This transcription is provided by artificial intelligence. We believe in technology but understand that even the smartest robots can sometimes get speech recognition wrong.

Today on This Week Health.

it's loud and clear what the question is there. Where does this technology, where does it fall on its face and what is it ceiling. Right? And how are we gonna be able to break through the ceiling?. 

all right. Another keynote episode. We're excited to have Taylor Davis, join us. once again, Taylor. welcome back to the show.

Oh, it's great to be here and on. I think my very favorite topic, so let's talk generative AI and how that's gonna impact healthcare.

we are definitely gonna do that. and, I saw your post on LinkedIn and I thought, it's been too long since we've been together and you seem to be really passionate about this topic and, I was looking for somebody to have a conversation with. It seemed like you were perfect. For those who don't know, formerly, president of class research, you had a role at, h I T Peak Advisors and, anything else we should know about?

Yeah, I am just right now launching a startup, called Care luminate, careluminate.com. And, it is probably not pertinent for today's discussion, but we're gonna be trying to help healthcare improve, so we're really excited

about it. Yep. And you're gonna be, you're gonna be in that research space again and giving us all sorts of insights.

I mean, the last time you were on, we talked the, the arch collaborative. And that has had a profound impact on healthcare, I believe. so many people I've talked to have talked about the, the benefit of having that research, really understanding what, what we can do with the, EHR platform, how to make it better, and some of the findings in that were so, helpful in directing the, the work of the local hospital.

So I'm looking forward to what you're gonna do in this, next round. Generative ai. let's start with the softball. I really do want this to be a conversation. Let's start with a softball, which is, how do you see generative ai, and this is the question of, my report is the world and all of this within it.

So here's the question. How do you see generative AI transforming healthcare as an industry over the next five to 10 years?

Okay. No, that sounds great. The. Let's talk about generative ai. Let me give a corollary to something else that humankind created a long time ago but didn't really understand. So most things that humans create, we actually understand pretty well how they work and what they do.

There was a recent Scientific America paper talking about generative ai and there's a whole field of study that has popped up in the last few months. Of researchers studying why generative AI is working as well as it is, and trying to understand why is that and the complexity. Essentially the way that we built this is we copied

our best ideas and understanding of the human NeuroNet. and we built that into a computer and then we threw incredible computing power and incredible information at it. And its ability to actually retain that information and adopt that information and utilize it, has shocked everybody.

So let me give the corollary, some point we don't know the date. Humans discovered how to make fire and, and that was, they didn't understand. They didn't understand how fire works. They just understood how to create it and and now you all of a sudden have something that is incredibly powerful.

I'm sure it was a game changer back then for them. But also somewhat dangerous. And they didn't understand how it worked. They just knew how to create it. Generative ai, we have to face the reality. Generative AI is the same type of thing. We understand a little bit about it. We understand how to add fuel to it.

We understand that it can burn. We understand that it has heat. But we don't understand all the way, why is this thing lighting up and why is it working as well as it is? That creates, as you look forward into what's gonna happen in healthcare, that creates huge questions about how it's gonna impact.

don I, I'd like to start, but it's really important.

Gives talk about yeah. No, I like the start cuz it gives us, there, there are things that we cannot explain that there, and there was the, I what most people point back to is when, machine was programmed to play the game. Go. And they programmed it with like all the, and it's a very complex game. it's a lot more complex than chess. And they programmed it with all these strategies and then all of a sudden it made a move that was like outside of the realm. and people were like, why? Why would it do that? No one had any idea why it would do that. But essentially it had consumed all these strategies and come up with a new strategy and they said, That almost looks like AGI almost looks like, Yes. Artificial general intelligence. And there's been a couple times with open AI and other things that people are like, that's interesting. That's they really do know. and if they're honest, they say, we understand why it does a lot of these things. It's just math, it's, it's computations and it looks at words and it says, these words go together, these words go together, these concepts go together.

There's actually mathematical algorithms. That they said, the, this paragraph, comes down to a mathematical equation and they haven't really played around with this much, but they're looking at their going, these things can be linked by these, by, by these math equations and whatnot.

But there is, and I think this is where the fear comes from. There is a sense in which there are times where it does stuff and we go, Do you know why it did that? No. Do you know why it did that? No. It's like we don't, the programmers don't know why it's coming down and I think that's where some of the fear is coming from.

it's interesting cuz you have, as I mentioned, you have this branch and I just say this preamble because I don't wanna start talking about how this is gonna help and impact healthcare. Without raising my hand and saying I'm one of those fairly educated humans about how this should work, and I also don't understand how this should work.

And anybody who says that they know exactly why it's doing what it's doing, frankly, is either holding out on the rest of us or is just not being honest. And so we've gotta be really careful. but we do understand a lot of how it works. But I just had one of those experiences yesterday, that, that was shocking to me. I do some programming. I'm not your best programmer and, but I do it in a statistical language called R, which is sort of like Python. I needed to do something really complex, bill, so I needed to simulate. a certain data set that needed to be made. I had all sorts of parameters. I had paragraphs and paragraphs of input, and I took that, I took those paragraphs of input and I already needed to kind of map it out in my own mind and see, okay, how's this gonna work?

I feed it into chat, g p t 4.0, and it pops out with two small mistakes. It pops out perfect code. That would've been almost two days worth of coding for me to, to be able to go too. What's shocking to me is I know how a lot of this technology I know foundationally, how it works, it transforms all of my words into numbers.

It takes those numbers, it puts 'em into it. it's already had, supervised learning. And then there's the transformer. it puts those into the narrow net. It reads through this giant matrix, and then it pops out and. Chat, g p t largely is just only predicting the next word.

And, however it keeps popping out with strategies that are, that, that are very connected, that appear as though, even though that's not how it's written underneath, it appears as though the technology is almost kind of taking a look back and say, okay, well here's the strategy, how we're gonna go down.

But it's actually just putting the next word out there is how the whole model, works. And, and I have to say that I sat back as I, I looked at the code and it's near perfect. It's far better. I've hired full-time r programmers, multiple of them. None of them would have had their first draft be anywhere close to the quality of what chatGPT put out.

And, and I just am looking at the quality of this code and it's really complex code. And I'm like, that is absolutely incredible. So, let's go back to your original question. we're taking the long round around here. Let's talk about healthcare. What do we know about where this technology works?

We know that this technology actually has the ability, so where we thought before that, that a NeuroNet and some of this generative technology. if you would've asked me one year ago, I would say it's very good where you can create a reward function and where you're doing the same things over and over again.

However, what we're starting to learn is that it's actually quite good at some of the creative, aspects of human performance. And we start to actually realize that maybe some of the creative aspects of human performance are actually maybe a little bit, formulaic it in the way that we go about doing these things.

Right? Right. So the funny thing is that we're realizing that this technology has incredible opportunities beyond just repetitive tasks. But if we just look at repetitive tasks, let's take a look at where we are in healthcare alone.

So there, there's a few of these in healthcare, is that what you're gonna

say?

Yeah. There's a few of these in healthcare, right? And, if you know a hospital or a ambulatory or post-acute practice, if you know these types of groups, you, you see, huge amounts of labor being put in these repetitive practices day in and day out. and you look at those and you say, those are the types of things that this technology is gonna specialize in changing.

Let me focus on one area Bill, if we can. That's gonna be really interesting. Some of the most repetitive practices that are, that go on today are in the revenue cycle, and the revenue cycle. you've got lots of people who are employed to do all of these pieces, so, You have the clinical encounter, you, you post charges based off of the clinical encounter, those charges. you have C p T codes and, those. Anyway, there, there's, a whole code system , ICD 10 codes. there's a whole code system, and then you take those codes, you take the posted charges, you look at the charge master, you look at the contracts you have anyway, you build a claim, you submit the claim is denied. And then you resubmit the claim. Anyway, there's this whole process, right? What's fascinating about the process is that. When a payer denies a claim, they're not actually even looking for all of the supporting evidence as to why the claim is there. They just wanna look at the clinical chart and see if the clinical chart supports what the claim is.

And you've got some reporting coming out that payers are already using quite a bit of AI to review, denials and it's. In some cases, and in many cases today, it's not even a physician, even though states have laws that it has to be a physician, the physician is barely looking at what's going on. the chart is being reviewed by AI and the claim is being denied and it's being done extremely quickly. And so on the payer side, you've got a lot of that happening. However, on the provider's side, There's very limited instances where the bill is actually being created by ai and we continue to walk down this kind of tricky cow path, where we're, where we feel like we have to do all of these steps in a row in order to drop a claim, in order to submit it.

Right? and so I think that, if you look at both payers and providers, You've got payers already jumping out in front in some of the adoption of this AI for claims denial. it is going to be very quick that you're gonna see providers realize, well, we need to use a lot more. When it comes to our claims creations, and it's really, I think it's a fairly simple process to see, down the road in a little bit, there's going to be whole parts of human capital that right now is, jobs that are going into doing these different steps that are gonna be cut out. and there's even technology vendors that. Are running some of these that are running the human automation of some of these steps that are gonna be found without a market and it's gonna be pretty dramatic, in the next few years. And then some of the question really comes to what is gonna happen when you have payers and providers.

Are submitting claims back and forth at really zero marginal costs and the amount of traffic going back and forth is going to be incredible. And you have this AI arms race, to be working on claims back and forth. It is gonna be just fascinating to see what goes on and just revenue cycle alone in healthcare in the next few years.

So yeah,

revenue cycle's interesting cuz you could even back up to the coding. Process. So a lot of errors happen in the coding process, and we're seeing. we're seeing nuance come out with Dax Express, which essentially is the computer taking the conversation, turning it into the code.

So now the computer's even going all the way back to the initial visit and generating the note, generating the coding, so that the coding's more accurate to what's going on. I think you're gonna see AI applied to. I mean, we were applying, I don't know if it was AI or if it was predictive models back then, but we were essentially looking at all of our denied claims and determining, what would cause that.

And then we were putting a prediction score on our claims that said there's an 80% chance that this is going to be denied, like even before we send it in. And so somebody could look at it and say, all right, if this is an 80%, how do we get this down to 50% or 30% or 20%? and try to get ahead of it. it, but you mirror that with the automation that you're talking about, essentially from the conversation that the doctor has to coding, to dropping the, claim to submitting the claim, to having the claim be denied to being resubmitted.

And you can see that whole cycle. Just be computer to computer interaction.

It's really hard to say where it's going to go. and so there's the technology piece of this and there's the human piece of this. and both of them are gonna play a role in how does this, how does this take off the technology piece?

We haven't created technology before as humans that has the ability of recursion, of actually creating new technology itself. and that ability opens the door for possible exponential growth in terms of what we're able to do. and so the fact that we have created a technology that can create technologies, that's something that makes people really uncomfortable with, but you're similar.

This is like the iPhone being released. In December, which was G P T 3.5, and then the iPhone eight being released in late March. And that type of progression cycle is pretty startling in terms of where it goes. Let's talk about the human element, bill a little bit. Cause I think this is gonna be interesting and here's a prediction that I'm gonna have fun making.

we've watched for a long time and there are good reasons why very smart leaders. act this way, but technology adoption in healthcare, because of just the high acuity and the complex, service lines that we run, technology adoption in healthcare oftentimes lags that being seen in other industries, right?

It's just harder to implement and you can't implement, or you can kill people. And, and so there are good reasons as to why it should actually lag in healthcare, but, if a lot of the predictions, and it's interesting, you notice a lot of the stories coming out, right now if you're reading the Wall Street Journal or some of these things, there's a good number of stories where, leaders are saying things like, well, it's not necessarily gonna replace jobs, it's just gonna make our average worker a lot more efficient. and I'm here to tell you, as a business leader, if your average worker is a lot more efficient, that just replaced jobs. So, and I think that really what you're seeing is that you're just gonna have less people, you're gonna be able to hire. just like I experienced yesterday, I don't need to hire with our new business.

I don't need to hire some programmers that I would've hired last year if we were launching this new business. Yeah. there are some things that, that I'm able to do just with my own programming knowledge as, deficit ridden as it is riddled as it is, I'm able to do those things because I have this kind of really smart technology with me.

We are going to see an American. business over the next three years are transformation in reduction in this, human capital that are doing non-value generating activities or very repetitive activities and even some creative activities. You're gonna see some of this in the film industry or some of the predictions, and I think that with some really good reasons and whole companies are gonna have very different cost structures, so, some of the predictions right now that, and I think that are well founded, indicate that you'll see at least, short term, you'll see some increases in unemployment, but you'll see some companies with, significantly different profitability, profiles than what they had a few years ago.

And some companies that don't keep up and are gonna be falling under healthcare if we assume that healthcare continues to be a slow adopter. Your healthcare CEOs are gonna be watching what's going on. They're gonna be watching some of these companies that fall behind, and we're all gonna be shocked to say, wow, that company was looking really healthy three to four years ago.

Now they're gone. And they're gone because they weren't able to keep up on this incredible acceleration curve that was going on. you're gonna see some other companies that shed you, the increase their bottom line by two or three x. And, and they were able to do this by harnessing this technology.

And hospital CEOs are gonna realize in, in, I'm guessing, two to four years, they're gonna start realizing, oh my gosh, we're next. And, and so today where you've got a little bit of a, okay, well let's, kind of be a little bit careful. In two to four years, there is gonna be a drum beat of we have to do this and we have to do this really fast.

Because there's a competitive pressure, and then if we don't do it, we're gonna, we're really gonna fall behind. So two to four years, healthcare is really, it's gonna pick up and it's gonna look a lot different. And so don't kid yourself in the next two years if you don't really see it picking up in the way that you expected to in healthcare.

We'll get back to our show in just a minute. I am excited about our webinars this year. They have been going very well. What I've done is I've gone out and talked to people in the community and said, what works in webinars?

And they came back and said, look, this is what we want. We want a webinar that is not product centric. It's really focused in on the problems of health care. And we want people on there that are actually solving those problems. And so we have done that. And the response has been fantastic this year. We have another webinar coming up.

It is the future of care spaces. Where care is being delivered is changing rapidly. Even the care spaces within the hospital themselves are changing. Technology is being added in different types of technology. A. I obviously computer vision and whatnot is changing that modality as well as what's going on in the home and whatnot.

So we're gonna have that webinar June 8th at one p. m. Easter time. We usually have it on the first Thursday. Happens to be a little too close to my anniversary. So we're going to do June 8th at 1 p. m. Eastern time future of care spaces. We would love to have you be a part of it. If you are interested in being there, go ahead and hit our website.

Top right hand corner. We have a card. You can click on that card and go ahead and fill out the form and get registered today. We would love to have you join us we look forward to seeing you there. Now back to our show.

So let's talk about two things. See, I think some people will listen to this and go, oh my gosh, Taylor is so wildly crazy. Totally. I've used chat g p t for, it's not doing this, and this. Let's talk about two different things. Let's talk about, let's talk about, I think it's called auto, g p t. have you used that? yes, I

do.

Yes. All right. So, so, we'll do that and then we'll talk about, gosh, I'm blanking on the graphics program. mid journey. Mid journey. Yeah. Talk about mid journey a little bit and some of the things that are going on there. But let's start with auto, G P T.

Because I think people would be surprised what's going on in that space. You could actually have a, an AI engine auto G P T that is spawning other AI things and actually creating or solving problems. So think about it like, Hey, I want to, let's take something very simple. I wanna write a new website.

And you feed it a bunch of information, and then it'll go out there and spawn and do different things and actually, generate all the code, everything you need to do the, to essentially launch that website. And I could see, it, it's not hard to see it from here, where essentially at some point you put that in there and it actually launches the website.

Puts the website out there and starts populating the content. In the website and you're done. And think about all the people you just displaced, graphic designer. web developer, ui ux and all that stuff because the, it's able to tap into all these bodies of knowledge to say, good ui, UX looks like this.

Good design looks like this. Good website design looks like this. Good positioning and marketing looks like this. what are some ways you've seen or you see that kind of, an auto G P T kind of thing. taking it to the next step where it's not just, prompt response.

Prompt response.

auto G p T is interesting because it's not, releasable code. I mean, it's just on GitHub, right? So it's just developers that are implementing this, and you've gotta have docking, you've got, you've gotta implement, right? So today, getting that going, and there's there's a website God mode that allows you to take your a p i key from chat G p t and plug it in.

And you can get a lot of the same benefits today. it's, there's still a pretty big gap as you're using this.

it's na, it's nascent technologies,

it's nascent technology. Right? But the fact that it's able to do what it's able to do, and so for the next little bit, we're going to see in a lot of places, we're gonna see this technology needing to run side by side with somebody who's really experienced to be able to run it.

So I, I use chat g p t to build the R code that I did yesterday. It would've taken me a long time if I don't know, r very well to look through and to find those couple errors. However, I know r quite well, so I'm able to look at it and to be able to find those areas very quickly. I don't have a lot of good experience or knowledge or, professional ability when it comes to creating movies.

So if I take Mid Journey and I go create a movie, I've got all sorts of gaps because it gets me 80% of the way there. as any of us who watch football knows. If your team always goes down the field, 80 yards to the 20 yard line, you still don't win the game. And so, current, currently people are looking at some of what these technologies they're saying, well, it only takes you 80% of the way. and that's not good enough. but the fact that a nascent technology can take you 80% of the way is incredible. And so auto, G p t does exactly what you're, what you talked about it. Chat G p t right now can't go browse websites. Auto g p t can go browse the internet for you. That's a huge difference.

And and so you can ask auto g p t today, you can say, I really need to, let me give an example. I needed all of the names and emails of, of major contacts for all of the media outlets in the Cleveland area. That was something I needed last week. I gave auto G P T, I said, could you find that for me? It comes back and it gives it to me.

And it was a really well done list. And, And that's something that before you would ask an intern, you'd ask somebody to go and work on. That would take you, I don't know, an hour and a half for somebody to go through and do that. Auto G p T goes and does that in just a matter of a few minutes.

But what's fascinating Bill, is you look at, how did it do that? It actually wrote Python code that could scan the web. According to the parameters that you were looking for. And then it implemented that Python code. Then it scanned the web, and then it brought down what it found and then it wrote another code to go double check what it found to make sure it was accurate, and then it popped it out.

So when you actually look under the covers of what it created and what it did is you're looking at auto G P T, it starts to be pretty sobering. You start to realize, wow, this actually. It. It created all of those steps. I didn't create those steps. It created all of those steps to know what to do.

It created these two different programs that ran them against each other to check its own results, and then it popped out. The result. When you start to see what goes on underneath the covers, you realize this technology has incredible potential to, to really move us forward in an accelerated rate.

And, in a similar way, that's what this technology has. Hopefully we make some good choices as a society, and we use it in wise ways. It's not a pandemic, but it's a really positive thing for us.

you bring up a great example of how it's being implemented in healthcare, so Stanford.

And, Stanford, uc, Davis are looking at it in terms of responding to the inbox. The inbox has been a yes. Ongoing problem. Great. Yep. and so it, it creates an empathetic response and a fairly accurate response. And it, it saves it as a draft. So, instead of having to create the entire response, the physician can come in there, review the. the prompt from the, from whatever the prompt is, and then review the, the response and say, yes, that's accurate. Yes, that's good send. Instead of creating it from scratch, and now you're talking about, something like that. You're talking about saving a couple minutes, a message, but you're also talking about hundreds and hundreds of messages.

So you could potentially save physicians a fair amount of time. but it's that, human supervised oversight that gives us the ability to use that kind of tool in that kind of setting and not be concerned about, oh, chat g p t didn't go to med school. Yes.

So, so here's another prediction.

This is all forward looking and. most predictions are wrong, but here's another prediction in terms of where this goes. if I'm guessing so today it's human supervised. Your average physician at 6 45 at night doesn't feel that empathetic as they're responding to messages from patients, before they leave for the day.

And so it fills this great need. I can just review it really quickly and correct press send after I've seen 20 of these messages that all look really good. I barely look at the messages and I press send at night and that's already happening. Right. We just, we know how humans work.

Right, right. when it comes to, AGI or these large language models actually being able to replace physicians, let's talk about where they can replace physicians and what that looks like and some of the benefits that actually this brings to us as a society. and so, now we're not talking revenue cycle.

We're talking some of the most complex parts of healthcare, however, There are some specialties that are very repetitive in terms of the types of what they're seeing and the types of instances they're seeing and what does this look like? so a prediction is that already you start to see chat, G P T, helping improve care in countries outside of the United States.

So if I am in an African country or a South Asian country, and I really, I'm fairly isolated. I struggle to get to a physician. Any physician that I get to, maybe wouldn't even have the certifications of an RN in the US right? but I have access, I have a cell phone and I have access to chat, G P T. already and increasingly I'm going to be able to go to that language model in order to, to receive care. An inflection point happens when, we as a human race, start to study some of the care that's going on there and we realize there will be a place. I predict that. and many are predicting where the interactions.

There's so much of diagnosis today of that initial interaction with your physician that, that there are biases that really create issues as you come in. There's a lot of studies that say that as a human, I can only think about four to eight factors, when it comes to diagnosis, and I can't think about more than that.

And so gender is a factor. Gender is a dimension that I can think about. weight is a dimension, a symptom is a dimension. So I easily have patients that come in and they will drop on me 25 different dimensions to their problem. And, and it's going to make sense that very soon this technology will be a lot better and more accurate at diagnosing where things should go than humans are. and we likely may even see some opportunities. On the diagnosis part of healthcare, which isn't everything by any means, where other countries almost leapfrog the United States because they start to outsource, this piece to, to a technology that can do it better than humans can. And, it's a long time before a surgeon is able to do this better or to where humans trust a surgeon, but, Given the experiences that we're having with chat G p T and that your average American is having with chat G p T, it seems likely that they're going to start feeling even more comfortable in many cases, receiving a diagnosis from chat G P T than they will their physician. and that's gonna be a very interesting point where you see that working well in other countries and you see, Americans starting to really trust these language models quite a bit. And going to the language model, which they're already doing. They're going to Dr. G P T and then they're struggling to understand the distance between Dr.

G P T and actual physician.

It's, it's interesting cuz yesterday there was an announcement yesterday, the day before, Hippocratic is building a large language model for healthcare specifically. It's funded by General Catalyst Andreessen Horowitz, and they, essentially large language model taking it to med school.

And this is what we thought would eventually happen, is we'd have these. Massive models become more focused. maybe one's gonna be focused on FinTech, maybe one's gonna be focused on healthcare. And so this one, I mean, their claim to fame here is they have all these different tests that you take, nursing tests and whatnot.

And when you look at Hippocratic, it gets essentially an A or a B. In most of these tests, there's like 15 of them have hospital safety training, the nurse, Nurse practitioner tests and others that, that are here and they get A or B rating. And then you look at G P T four and it gets a C rating in most of 'em, and one or two of 'em, it actually fails.

And you go is, yes. I agree with you that we are gonna have some epic failures. I mean, just somebody goes to chat, g p t, they get the wrong diagnosis. The hallucinations. Absolutely. Yep. it's gonna go. And, but is that gonna lead us? I think because of the momentum it already has, we're gonna plow through that.

We're just gonna say, all right, we need to train the model better. We need to get it more data in order to make sure. Cuz that's, if that's where people are going, we need to make it better. And we saw that with Google. Google search went out and hired all these healthcare people cuz they're like, Hey, you know what?

People keep coming here for healthcare. let's bring on, really credible people who can make sure our searches are as accurate as possible. It's putting out credible medical. Data. It also has the warnings in there that says, go see a physician if you have this kind of diagnosis. I'm not sure what the question is there other than Well, no,

I, it's loud and clear what the question is there. bill, where does this technology, where does it fall on its face and what is it ceiling.

Right. and how are we gonna be able to break through the ceiling? I cut you off and I apologize.

I get so excited. no. that's a really, it's an interesting question. I, by the way, I think there's tons of low hanging fruit. You mentioned revenue cycle. I think there's a ton just within it.

Like administrative task programming. You mentioned programming. I've used it for programming. Yep. And yes, I know the programming language that I asked it to program in. So I was able to look at it and go, ah, I wouldn't do the variables that way. Let's change this. Oh, by the way, let's document the code.

And, and so I changed a bunch of things because I had skills and you had the skills in r to do the same thing. But at the end of the day, when you look at the things it can do, I would imagine I could set it on a, a log file. let's take something you're very familiar with. I can set it on the Epic log file.

And I can say, yeah, tell me, tell me the 10 physicians that are struggling the most. Now I, I realize that Epic does give you some of that information already, but this thing could get more specific and say, Hey, it's struggling here. It's struggling here. this physician's struggling here. we could be 20% more effective if these physicians were taught these things, around that.

Would you like me to generate, some training for these? I mean, you can see that kind of thing. On the administrative side, it revenue cycle, just your basic ordering tasks, the scheduling tasks, the call center stuff. I mean, there's a whole host of areas where you can see significant efficiencies pretty quickly. and

let's talk about, building on that bill and I love where you're going. This is going to be a divisive technology. Health systems that know how to adopt this well are going to plow ahead. They're going to reduce their costs. They're going to see possibly huge improvements in the quality of care.

Those who don't know how to adopt this well are gonna see significant challenges and a lot of the hinge. Between those who do well. And so the haves and the haves nots when it comes to this technology, and everybody's gonna have it. But a lot of the hinge will actually be the dos and the do nots.

Do you know how to adopt this? Do you know how to utilize this? Do you know where to utilize this? Just yesterday for something I, I am constantly watching. and I realized a way that, that this technology can help me. All of a sudden I was like, oh my gosh, I did like four hours of that last week, and I could have done it in under 10 minutes if I had, utilized this technology.

It was a non-standard way, to the point where you talked about with the chest moves where the computer says, do this move, even your grandmaster goes, that looks a little weird, but then it ends up winning the game. You're gonna have some health systems that help their team through those moments and help them push through wisely when they need to, when the computer is pushing something that they didn't know and what we're gonna have to get really better.

Really good at is humans is in complex situations, not disregarding immediately something that these models bring to us that are interesting, but being able to quickly be able to dive in and diagnose, was that a hallucination or was that actually a really helpful solution that just really doesn't mirror what typical human solutions are.

My best guess is that there's going to need to be training around this. There's going to need to be firm strategies around this. we haven't even talked about, what are some of the HR strategies, because you're gonna have a, you're gonna have, one department where, either through attrition or through layoffs, they're able to, they're able to, to move forward.

And now other departments are gonna say, Well, we don't wanna bring this technology in, we wanna fight it because we just watched what happened over here in revenue cycle where they laid off 70% of the pre-auth, or 90% of the pre-auth team. We watched what just happened there, so we're not gonna be lining up to go adopt this technology really well.

Right. So there's. That having a comprehensive strategy as a health system today to be able to navigate those points of distance between the human and the model to be able to navigate those HR situations, to be able to say, where are we gonna adopt this first, and what is it gonna look like? And to have a vision as to what the final destination is gonna look like.

Doing that is gonna be critical. and the execution of those strategies are gonna be, are gonna make or break health systems over the next 10 years. It's gonna be, or even five years. It's gonna be fascinating to watch.

Yeah. it is gonna be fascinating. as I think through what you're saying there, the, the implementation becomes really interesting and obviously as you're talking, I'm thinking there's a room for a consulting firm that just. helps healthcare organization or any organization, work through a lot of these, a lot of these challenges that are, and opportunities that are right there in front of us. I will remind people. That when the PC came out, there was a lot of this kind of talk. It was like, oh my gosh, we're gonna, we're gonna lay off all these people, we're gonna lay off all these people.

And at the end of the day, what we generated was new industries just completely never existed digital industries before, and people got jobs in those digital industries. So, there, there are gonna be opportunities. I'm curious, from an implementation standpoint, I think if I were a CIO for a health system today, one of the first things I would be doing.

Is, I'd be looking for a model that I could train on our data in a, with a, b, a with the right protections around it, but I could start feeding literally our e h r data into it, our financial data into it, our, workflow data into it so that I could start asking it questions like, where do you see, again, a lot of training has to go on, but, where do you see flaws in our workflows?

Where do you see efficiencies that can be gained? Where do you see. where are we, not billing, where are we losing? I mean, that's, these are the kinds of things it can make those connections and those inferences, maybe not in its current form, but in a form. Soon to be released. We're gonna see it.

In fact, we're seeing Moose of You said, earlier, Hey, chat, g p t can't read the internet, and I'm on it today. And it says, Hey, would you like to do plugins? Oh, by the way, we added a plugin where can read the internet. I'm like, oh, I'm excited. That's great. You gotta be kidding me. I mean, it's like, it's moving that quickly. I know that Microsoft offers access to G P T four through a baa. You could stand up your own instance, feed it with your own data. so, so you have that opportunity. The other thing is how do you create a culture that's accepting of it? Yes. How do you encourage it? How do you, what kind of training makes sense at this point it's, at a basic level it's prompt engineering.

Right? You're probably getting better and better every day at absolutely writing the prompts to get the things you want out of it. What other things would you recommend, do you think, if you are. Trying to prepare your health system for, to really take advantage of generative ai? what are some of the things you would be doing?

just because the technology is moving exponentially faster doesn't mean that the humans are going to, so things move fast, people move slow, and at the end of the day, healthcare is a services industry. there, we are going to need to continue to value our people. They're going to do great things for us.

There is gonna be some shifts and health systems that fight this. There's just gonna be no way to fight this. So there will be some shifts and there will be some team members who are gonna have to, possibly leave reskill, come back in a different role. But the, there's never. been a time, healthcare has already bruised coming out of the pandemic. quality measures are down. Staffing levels are really challenged. There's never been a time where leadership was needed more in healthcare than I would say today, at least in the last 50 years. and so leaders are going to really need to lead. And this is not all about the computer element.

This is gonna be largely about the human element and how are we going to adopt this technology and how are we gonna implement it? and we're gonna need to have some empathetic discussions as leaders, with our teams and really listen to them and work with them. And then we're also going to need to effectively bring reality into the room as we talk about where this is gonna go and the faster.

We educate ourselves to be able to understand what some of the possibilities are and where this can be adopted faster, and that we're able to, once we're educated, be able to work with our teams is gonna be really critical, to helping our organizations be able to adopt this. Well, so you bring up some great points, but step one.

Is as, as healthcare executives is, we've gotta educate ourselves about what the possibilities are. What does this technology entail? What can it likely do and not do? It's like fire. We don't know all the way what it can and can't do, but what are the likely can and can't do? and then from there, we need to start building strategies.

And we need to start communicating and building holistic plans.

Right. Taylor, I want to thank you for your time and, I look forward. Oh, that's so fun. Yeah. Well, I look forward to what's next and we'll have to, we'll have to come back and see how many of your predictions come, come true. Well,

I got 'em all from chat, g p t, so they should all be honest.

Chat G p t predictions about generative AI should be sold. exactly. Hey, thank you so much. Thank you. Take care.   (Main) 📍 📍 I love the chance to have these conversations. I think If I were a CIO today, I would have every team member listen to a show like this one. I believe it's conference level value every week. If you wanna support this week health, tell someone about our channels that would really benefit us. We have a mission of getting our content into as many hands as possible, and if you're listening to it, hopefully you find value and if you could tell somebody else about it, it helps us to achieve our mission. We have two channels. We have the conference channel, which you're listening. And this week, health Newsroom. Check them out today. You can find them wherever you listen to podcasts. Apple, Google, overcast. You get the picture. We are everywhere. We wanna thank our keynote partners, CDW, Rubrik, Sectra and Trellix, who invest in 📍 our mission to develop the next generation of health leaders. Thanks for listening. That's all for now.