you're trying to do, and AI has got, you know, all the data in the world available at its fingertips and it can, it's programmed to be, uh, compassionate. (40:10) Um, the other thing we're doing now is the, the one that's getting all the pub is the, um, it's a partnership with, uh, Nuance DAX, now owned by Microsoft and Epic where they're, it's doing the ambient listening.(40:22) And this thing is finally the real deal. (40:25) Um, we're, we've, we're live on about 70 docs, something like that, uh, across multiple specialties and what it, it listens to you and it takes the notes. (40:35) It does the H and P and everything.(40:36) And then the doc has to approve it. (40:38) And the feedback we're getting from the providers is they're like, this is amazing. (40:42) And some of the funny stuff, like some pretty compassionate, some of the funny stuff, like one of the docs was like, well, I don't know what to do with my hands anymore in a visit, you know, because they're like, I'm, it was always used to typing.(40:51) I don't have to type, but, um, it's improving the patient experience. (40:55) We had a provider that said I had a first time initial visit patient that was going through a really, really tough time. (41:04) And I got to hold her hands during the visit.(41:07) And I've never been able to do that, you know, since we went to EHRs. (41:11) And so it's, it's, there's some really, really amazing outcomes potentially from it. (41:18) And our CFO, we were talking about it.(41:20) We're trying to figure out funding for the next year, how to expand the program. (41:22) She's like, this is just going to be table stakes.
William "Bill" Phillips(41:25) Right. (41:25) And you know, we're doing the same with DAX and nuance, and we've got it pretty much in a lot of our ambulatory systems that were, we basically came to agreement with our physicians because it's not cheap as we all know. (41:37) If we give you this tool, can you see like two more patients a day?(41:40) Because we think we can save you that time. (41:42) So now we just made an ROI out of it.
Eric Reid(41:44) Right.
William "Bill" Phillips(41:45) Part ROI that's easy to sell. (41:47) And we're taking that ambient one more step. (41:49) I'm designing two new community hospitals right now, and I'm looking at no computers in the room, full ambient listening, full charting, all automated, all without hands-free.(42:00) You know, one of the things that we've heard throughout our careers is, you know, I can't talk to my patients. (42:04) I can't look at them because you IT guys put us in front of keyboards. (42:07) So now we're trying to turn that table and take those keyboards away.
Matthew Chambers(42:11) I always remind them the federal government did that.
William "Bill" Phillips(42:13) I do too.
Matthew Chambers(42:14) It definitely changes the dynamic. (42:17) How many of any of you have ever gone into a clinic, just to check up, whatever, and your provider, you know, warmly greets you, turns to their back to you, starts interacting. (42:26) Is that everybody?(42:27) Or is that just me? (42:28) Nope. (42:29) Well, here's what I do when I go, if I go from work and I've got my badge, I try to, I'm not from IT, but anyway, it completely changes that dynamic to now you're interacting directly with a patient.(42:42) I also, that real deal quote, I would heartily endorse that, that ambient note capability is the real deal. (42:47) It's the closest thing to magic. (42:51) You know, any technology sufficiently advanced is essentially magic.(42:54) It's the closest thing to magic I've seen. (42:56) The flip side of that would be, what was mentioned is, yep, that note is captured. (43:02) The physician must inspect and sign off of it.(43:06) What happens when they trust it so much that they sign off on it, but they don't actually inspect it? (43:11) Right. (43:12) Do you think that that could ever occur?(43:14) I think the answer is yes. (43:15) Absolutely. (43:16) We could establish a level of trust that's so implicit that it's just the AI taking it end to end, right?(43:24) That's the plus side. (43:26) The unsettling side for others might be, is like, wait a minute, now AI is actually making that decision, right? (43:33) Right now, or recently, there was a nursing staff strike at Sutter's and there was a similar protest at Kaiser around AI to say the nature of healthcare is being undermined a little bit, in their view.(43:50) I'm not weighing in personally. (43:51) I hadn't heard about that. (43:53) Absolutely.(43:54) Yeah, it was a week ago, 10 days ago. (43:57) Yeah. (43:57) And they're like, look, AI is not the same as a human being.(44:01) It is not the same as the care provided by a nurse or a physician and so on. (44:06) So, you know, we've got this, again, this tension between massive productivity lift, massive improvement in physician satisfaction and being able to do their jobs and maybe going home and have a healthy life versus handing off something that you trust that probably is mostly right, but sometimes could be a little bit wrong. (44:27) I think back to things in our past, our collective past.(44:33) How many people know what happened about 600 years ago? (44:36) You wouldn't know. (44:37) This is a strict question.(44:39) Printing press was invented. (44:41) What happened when the printing press was invented? (44:45) Massive productivity increase.(44:47) Other people lost their minds. (44:50) They went on strike. (44:51) They broke into printing press shops.(44:54) They smashed them. (44:55) They destroyed them. (44:56) It was an existential threat to their ability to feed their families and so on.(45:03) So these things can be good, but they may have, they may be like throwing a rock in a pond. (45:09) They may have rippling effects on society. (45:12) Something to think about.(45:14) I went to a Gartner conference at the end of last year, and one of the things that I cannot get out of my head even now is one of their featured speakers said in 100 years, Gartner believes that in 100 years, human beings will be penalized if they allow anyone other, anything other than AI to raise their children. (45:34) Other than themselves. (45:38) Grandparents, neighbors, providers, if they're human beings, society will frown on that to the degree that people will be either incentivized or de-incentivized regarding raising children.(45:52) So you might think, what? (45:54) Well, think about that. (45:55) If we implicitly assume AI is perfect, AI never has a bad day.(45:58) AI is never frustrated, grumpy, anything(46:01) It's perfect. (46:03) If we assume it's perfect, what is more important to a parent than raising your child in a perfect world?(46:09) So I'll just leave that as a question mark. (46:13) Societal implication is at the door with AI. (46:16) I think you just wrote a Black Mirror episode.(46:20) Or has that already been out? (46:22) I think that's coming up in the upcoming season. (46:25) Yeah, the Black Mirror.
David Stuart(46:27) With the pace of the tech advancing, where are you at with policy, data governance, keeping up with regulatory around the data that is being utilized, the use of that data, et cetera?
William "Bill" Phillips(46:46) Yeah, so a difficult process. (46:49) We are in the process of actually writing policies around this. (46:53) But a document we've put together, and I'm sharing actually this Thursday with our leadership, is our AI strategy.(47:00) Because again, AI is so many different things. (47:03) So we've taken and put together a document. (47:06) Here's what we have, because you ask your organization, are you guys using AI?(47:11) And probably most people are going to say, just a little bit, but they really don't know. (47:14) So we've put together a document. (47:16) It was amazing because I blew my own staff's mind away when I started asking these questions.(47:21) Start documenting all the AI tools we have. (47:25) And it's a massive list because a lot of it's already embedded in your applications, and a lot of people are unaware. (47:33) What is the deficit of what we think we need?(47:35) And again, I've got a big emphasis on patient facing. (47:38) And then where do we think we're going long range with it? (47:41) So it's kind of three phases.(47:43) It's going to be educate, and then what's next, and then what's long range. (47:48) And what's coming on top of all that is a policy. (47:51) Because like the examples with CHATGBT, we played with it just a little bit, and we asked for the history on one of our facilities.(47:59) And the history it gave on our historic building downtown was not our building. (48:04) We don't know what it was. (48:06) It was a cotton field.(48:07) And so we said, okay, we're not going to trust it. (48:11) Our CEO is an attorney, and he found an interesting article where an attorney, and I think it might have been in Houston, but a recent case, he had his whole case built through CHATGBT(48:23) And he got up to the judge and presented it, and guess what?(48:26) It wasn't for his case. (48:28) So you've got to be careful with where you have it. (48:32) So we're going to put in policy on what you can and what you can't do.(48:38) And if you're going to, we're really not going to allow it. (48:40) We deliver all our apps through a secured mobile app. (48:44) So if we don't allow it, they can't get it and use it on our network infrastructure(48:48) So I think it's going to be a lot about policy, but it'll be almost like a cybersecurity policy. (48:53) It's going to change and change and change. (48:54) It's going to be a very fluid policy until we get some of the smoke clears around it.
Matthew Chambers(48:59) Yeah, I'd agree. (49:01) I think we're in the same place. (49:04) I've got two teenagers, excuse me, one in college, one in high school.(49:08) Like a good AI, I always read them books when I was little, going to bed. (49:13) And there was one of my favorite books was, I don't remember the name of it. (49:18) I've been looking for it because I think it's a great book around information security.(49:23) It was, maybe it was always the Scaredy Cats. (49:28) And it's about this family of cats that they're too scared to get out of bed. (49:35) They're too scared to go to the door when somebody knocks on the door.(49:38) They're too scared to cook breakfast. (49:41) Well, I can burn myself. (49:43) And it's funny, like everything, they're just like too scared.(49:45) And then finally, the baby cat says something along the lines of, well, if bad things can happen, doesn't that mean good things can happen too? (49:53) And so they have this cathartic moment where they go out and swing on the swing and it's all wonderful. (49:58) But I kind of feel like one of the Scaredy Cats every day because it's, well, we want to do this great thing and give them all this data and great things can happen.(50:08) And I'm like, well, if good things can happen, can't bad things happen too? (50:14) They're like, what are you talking about? (50:16) So it's, you do, you have to be, you have to balance the good and the bad and the ugly to that point.(50:22) But the data governance policies and just freely turning over your information, I mean, those things are gone. (50:29) And then, you know, it's, I can't tell you how many times a day somebody's like, well, we'd like to take all your data and do X. (50:37) And I'm like, you lost me at, we want to take all your data, you know, so no thanks.(50:42) And we're doing the same thing. (50:44) We're working hard, legal and compliance are trying to get their minds around what's an effective way to enable our business and yet protect our business. (50:53) So that's the normal thing of any new technology or any new capability.(50:58) What policies do we have? (50:59) Well, what policies should we even think about? (51:02) What are other people doing?(51:04) What are our third parties advising us we should do? (51:09) To me, it's an extension of what we already do. (51:11) We already protect the company in many ways around many aspects.(51:14) It's that notion of our information is outside our four walls. (51:18) You have to assume that anything you put on the internet is consumed. (51:21) If you're okay with that, you're good to go.(51:24) You may want to jot it down that you're good with that and so on in case you get a legal challenge. (51:28) If you're not good with that, well, then you have to describe what are you okay with? (51:32) What are you not okay with and why?(51:33) Under what conditions can information be used professionally and so on? (51:40) That's not a simple thing. (51:42) A sample list of what we're thinking about is many, many, many pages long with all sorts of implications.(51:49) It's like an analogy would be, have you ever thought about trying to change a tire on a highway as you're whizzing down the road? (51:56) I-35, especially here in Texas, not the safest thing to do. (52:00) Yet, you have to(52:02) It's here. (52:03) It's fast moving. (52:04) If you don't do it, well, then you'll be at the mercy of the market or perhaps someone making a mistake, putting something out harmlessly that then is consumed.(52:13) If it's your company moniker on it, well, now you've got some skin in the industry and you've got some potential liability and so on. (52:21) Back to the fraidy cats, that doesn't mean that you should just say, well, we won't do anything because that's not a reality either. (52:29) You'll get left behind.(52:30) You need to navigate that. (52:32) You need to figure it out and you need to pick and choose and figure out, well, what is an appropriate and safe use of this new capability?
David Stuart(52:41) I think that's what we're maneuvering right now. (52:44) We're in the process of writing our policies as well. (52:47) You mentioned IP earlier.(52:50) IP is becoming a big part of the conversation.
Matthew Chambers(52:52) Intellectual property, yeah. (52:52) How do you protect and guarantee appropriate use of what you would consider intellectual property? (53:00) Not just the things that you're doing that might be publicly available, but the property that you create as a company.
David Stuart(53:06) What do you do with that and how do you protect it? (53:08) How is it being aggregated? (53:10) Who else has access to the use of the models that learned from that data?
(53:18) Let's talk a little bit more. (53:20) We've touched on a little bit, but what are the main areas where your strategy and focus is for where you plan to deploy AI, let's say, in the next 12 to 18 months? (53:33) I'll start with you, eric.
Eric(53:34) Sure. (53:35) We signed up for Microsoft Copilot. (53:37) That's an easy one.(53:38) We were an early adopter. (53:42) I've got a couple hundred of what are called personas, people in different roles. (53:47) I've got HR, legal, clinical, IT, and so on.(53:52) We want to get a good sense of if we enable Copilot for Outlook, Teams, Microsoft products, GitHub, Power Apps, really anything Microsoft. (54:05) For better or worse, we do use a lot of Microsoft products. (54:08) What does that mean?(54:10) Is that worth that $300 a license times X number of employees? (54:17) Is there a lift? (54:18) Is the benefit greater than the cost or not?(54:21) It's not going to be everybody gets it because the cost is non-trivial. (54:28) We also don't want to just chase technology for technology's sake. (54:31) We want to put things in the hands of people that will get great benefits(54:35) So far, it's our easiest win. (54:37) It's a simple way to tiptoe into AI. (54:41) Everybody can benefit from writing an email and MIT you up with an email.(54:45) Join a Teams call, 30-minute Teams call, join 10 minutes late, no problem. (54:50) If you'd like a summary, you bet, right there. (54:52) You're instantly caught up and I can tell you it's highly accurate.(54:55) It's very beneficial. (54:56) It's scary. (54:57) You miss that meeting entirely, it says, no problem, let me send you this little thing and 150 words would you say to a fourth grader, I'm just joking, and so on.(55:07) You get a summary that it's like you attended the meeting. (55:10) It's that good. (55:11) So those are like simple non-risky productivity lifts that if you're not doing that in some form regardless of your provider, I think that's really safe to do.(55:21) That does not involve putting your IP at risk. (55:24) It does not involve having to decide what to do with data. (55:28) It's just enabling something like using a calculator versus long division in my sense.(55:34) We are using the ambient notes. (55:36) There's a couple of leading providers. (55:37) We are working with Epic.(55:39) We're looking into how do we take in like the human resources area. (55:44) I'm sure we all have these software products that have a gazillion menus and clicks and drop downs and this and that and the policies are all there. (55:53) Maybe all of your policies are in there.(55:56) Maybe not. (55:57) If you say, I need to check my PTO hours, well, sure, you go here, click and do this and do this and do this. (56:03) Great if you know how to do that or if you remember it.(56:06) So we're going to be using AI to make that much simpler. (56:11) Just like interacting with Google now. (56:13) How do I change my PTO?(56:15) Well, here you are. (56:16) Would you like me to change PTO for you? (56:18) It's going to take, again, the drudgery out of interacting with corporate systems and it's going to enable people to get things done more quickly.(56:26) Those are the easy answers. (56:28) The more complex one is what do we do about AI embedded in clinical systems? (56:34) So now we are potentially using AI for interacting with human beings.(56:39) That's the one that is a longer discussion but maybe eventually has the most pervasive benefit. (56:48) Yeah, we're looking at some of the similar things. (56:51) I was thinking about the Microsoft Copilot(56:54) I mean, it's pretty pricey. (56:55) It is pricey, yeah. (56:56) So our pilot group is much, much smaller.(56:59) It's basically us and IT trying to figure out what the risks are associated with it because the Teams meeting thing is a very cool example but at the same time you we instituted a policy across the enterprise that you're not allowed to record meetings. (57:14) They're discoverable. (57:15) There may be IP.(57:16) There may be whatever reason that you don't want those things to be recorded unless you go through a specific process. (57:22) Well, they have to record it to do these things. (57:26) So we're like, okay, did we just invalidate our own process if we turn this stuff on?(57:29) That's true. (57:30) So yeah, that's the scary cats coming back out, right? (57:34) So it's a good thing but something bad might happen too.(57:38) But we've actually looked at developing, or I think we are developing something similar, the HR policy thing. (57:46) I totally agree. (57:46) That thing's awesome.(57:48) Just the ability to... (57:49) I think the stuff that's the really low-hanging fruit for large language models is summarization. (57:57) So if you look at things like...(58:00) I see the MacBook here. (58:01) You think about a MacBook versus an iPad. (58:03) When iPads first came out, everybody was trying to use them to replace their MacBook or their PC.(58:09) That's not what they're made for, in my personal opinion. (58:11) It's a consumption device. (58:14) That's a creation device and this is a consumption device.(58:17) And if you look at AI doing those things as well, it can certainly create data. (58:21) But where it's really good right now, in my opinion, is that consumption and summarization. (58:27) One of my favorite things in the world.(58:29) So like I said, our pilot's really small. (58:31) It's like seven or eight of us. (58:33) But if you're not on your email and there's five or six people go back and forth in a thread and you're out for a day, you just go at the top and it just summarizes email.(58:45) And it is incredibly accurate when it goes through and it just tells you here's all the discussion that happened. (58:51) And it's amazing some of the stuff that it does. (58:54) What we're looking at next though, honestly, is our CFO actually asked us to get together and say, where can we use this kind of next generation of tools?(59:04) I'd lump automation in with AI as well. (59:08) Where can we use some of these tools to reduce cost and see cost savings across the organization? (59:15) So that's the next step for us.
William "Bill" Phillips(59:19) So I've already talked about patient-facing AI. (59:22) I talked to what we're doing with clinicians. (59:23) What we can't forget is our internal workforce.(59:27) And so we've already started playing with some AI and I'm actually glad I'm at this financial conference because I saw all the vendors and so I'm going to be getting cards from everybody because we're looking. (59:39) We played with bots and AR in our financial system to try to help our workforce take the burden of managing work queues off of staff and let them do some other things. (59:51) Epic has many, many work queues, pre-offs, and the lists go on and on(59:56) And we just did a quick kind of proof of concept and we found out that the AI was processing 10 times at the rate of a human. (1:00:05) We're not doing it to put people out of jobs. (1:00:07) We're doing it to allow those folks to do other things to better our organization.(1:00:14) So I think a big focus for us now is kind of that three-sector. (1:00:19) What do we do for patients? (1:00:20) What do we do for providers?(1:00:22) And then what do we do for our internal staff?
Matthew Chambers(1:00:24) I think we're looking at the same things and I think about it as allowing everyone to practice at the top of their license. (1:00:30) Exactly. (1:00:31) Right, that's what we talk about for docs to be more productive.
David Stuart(1:00:34) So the same thing for people. (1:00:36) Put the same button every day.
Matthew Chambers(1:00:38) A bot can do that.
David Stuart(1:00:40) But let them think and use their skills.
Matthew Chambers(1:00:44) Yeah, the rote stuff, let the AI and the bots do it. (1:00:49) The exception handling, let the people do it. (1:00:51) Exactly.
David Stuart(1:00:52) So we've touched on this throughout here, but to put a little bit more focus on it, how are you utilizing these AI tools? (1:01:02) Are they external? (1:01:04) Are you getting them from third parties(1:01:05) Are you building your own models internally? (1:01:09) What is it that you're doing with AI?
William "Bill" Phillips(1:01:13) I think it's all the above. (1:01:15) I mean, I've already mentioned many of our products are coming with AI and half the time we don't even know it's being embedded. (1:01:21) Right.(1:01:21) And then we're out looking for specific AI components. (1:01:25) And so we shop the market for specific. (1:01:27) And then we're also meeting with many integrators along the aspect of, okay, if we can dream it, they can build it.(1:01:34) So we're trying to bring all these together. (1:01:37) And we've actually started asking our organization, hey, think about AI and let's start creating a list of what we could do in AI that would help you do your job. (1:01:48) And we're gathering that list.(1:01:50) And then we're going to an AI company and say, okay, now this is not off the shelf. (1:01:56) This is invent the process and help us along with that. (1:01:59) I mean, we're even looking at Amazon AI to take into our food services, into our cafeterias, where AI is monitoring and you don't have to have cashiers and these type of things.(1:02:12) It's more of a grab and go type. (1:02:14) So we're looking at it from all components.
Matthew Chambers(1:02:18) Yeah, I'd say pretty similar for us. (1:02:22) If anybody is trying to build their own AI engine, like, I mean, that's like 10 times as crazy as trying to build your own EHR. (1:02:35) I mean, you need to think about what you're good at and what you're going to get your lunch eaten if you try to do.(1:02:44) But yeah, like Bill said, I think it's coming up with your personal use cases and trying to figure out how to apply them. (1:02:50) That's where the power of it is. (1:02:52) And same here.(1:02:53) And I would absolutely echo for those that have been in the industry long enough. (1:02:57) Do you remember when people used to write their own software? (1:03:01) And for good reason, because the big providers didn't exist yet.(1:03:04) Pre-Epic, pre-Cerner, pre-everybody. (1:03:08) So hospital systems wrote their own. (1:03:10) They had teams of programmers.(1:03:11) We all moved away for that for obvious reasons. (1:03:13) So if we think about that and say, well, let's talk about AI and LLMs. Let's go write our own LLM. (1:03:20) Okay, great.(1:03:22) You must have 50 billion or so and hundreds of people in a workforce of thousands worldwide that you're not sure what to do with. (1:03:29) So you're going to create a new AI because that's what it will take. (1:03:34) So that's not the answer, clearly.(1:03:37) It is to identify a partner, whether it's Google, Microsoft, others that have that investment. (1:03:46) And now you figure out what do you do with it. (1:03:48) You either use it via publicly consumable models like Copilot.(1:03:51) That's easy. (1:03:52) You use it or use a competitor and so on. (1:03:54) You might interact with it via an API, so a third-party integration.(1:03:58) Okay, well, that's normal. (1:03:59) You do API third-party integrations all the time. (1:04:02) The stuff you say, well, we want to create our own.(1:04:05) Well, okay, you are going to create something, in our case, Krista-specific, but we're not starting with a blank sheet of paper. (1:04:11) We're engaging with a third party. (1:04:13) We're going to have an LL model that we then layer our IP on top of that.(1:04:18) In that use case, I think of AI as glue, the glue that allows other things to fit together. (1:04:25) If it's something that, well, more properly belongs in a provider's roadmap to say, well, let's write something for Epic, well, let's make Epic aware, make sure it's on the roadmap, but maybe we write some glue in the meantime, something like that. (1:04:38) So we're doing that.(1:04:40) We've figured out some use cases. (1:04:41) It would be a lightweight use of a commercially available LLM that we then Krista-size, so to speak. (1:04:48) Our information stays internal, and it becomes something that has benefit, low risk.(1:04:55) Our data doesn't leave Krista's, and that could be replaced by something that a third party provides down the road, but for now, we have a business need, and we think that's a smart use of AI with moderate investment and limited risk.
David Stuart(1:05:15) So we've touched on this in multiple ways too, but I want to come back to the patient engagement with the use of AI. (1:05:24) For us, it may be patients that are calling into a call center, maybe interacting with a portal, and we're looking at use cases of how could AI help those that wants to be helped that way, but still get to the human touch as needed, but how are you looking at it?
William "Bill" Phillips(1:05:44) So the biggest thing, and I use this example on what I'm trying to get to, and I'll use this as me as a patient. (1:05:52) If I call in and say, hey, I've got a 2.30 appointment. (1:05:55) I'm running late.(1:05:56) I can't make it. (1:05:57) AI picked that up and says, Mr. Phillips, we understand. (1:06:01) Can you make 3 o'clock?(1:06:03) Yes, I can make 3 o'clock. (1:06:05) Great. (1:06:06) Well, we've already changed that.(1:06:07) We've just sent you a MyChart message. (1:06:10) By the way, while I have you on the phone, I noticed there's a $15 copay. (1:06:15) We have a credit card on file.(1:06:16) Do you want to pay that now or just deal with it when you come in? (1:06:19) Well, pay it now. (1:06:20) I'm done.(1:06:21) So now I've been talking to AI. (1:06:22) I'm driving. (1:06:23) Hey, just let you know I just pulled in, but the parking lot's full.(1:06:27) Okay, Mr. Phillips, if you pull out of that parking lot and turn left, we have a parking lot next door that has parking spaces in it. (1:06:36) So that's what I mean, patient-facing AI. (1:06:39) That's a brief example, but how can we utilize it to really help the patient navigate through the system?(1:06:47) We all have very complex health care systems, and we live in them, and so taking some of that complexity out and making it easier for the patient is what I'm really about.
Matthew Chambers(1:06:59) Yeah, I think that's a great example. (1:07:04) Yeah, and I don't even know that I could add on that. (1:07:06) Actually, I'm taking notes.(1:07:07) That's a great idea. (1:07:09) What I'd say is I would much rather use a bot because it's typically faster, and we're doing it to our patients and customers just like the companies that we hate do it to us, right? (1:07:23) It's going to be cheaper and faster to put a machine out there answering the phone, but the only reason I start banging through and saying representative on American Airlines is because the automation is so bad.(1:07:37) You know, if it's good, I'd much rather use that because it's going to be faster, but when you're, you know, it's like, how can I help you today, representative? (1:07:46) It's because I know it's not going to be able to help, and if any of you work for American Airlines, I'm sorry, but I mean, it's just crap. (1:07:55) I'm sure I speak for many.(1:07:57) If you all have a help desk, does anybody call it the helpless desk? (1:08:00) Has that you ever heard that term or something similar? (1:08:03) I mean, I couldn't personally do it.(1:08:06) I know it would kind of wear me out because you just try so hard, and yet it's impossible to meet, you know, the wishes of everybody. (1:08:13) I think that's where AI is going to step in. (1:08:15) I think AI is going to be, and I'm seeing some evidence of it already, AI is going to be able to provide a person-specific experience that might interact with me as a CTO, you know, rather crisp and so on.(1:08:30) It may interact with me as an elderly person, kind of fearful about whether I'm going to survive my medical event or not. (1:08:37) It's going to customize that message to me. (1:08:40) It's going to say, hey, you know, this will be okay.(1:08:43) We'll work. (1:08:43) This will be worked out. (1:08:44) It'll help provide not only a quality experience, but it will lower the stress level.(1:08:51) It'll provide a level of reassurance. (1:08:54) It'll also speak to me if I happen (1:08:55) to be from New Mexico, perhaps, versus Texas in a way that if I'm out in the more rural area of New (1:09:02) Mexico and I don't necessarily speak English as my first language, it will be able to address that (1:09:07) and, again, connect with me at a level as if it were human that will solve the problem, no doubt, (1:09:15) but in a way that a human would appreciate, lower stress, and allows them to get through it more (1:09:22) easily.
William "Bill" Phillips(1:09:23) And I'll add one component. (1:09:25) I gave you an ambulatory example. (1:09:26) I'm going to give you an inpatient example.(1:09:28) I told you about ambient listening, and that was on the provider side, but we're also looking at and working with a couple companies on ambient listening for the patient side. (1:09:36) In other words, laying in the bed, forget that daggum nurse call button that everybody drops and can't find it, right? (1:09:42) It's, I'm cold, turn the thermostat up.(1:09:46) Put it on channel 13. (1:09:48) I'm ready to order food. (1:09:49) I want a cheeseburger today.(1:09:51) No, sorry, sir, you're on a cardio diet. (1:09:53) You can't have it. (1:09:53) Well, what can I have?(1:09:55) That would be me, right? (1:09:56) What can I have? (1:09:58) And making it, again, hey, I don't speak English.(1:10:02) Can we talk in Spanish? (1:10:03) And all these type of things, again, to really aid that patient through difficult healthcare. (1:10:12) Those are great examples too.
Matthew Chambers(1:10:14) A quick question. (1:10:16) You've all interacted with AI from all sorts of products. (1:10:19) They all have similar AI look and feel.(1:10:24) Have you ever noticed that none of them sound like Mr. T? (1:10:28) Did you say Mr. T? (1:10:30) Mr. T. (1:10:31) You know, like, hey, real gruff, right?
Eric Reid(1:10:32) Yeah.
Matthew Chambers(1:10:34) What does he say? (1:10:35) I pity the fool? (1:10:36) I pity the fool that interacted with me as an AI.(1:10:38) Doesn't do that, right? (1:10:39) That's highly intentional. (1:10:41) And so back to my example, it will probably become very personalized, something that you could easily interact with and trust.(1:10:48) Well, there was, I don't know if you all remember this, but there was that study where it said AI generated notes were like far superior to human generated, or it was AI generated questions, answers to medical questions were deemed by judges to be far superior to physician generated answers. (1:11:10) And you look at the study and it was such a ridiculous study because it was, they might use Reddit, any Reddit users? (1:11:20) Okay.(1:11:20) So it was people asking medical questions on Reddit and docs going in and answering in the middle of the night. (1:11:26) First of all, they're off the clock. (1:11:28) Secondly, it's some moron on Reddit that they don't know, and they're just giving them their opinion.(1:11:33) And then you compare that to an AI tool that is designed, engineered to sound like a compassionate, caring human person, and it comes back. (1:11:42) And so it's like bringing a knife to a gun fight. (1:11:44) I mean, that's not a good study.(1:11:47) Now, what I have seen though, is I was at a function of the day sitting with some of our docs that are using it, and they're like, it's better than I am. (1:11:56) They're like, it takes the time, it writes a good answer, and it does exactly what you said, the way that it's programmed to act. (1:12:02) And I would offer a thought bubble.(1:12:04) In the near future, and you can define there, will AI appear to be more human than humans? (1:12:12) Episode two of Black Mirror next season. (1:12:15) If you all see that someday, you'll know the idea came from me.(1:12:18) yes.
William "Bill" Phillips(1:12:20) Well, just Google the growth rate dollars that are expected in AI over the next 10 years. (1:12:26) It's billions upon billions, like 400 billion over what it was a year ago. (1:12:31) Amazing dollars coming.
David Stuart(1:12:35) Now, Matt, I want to go back to something you said about, I'd rather use a bot for that.
Eric Reid(1:12:40) Yeah.
David Stuart(1:12:40) This has been a difficult area, especially for me, and I spoke about it last year. (1:12:46) The difference between RPA, machine learning, and AI. (1:12:52) And it seems like we're grouping everything in the AI today, but really trying to understand which is the best tool, and then ultimately, AI may drive RPA.(1:13:04) But how are you seeing the use of those three things within your organization today?
Matthew Chambers(1:13:10) I'm old enough to remember, you all remember Windrunner? (1:13:13) I do. (1:13:14) Yeah.(1:13:15) I thought I heard somebody giggle, maybe somebody's used it before. (1:13:19) But robotic process automation is really, in its most basic form, it's just automating the key strokes that a human would do in a Windows environment. (1:13:30) That's the most basic, and that's what first came out.(1:13:33) It was called Windrunner 100 years ago. (1:13:36) And robotic process automation is, in my estimation, it's simply that. (1:13:41) It's you're taking a process that may be manual and automated together, and you're trying to automate it and simplify the amount of data entry that a human has to do.(1:13:54) There's still a lot of low-hanging fruit for that kind of stuff, if you can find it. (1:13:59) Especially in really, really paper-intensive situations like revenue cycle, right? (1:14:04) You get a denial, you have to take this form, that form.(1:14:08) So I think there's some low-hanging fruit there. (1:14:10) I think AI, while it may achieve some of the same things, it's kind of the, I don't know, it's the difference, I'll mess up a metaphor, but it's like the difference between an axe and a chainsaw. (1:14:20) I mean, AI is answering questions based on what it's learned.(1:14:28) It's not necessarily just going and clicking the buttons and doing the things that our PA might do. (1:14:36) Any other feedback? (1:14:38) Yeah, I would add to that.(1:14:40) As Matt said, it's much more than just, let's say, screen scraping of the past. (1:14:46) Capture keystrokes, replay them, detect events. (1:14:49) If this event occurs, do this action.(1:14:52) If this occurs, more than like that cookbook, definitely not intelligent approach. (1:15:00) So take that, which it still has a use case. (1:15:02) RPA is a big deal.(1:15:04) It saves a lot of money, and it increases productivity, and things like machine learning. (1:15:10) AI doesn't replace machine learning. (1:15:12) The radiology image reading, that's machine learning right there.(1:15:14) It'll become more and more accurate over time. (1:15:17) In fact, I saw a study that said machine learning based radiology reads surpassed human capability in 2014. (1:15:26) So a long time ago.(1:15:29) So AI is informed by literally every piece of available knowledge worldwide, publicly available. (1:15:35) So that's not something you need to use to script a business function, and so on. (1:15:41) But it could be something that is much more meaningful.(1:15:44) So AI is, in my opinion, a superset of all those sub-technologies, whether it's machine learning, whether it's robotic process automation, and so on and so forth. (1:15:54) It's a good way to look at it.
William "Bill" Phillips(1:15:55) Yeah, I mean, we sum it up. (1:15:56) We roll it all in one basket. (1:15:58) Healthcare IT is full of acronyms.(1:16:00) We love, IT guys love acronyms, just to mess all you guys up, and hey, they know what they're talking about. (1:16:06) No, we don't. (1:16:07) So we sum it all up in AI, just to make it easier for everybody else to understand.
David Stuart(1:16:16) So I'd like to open it up to the group here for any questions. (1:16:21) You've got a group of CIOs up here. (1:16:24) It's not often you get them cornered in one place.(1:16:27) Were there any questions about where we're going with it?
Eric Reid(1:16:38) No. (1:17:08) So how do you balance that in terms of taking away, just in general, you have a lot of used personas that you use. (1:17:18) How do you take that into consideration and avoid people getting comfortable with that and trusting AI to the point that they put themselves at risk?(1:17:27) Their license, you know, when you look at the provider side, they're ultimately accountable for those clinical records.
William "Bill" Phillips(1:17:33) Well, you know, I think it goes to, they were always ultimately accountable for those records. (1:17:39) And you saw a phase come and go, which was scribes, right? (1:17:43) And so you had someone else typing it.(1:17:45) Some docs had scribes in the room with them. (1:17:48) Some had online transcription. (1:17:50) But again, they knew they had to review it before they signed their name on it.(1:17:54) They've already come through that pace because of those two jumps. (1:17:58) And they're adopting and accepting the Nuwok stacks type of products just overwhelmingly. (1:18:07) It's just, it's the next transition.
Matthew Chambers(1:18:10) When we, when I first saw the Nuance as the vendor, DAX is the product, for those of you who don't know, I first saw it actually at an Epic conference. (1:18:21) It was a healthcare system presenting what they'd done with it. (1:18:25) And it was the version before what we're on now, I think now they call it Copilot, is it was just recording all of it and then sending it offshore to, you know, and there were humans involved.(1:18:40) And I said, this is no better than the transcription that we've had for years and years. (1:18:44) I don't know why you invest in this. (1:18:46) But then the first time I got the demo with the real AI, I was like, okay, this is legit, this is real.(1:18:53) And I think the way that you have to look at these things is first of all, like, so diagnostics, right? (1:18:59) You don't necessarily, there's just all these ethical, moral, esoteric questions you have to ask. (1:19:07) Do you want machines doing diagnosis?(1:19:09) Right now, I think the answer for people who are not selling AI is no, not yet. (1:19:14) We're not ready to go there yet. (1:19:18) And so then you've got to set the policies and be very, very thoughtful in advance of don't ever even put the AI in the potential of doing the diagnosis if that's not where we want it to be.(1:19:30) But it is a slippery slope, people are going to fall down it fast. (1:19:33) And I'll echo the same thing. (1:19:35) I think it's a slippery slope that becomes something you don't even think about.(1:19:38) If anybody used Google Maps or Waze or anything else, which might be a minor form of AI or some automated thing, how many of you say, oh, I don't know, let me double check that? (1:19:48) You don't do that, you trust it implicitly. (1:19:50) How many of you have been sent down a sketchy road and maybe you were too happy about that?(1:19:55) Or maybe for me, one of my biggest beef, I wish there was a keep me on what I would call the main drag, keep me on the primary road, do not send me through a residential area to save four feet.
Eric Reid(1:20:06) There's children in residential.
Matthew Chambers(1:20:08) Stay away from Waze. (1:20:09) It's insane. (1:20:12) It goes back to trust.(1:20:16) So radiology diagnostic, that is proven to be highly accurate, surpassing the ability of any given individual human being. (1:20:24) Absolutely. (1:20:25) So trust has been established.(1:20:28) Physician notes we've spoken to, that seems to be the real deal. (1:20:32) The real deal is that the accuracy is spot on. (1:20:37) It's not sometimes okay and sometimes not.(1:20:39) No, it's the real deal because it's always right. (1:20:43) How many in the room here have children that are going through your STAR exams or did maybe a week or two ago? (1:20:51) Are you aware that the Texas Department of Education decided to allow AI to make final decisions on the written portion of all STAR exams?(1:21:01) Final decisions. (1:21:02) In the state of Texas? (1:21:04) State of Texas, Department of Education.(1:21:08) Well, let me show you the flip side. (1:21:10) It saved between $15 and $20 million is what it says. (1:21:14) So my analytical mind says, well, wait a minute, what is it, 15 or 20(1:21:18) And do I get that money back or not? (1:21:21) It put 6,000 people or companies or services out of getting that contract. (1:21:28) So now there's a societal impact of that.(1:21:31) But the more important thing for me is my kiddo, ninth grade, so not quite ready for college yet, but these tests matter. (1:21:39) I gave a speech a while ago where I talked about this STAR thing and someone raised their hand and says, my niece was denied admission to the college she wanted to get into because of that test. (1:21:49) She missed some random score by one point.(1:21:53) It was scored by AI. (1:21:55) Did AI get it right or wrong? (1:21:57) We don't know.(1:21:58) There was no double blind run. (1:21:59) There was no, hey, we're going to score it with humans, we're going to score it with AI, see how close they are. (1:22:06) And eventually if AI is as good, okay, well, then now we can trust it.(1:22:10) That trust aspect was skipped. (1:22:12) The kicker was, well, hey, if you don't trust AI, it's no problem. (1:22:16) You can go to this website, you can put in your name, you can pay $50 and ask for an appeal.(1:22:22) So there's an immediate impediment to challenging AI. (1:22:25) So imagine that a person that doesn't know any of this, didn't see that article, doesn't know that, hey, you missed that score, but maybe you did, maybe you didn't. (1:22:33) How does AI know how a person who maybe doesn't use English as their primary language happens to write it in English?(1:22:39) Maybe there's nuance and context that a human being would pick up on that AI cannot. (1:22:46) So for me, it all rolls back to demonstrated trust. (1:22:50) We're all highly qualified professionals.(1:22:52) We know what is something that's trustworthy. (1:22:55) I think you know it when you see it. (1:22:57) In some cases, AI is already there.(1:22:58) In some cases, I think it's to be determined. (1:23:01) How are they going to burn the printing presses when they're all in the cloud? (1:23:05) There you go.(1:23:06) Can't smash a printing press in the cloud. (1:23:09) Question? (1:23:09) In health care, I think it's a key when you look at it, but think of the ethical, moral, business reasons that you can't, at this point, an internal med doctor records his interactions or her interactions with patients over a period of time sufficient to build a body of knowledge of his practice.
Eric Reid(1:23:51) Now, the role that AI model and is able to extend chemical herself to a wider population.
Matthew Chambers(1:24:02) How do you see that unfolding? (1:24:06) Black mirror episode three. (1:24:09) No, in all seriousness, that absolutely can happen.(1:24:17) It makes all the sense in the world if you have a shortage of physicians. (1:24:21) I mean, in all seriousness, if you want to take computing capability, which is probably, I don't know, two, three, four, five years away, and a bot, like a physical robot, which now they've got these jumping dogs that can shoot flames and what is it, Boston Dynamics? (1:24:38) Yeah.(1:24:40) They're not far away from, in my opinion, a sufficiently humanoid Android machine that you could embed that AI in it(1:24:50) And the first thing that I would think is, that's a combat physician. (1:24:54) You can drop that into a dangerous scenario.
(1:24:59) Granted, they're probably going to cost 10 million bucks a piece, but the software is not, I mean, the software is in the cloud. (1:25:04) You got to just drop another body in there and drop the software back on it(1:25:08) I think that makes all the sense in the world.(1:25:12) And it also leads to all these robot revolution sci-fis, once we reach the uncanny valley or the, crap, what's the word for when AI is smarter than we are? (1:25:24) Yeah, thank you. (1:25:29) It's fascinating when you do that.(1:25:34) I don't know. (1:25:37) But then I go back to, at some point, much like, you know, you've got to go a little different. (1:26:00) Because people, and I've said it to my own organization, there's enough video of my CEO, he created an AI image over Teams, LinkedIn, whatever, that says, yes, it's okay to go request and approve, and this value is just going to be...(1:26:22) Yeah, that's interesting. (1:26:24) I think this is where we may finally see a use for blockchain to validate whether or not this person talking to me is real. (1:26:33) Deep fakes are the real deal, and what are called shallow fakes.(1:26:37) So it takes just a tiny bit of technology to do exactly what you're talking about. (1:26:41) It costs almost nothing to do what you're talking about. (1:26:43) It's in the cloud.
William "Bill" Phillips(1:26:44) And going back to your original question, we're doing a lot of this now with the learning. (1:26:49) So think about it, probably all of our ORs, we're doing it. (1:26:52) We all have intelligent ORs where every surgery is being filmed, and it's used to quality control, it's used for training, that's learning, that's being fed into systems.(1:27:05) You have virtual reality now in healthcare, where surgeons and many care providers are being trained through virtual reality surgeries. (1:27:14) We use... (1:27:15) I've got it on my desk, I'm looking at virtual reality to help train nurses how to use beds, because beds have become smart with AI in them, and they're telling you what the tilt rate is, is a patient at fall risk, are they coming out?(1:27:28) So we're doing it today already. (1:27:30) It's just going to blow up on us quickly.
Eric Reid(1:27:35) Thank you. (1:27:37) Obviously a very engaging panel, so thank you very much, and I have donations in your name for each of you. (1:27:43) Remember to fill out the survey in the app, and also come back at 11 o'clock, but please visit our business partners during this short break.(1:27:53) Thank you so much.

Executive summary

In this panel discussion, CIOs from major healthcare organisations tackle key challenges in healthcare IT, including cybersecurity breaches and the role of AI in transforming patient care. They share strategies for deploying AI in the next 12-18 months, focusing on patient engagement, clinical decision support, and operational efficiency, with examples like AI-powered bots and ambient listening. The CIOs emphasize balancing AI's benefits with ethical concerns, stressing the importance of data governance and ensuring AI complements human expertise in healthcare delivery.