Webinar - AI: A Valuable Crew Member, Not the Ship’s Captain

Aired on: Thu, Aug 08, 2024

314e at HLS Symposium 2022 | 314e Corporation

Discover invaluable insights on the use of AI in healthcare IT from our recent webinar, “AI: A Valuable Crew Member, Not the Ship’s Captain.” Learn how healthcare IT experts are leveraging AI to transform the industry, with practical insights and real-world examples you can apply in your organization today.

Key Takeaways From the Webinar

  • Understand the true capabilities of AI and dispel common misconceptions
  • Discover the potential and limitations of AI in the healthcare sector
  • Learn how AI is transforming healthcare leadership and operational efficiency
  • Explore practical use cases of AI with Jeeves, Dexit, and Muspell Archive
  • Gain tips on using AI creatively to meet your health organization’s needs
Ryan Seratt

Speaker 1: Ryan Seratt
Director of Training and Development, 314e Corporation

Ryan Seratt is the Director of Training and Development at 314e Corporation, with over 20 years of experience as a trailblazer in EHR training and development. As an AI pioneer, Ryan has set new standards in the field, crafting innovative solutions that transform the way EHR systems are trained and upgraded. His deep expertise in software, leadership, and e-learning drives effective and impactful training experiences.

Alyssa Dennis

Speaker 2: Alyssa Dennis
HIM Product Manager, 314e Corporation

Alyssa Dennis is the HIM Product Manager at 314e Corporation, with over 12 years of experience as a leader in Health Information Management. A healthcare data management powerhouse, Alyssa is an expert in crafting AI-powered solutions that streamline the management of medical records across multiple facilities. Her deep knowledge of the latest trends and best practices makes her a trusted authority in the field.

Here is the transcript of the webinar

Casey Post 0:00

It's here. Let everyone join. All right. I see a bunch of people coming in. We'll give it one more minute and then we'll get started.

All right. Good morning. Well, thank you everyone for joining us today for this exciting webinar. My name is Casey Post. I'm the Senior Vice President with 314e. And I have the honor of introducing today's speakers - Ryan Seratt and Alyssa Dennis, who will be discussing the hot topic of artificial intelligence in healthcare. Before I introduce our speakers, I'm just going to take a few minutes and some housekeeping items and also provide some background on 314e Corporation. Today's call is being recorded. Everyone's mics are muted and cameras are off. However, if you do have questions, please put those in the QnA section of the webinar and we'll leave time at the end to answer as many questions that we can with the time allotted. The slides and the link to the recording will be available after the session and those will be sent to you.

So, let me take the next minute or so and share some background on 314e. 314e has a 20-year history exclusively focused on delivering modern healthcare IT software and services to provider and payer organizations. Headquartered in the Silicon Valley, we have a passionate team dedicated to crafting innovative solutions that drive positive change within the healthcare industry that leverage cutting edge technologies. We understand the unique challenges faced by healthcare organizations and have tailored our products and services to meet their specific needs. We take pride in our diverse range of products designed to address critical healthcare challenges. Our solutions encompass a healthcare data archive, a Document Management System that leverages AI for automated indexing, a Content Management System, a Just-in-Time AI-enabled training delivery software, a healthcare data analytics platform, a cloud-native integration engine. And each product is carefully crafted to enhance the efficiency, streamline workflows, and improve patient outcomes by leveraging AI and modern technology. And in addition to our product suite, we offer a comprehensive range of services that cater to various aspects of healthcare IT, including EHR implementations, integration and interoperability, analytics, training and e-learning services, Revenue Cycle Management, and automation and AI. Our team of experts is well-equipped to guide healthcare organizations towards success in this rapidly evolving digital era. Over our nearly 20-year history, our solutions have benefited many of the country's largest health systems, including academic medical centers, county and community hospitals, as well as critical access hospitals, FQHCs, and independent physician groups. Our affiliations and partnerships allow us to stay current with industry trends and provide strategic advantages for our clients. So, without further ado, let me introduce today's speakers.

Ryan Seratt is 314e's director of training and digital development. He has been a digital learner, learning leader for over 20 years. He is passionate about leveraging technology to connecting people with information they need to excel at their work. Recently, he's been very involved in how AI is affecting learning. He has given over a dozen talks and is a contributing author for the book Transforming Learning Design with AI ChatGPT. He has been a guiding force in the development of our very own Just-in-Time AI-enabled training software called Jeeves.

And Alyssa Dennis is 314e's HIM product manager. She has extensive experience in leading and participating in archival projects for major acute medical centers. She brings over 10 years of HIM operations leadership experience, combined with Six Sigma certification and credentials with AHIMA. She has been integral in the development of 314e's archive product, as well as our automated indexing solution and Document Management System called Dexit. With that, I will turn it over to our speakers.


Alyssa Dennis 5:23

Awesome. Thank you, Casey. Wonderful introduction. So we've got a lot that we're going to go over with everybody today. We're going to do a little bit of level setting early on, talking about what AI is, what AI isn't. We're going to be talking about some of the different examples that you see in today's healthcare sphere and how AI is already being utilized. And then we're even going to show you some of the ways that here at 314e, we're utilizing AI right now today with some of our products. But before we do that, we really just kind of want to see how everybody feels about AI. You know, it's a huge buzzword. Everybody seems to have an opinion about it today. But if I were to ask you right this second, how comfortable are you using AI today? Let's go ahead and take a poll. So let's go ahead and kick off the poll. And everybody, you're going to see it pop up on your window. Go ahead and take just a couple of seconds to answer. If you don't want to answer, you can skip it. That's totally fine. But let's just see how everybody is coming in. All right. So we're kind of across the board, Ryan, and I haven't seen anybody claim extreme comfort. There's one person. I stand corrected. Yeah, I think we're, you know, we're all kind of starting to use AI in different ways. Some of us are using it a little bit more than others, but a lot of us are kind of there is a lot to learn and it does change very quickly. So I think that's why webinars like this are so popular, as we're all trying to figure out how can I use AI going forward? And then how is it changing on a daily or weekly basis? Exactly. So. All right. So let's go ahead and end that poll. But from what I'm seeing is we have a mixed bag. We have folks that claim that they're not at all comfortable. And, hey, I get that. It can be extremely intimidating. And we have a couple of souls out there that say, hey, they're ready to tackle whatever's coming. They're very confident. So thank you for taking that time. That gives us a couple of ways to go through our material. So we're going to make sure that it's obtainable for everybody.


Ryan Seratt 7:27

Yeah and as we go through the presentation, as Casey said, if you've got questions, please do take the time to put them in the Q&A. We'll at the end of this, we'll take everyone's questions and answer them to the best of our ability.


Alyssa Dennis 7:40

Absolutely. So first off, before we can talk about AI, I think we've got to take a few minutes to really just try to understand AI as it stands today. Because, again, like you said, Ryan, it changes so fast. So, you know, if I know this is being recorded. So if you watch this webinar in six months, you're probably going to chuckle since a lot of what we say is probably very outdated. But as of right now, today, how we understand AI kind of comes in two different ways. So personally, I've found that a lot of people, when you talk about AI, you know, immediately we jump to the media. And that makes sense because we absorb media all day, every day. But a lot of people are either going to come in and say Skynet or Enterprise. There's very few in between. And I think that's really funny.


Ryan Seratt 8:27

Yeah, I think that when we talk about that, I think the Terminator is our fears of AI kind of come to life. There's a lot of movies, there's a lot of books, and there's some apprehension with AI. And I think that the Terminator movies represent that very well. I think on the flip side of that, Star Trek has always been a utopia of the future. And the way that they use AI in Star Trek is what I personally think that we're heading towards a little bit more. Is that the AI being a tool to help us do our work, help us be in our personal life as well. But really being an assistant to us and helping us create more and create better products, learning outcomes from what we're trying to do.


Alyssa Dennis 9:18

Exactly


Ryan Seratt 9:19

And I'd like to show a couple of videos specifically around Star Trek and what their use of AI is. And Alyssa, what are your thoughts on the two different viewpoints that we're demonstrating here?


Alyssa Dennis 9:33

You know, I think it is very, I think there's, to be honest with you, it's almost like when you're looking at it, the emotional response is more of that Terminator-esque because, you know, there is fear of the unknown. We're human. That makes very good sense. But then on the flip side of it, when we look at it from the Star Trek perspective, like you said, that is more of that utopia. And it's really interesting because we'll see that in just a few minutes in those video clips. But the way that they utilize AI is really what I personally feel that we need to start working towards and become better suited at is it's a tool. It is literally a tool. It's no different than a calculator or if you want to go way back in abacus. It's really a tool that is going to be used to augment how you're already thinking, how you're already problem solving. And I think that it's really important to focus on that versus the potential fear of that unknown.


Ryan Seratt 10:28

Right. The increase in productivity that it's going to bring versus going to lose my job.


Alyssa Dennis 10:33

Exactly. Exactly.


Ryan Seratt 10:36

Let's take a look at some of those videos. The first one that we're going to play is going to be with the - from the original with Captain Kirk. So I will let everyone know the audio on this is a little bit low, but this also came out 55 years ago. So this video is from 1968. And we're going to kind of pick it up in the middle of the conversation that Captain Kirk is having with the computer. But I think what's really important for you to see here is the way that they're interacting. And I think that's very much how we're interacting with ChatGPT or Bard or any of the other large language models that are out there. So let me play a short clip and I'll stop it and we won't watch the whole thing. And I'm going to make sure that I did share audio because that seems to be an important part of the video clip. And here we go.

(Video clip plays - 11:39)

Last nine years. Has daughter Lenore, 19 years old, traveling with. Give comparative identification between the actor Caridion and Governor Kodos.

No identification records available on actor Anton Caridion. Give information on actor Caridion prior to Kodos death.

No information available Anton Caridion prior to 20 years ago.

Photograph Kodos. Photograph Caridion. Now photograph both.


Ryan Seratt 12:39

So, Alyssa, what did you take away from that short clip?


Alyssa Dennis 12:42

Yeah, there's a couple of pieces that I love of that clip. So, like, right when you start playing it, he says, stop. OK, anybody who's used chat GPT, how many times have you typed in a prompt and it just starts rambling and you're like - ugh just stop, hang on. And then I also really love how you can see he's just sitting there at one point before he turns the monitor. He's just sitting there. You can, you know, the gerbils on the wheel. He's processing. I've been there, too. And again, remembering that this is from 55 years ago, because just yesterday I was using ChatGPT and it threw out a bunch of stuff. And I sat there and went - well, now what do I do with that? OK, because, again, rarely I to be honest with you, I don't think ever I've been able to just take the very first piece of information that you know GPT gives me and run with it. It always takes some refinement. It takes some decision points. It takes that human interaction to really mold it into what it needs to be. And 55 years ago, they hit the mark.


Ryan Seratt 13:39

Yeah, I completely agree. The thing I liked about this, this clip and the next one that we're going to see is that humans are doing the problem solving. The AI is providing a lot of the - a lot of the information. But really, the context is being provided by humans and humans are problem solving and they're getting information from the AI, which it's able to do very, very rapidly, much more rapidly than what we could. But it's able to sum that up and give it to us as we need it to help us make better decisions.

So I want to show another clip. So we're going to jump from 1968. This is about 1990-1995. So we're going to jump 20-25 years in the future and still stay with Star Trek. But we're going to go with Star Trek, the next generation, and we're going to see how the character Jordi LaForge is using the AI to problem solve again.

(Video clip plays - 14:40)

Computer, locate Commander Riker.

Commander Riker is in his quarters.

Computer, locate the electropathic pattern described in medical database 4 Delta 1.

Electropathic pattern located.

Okay, we're going to track down any possible cause of that brain pattern. First, what are the medical conditions that might... account for it.

Irisine syndrome is a cause of the electropathic pattern.

Right, Dr. Crusher already eliminated that, so how many non-medical forces or substances might cause the pattern?

There are 22 non-medical agents.

Okay, and how many of those are present on the Enterprise?

None are present on the Enterprise.

Are the ship sensors calibrated to detect all those agents?

All agents are scanned by the sensors.

Okay, let's start with the chemical substances. Name them.


Ryan Seratt 15:38

We'll just go ahead and stop it there, but 25 years in the future, using it very much the same way as it was envisioned earlier, but a little bit more refined. And then of course, we jump 25 years in the future, and that's where we're at today. But it's kind of neat seeing, you know, what this looked like 55 years ago, 30 years ago, and kind of where we're at today. We're finally jumping in.

Alyssa Dennis 16:02

Well, and I really like, too, that this is a perfect example of what natural language progression is, because, you know, he's talking to the computer as if it's somebody in the room with him. And whereas we saw that with, I almost called him Dr. Kirk, Captain Kirk, he was very, you know, cut and dry, black and white with his statements, you know, show me, Coridian, show me. Whereas he, you know, Geordi was just having a conversation. And really, that's, I think, where we are at today. I know I'm guilty of it. When I go to chat GPT, or when I go to Google, I'm very much into just typing it in, ask my question exactly the way it's fired in my brain. I'm not thinking about doing a Boolean search or anything like that. And I really think that's also another way that we can utilize AI as a tool, where you don't have to take that extra, you know, split second or the extra brainpower to reframe, you know, refigure your thought, to be able to put it in a way that you're going to get the best search results. Instead, you just throw it out like you're talking to your buddy, and you're going to get viable results back.


Ryan Seratt 17:06

Yeah, I know we're going to talk about Natural Language Processing a little bit later on. But really, that's the leap that took place almost two years ago, year and a half, two years ago, with ChatGPT, is now we can actually inquiry, we can create inquiries, and I don't have to have a programmer do it. I can just tell the system myself, in my words, what I'm looking to do. And it's going to do that. I don't have to have programmers, I don't have to get on someone's, you know, docket that three, four months later, a programmer comes back and says, okay, your search is now ready. I can actually do it myself, I can refine it in real time. And I can get something that helps me in my day, and helps me be more productive.


Alyssa Dennis 17:53

Which is great for me, because I am not a patient person. So I can really appreciate that, you know, give it to me now. And the fact that it is becoming more obtainable for just general use. So you know, like, as we've been talking about, these are really good use cases that have been showcased in the media. But I still think it's important that we need to also just level set what AI is not. Because again, you know, as we're using the media to show what it is, you still kind of get those perceptions of it's mythical, it's magic, it's, you know, it's a human behind that keyboard talking back at you. And it's really, really not. And I know, personally, I really have to remind myself, even today, that it's not a magic fix-all, I can't spend, you know, two seconds of effort, put something into again, ChatGPT, and get out viable results instantaneously, I have to be committed to being able to absorb what it's giving me, and being able to change it and augment it again, start pinpointing exactly what I'm trying to get. The only thing that it's doing for me is it's taking that huge, vast quantities of data that I as a human could not truly absorb. And it's able to funnel it down into a bite size piece, you know, essentially eating the elephant a bite at a time, and giving me that bite size piece that I can consume with my comfort level.


Ryan Seratt 19:11

No, I think that's absolutely right. I remember having a conversation with Ray Jimenez on AI. And he said that just like - he's like, AI is another person that you need to manage that's on your team. So as we create teams, and people are doing work, you always have to take a look at the work to make sure that it's aligned with the project, and that it's going to reach the goals. So going back, you know, number two stands out that it's not perfect. So sometimes AI will give you results that aren't exactly what you looked for, just like a person on your project team would. And you can rephrase that, send them back to recreate that work and bring it back. I think that, you know, the expectation is, if it's on the computer and on the internet, then it's true. Well, that's not always true. If you've ever been to Reddit, you know that that's not even true today. So if we don't expect that from each other when we're talking, how do we expect a computer to sift through all that and come back with absolute truths? That's not something the internet actually contains. So there are times you do need to edit it. It's not magic. It should, it does show up, but a person needs to manage the AI just like you would a person.


Alyssa Dennis 20:26

Exactly. And context is so important. Like you mentioned Reddit, maybe there is good information out there. I have yet to find it, but maybe there is good information there. But again, the context matters. You have to be able to look at where that information is situated, how it's being presented. And again, I'm not convinced that we're there yet when it comes to AI, because again, it's just going to provide you those facts. The human piece is you have to interpret it. You cannot give up that ownership. You still have to oversee it. You have to manage it. You have to have certain expectations. It - exactly like you said, AI is a team member. It's not a separate entity anymore.


Ryan Seratt 21:05

Yeah, absolutely. And I think that, you know, that kind of goes back into, you know, we talk about number five, it's not an expert and humans need to be the experts. And AI is doing a lot of the, a lot of the legwork to gather the information, to process it, but it's only going to do what you're asking it to do. It's not going to be able to make a huge work plan of what you should do. You can, it can assist you in that, but really a human's hand needs to guide the conversation.


Alyssa Dennis 21:38

Exactly. And I mean, and I think that's actually a perfect use case, even just in a personal life. I mean, I sometimes can feel very overwhelmed with my to-do list, even being able to just put in the to-do list and say, help me prioritize and start, you know, it'll might give you something where you feel that prioritizing is not right, but hey, you're already progressing towards that goal. You're able to look at it. You're able to augment it. Sometimes I think half the battle is getting started. So being able to utilize AI as a jumping off point to then come back in, reevaluate and start, you know, that creative thinking, I think is really, really helpful. And that's a lot of that where efficiency and productivity comes in because you're able to bypass that start paralysis that I think, at least me personally, I can be captured by that.


Ryan Seratt 22:23

Yeah, absolutely. And I think that on this list of what AI is not, AI is not an expert in your workflows and your business. The AI that's out on the internet is really the internet's version of AI. And it has a lot of great kind of common knowledge information. It even has very specific things that it's learned, but your business practices are not out on the internet. At least probably half of us in this room work to make sure that our business processes are not out on the internet. So whatever AI you're using and you're partnering with needs to have your specific business processes in there to be really, really useful. So something to keep in mind as you move forward, that the general public version of the AI is very different than the corporate trained AIs.


Alyssa Dennis 23:21

Yeah, absolutely. So now that being said, again, like, remember, if you come back and view this six months from now, this may be outdated because it is changing almost daily, I swear. But I do think that there is some key components that we're going to see continue to transition and continue to grow as these creative leaps happen. And so one of the pieces that I see a lot, especially in the products that I oversee here at 314e, I'll show you that shortly, is the Machine Learning. And that to me is, you know, that's where the machine is able to essentially learn from the process and the progress that's being made. And it's not necessarily a case of where you have somebody in a back room, you know, a dark room anymore, feeding in examples to the computer. You've used, generally, you've used a large language model and you've set a basis of knowledge. And now the system is going to continue utilizing that to make different leaps of logic, take corrections from humans, and it's going to kind of just funnel you as you go. And I think that's really, really important when we start talking about efficiencies and we start talking about productivity, because nobody likes to do duplicative work. So the fact that if you correct a mistake, and now you know with fairly decent certainty that that mistake will not be made exactly again, that is really going to push you ahead on the productivity aspect.


Ryan Seratt 24:43

Absolutely. So one of my favorite things is the Natural Language Processing. Being a trainer, if I have to train people on how to do something, and when they're doing a search, for example, then that's hard. It takes something that's very complicated, probably, and makes it more complicated, because you have to use, oh, you have to use a quote, plus, and use a plus or minus when you're doing the search. And people have never taken on to those Boolean searches and mastered them. Very few people have. So when we talk about Natural Language Processing, and we were talking about it earlier, I just type in what I want, it gives me a result, then I can say add this, or subtract this, or add this bullet point, expand this out, make it an eighth grade reading level, make it a twelfth grade reading level, remove any reference to this, because I don't do that. And so you can really kind of refine that with the Natural Language Processing.

The other thing that I think that exactly, what is AI? AI is not a singular element. AI is really made up of a bunch of small, you can call them programs, people call them robots, or applications, but really when you're talking about a tool, there's going to be many, many AIs that are working together, that are doing the language, you know, writing a poem will be done by one AI, creating an image will be done by a different AI, and they'll be all packaged together, and but they're all singular, and they can work together, but it's just not one AI, even though we talk about it that way often.


Alyssa Dennis 26:27

Yeah, exactly. I mean, it's this way, AI is not an, it's not a singular, it's an umbrella. And I think that's really, really important for us, because even when you're talking about like generative AI, generative AI is utilizing LLMs, the Large Language Models, that's part of Machine Learning, you know, there's so many different interconnected components. So that's another thing is if you're having that conversation with somebody, and they're just saying AI, AI, AI, well, part of AI, because, you know, it's not necessarily that blanket statement anymore.


Ryan Seratt 26:59

Absolutely. So let's talk a little bit more about going from what AI is to how people are actually using some of these things that we've got here on the screen.


Alyssa Dennis 27:12

Yeah, exactly. So we do, right now, today, there are really prevalent AI use cases in healthcare, and a lot of times, I don't think we even recognize it as AI, simply because it's how it's always been, you know, it's baked into our EHRs, you know, like the medical imaging analysis, that's a huge thing within radiology right now, where they're able to utilize that assistance to be able to identify tumors and tumor markings, you know, the clinical decision support, determining if there's going to be adverse reactions between different drug interactions. That's AI. And so it's interesting, because we haven't really put that label on it, except for probably the last couple of years. But some of those components have been around for quite a while.


Ryan Seratt 27:58

Yeah, we've been using AI for forever. For example, Netflix uses AI to suggest videos to you. That's something that's been happening for a long time. But really, it was the advent of the Large Language Models two years ago, that really AI has kind of come to the forefront, and just under that blanket umbrella term that you were talking about, Alyssa. So on here, you know, I think that definitely all these are different exciting areas. But I think that clinical decision support is something that's going to be a little bit new, is one of my favorite things about AI is that, you know, two years ago, we were in the information age, and we had access to more information than humans have ever had access to before. But the question was, what do we do with all that information? And we're getting reports, and we're getting a lot of details. And we didn't have a lot of time to go back in and really analyze those like we wanted to. AI is going to assist us in the future of moving from the information age to the AI age, where it's actually going to be able to summarize a lot of that for us, it's going to be able to help us, and it's going to be able to do a lot of that very quickly. What are your thoughts on that idea?


Alyssa Dennis 29:21

I mean, you're very true. And the fact that it's going to be done quickly is a huge thing. Because again, if you think about it, we're all being asked to do more with less, do more with less, including time, it's not necessarily resources. So the fact that you're going to be able to take huge swaths of data, kind of look at what I said earlier, and actually be able to digest them and make educated decisions based on that vast quantity of data very quickly, not only have you increased your productivity, you've again probably increased your efficiency simply because you're not going to have to rework if you find additional data you hadn't considered the first time around. So again, it's, you know, I know we just keep talking about this again and again, but it really comes down to that AI is a tool. It needs to be a tool in your toolbox, and it needs to be something that you do pull out routinely, you utilize routinely, but you use intelligently. And I think that's where I really, really want to stress is don't be afraid of it, but you've got to make sure that you understand how you're using it. And if you're not sure how you're using it, then you've got to ask some more questions, because again, that's the driving factor.


Ryan Seratt 30:31

Yeah, I think the big fear is will AI replace me? And I wish I could attribute this to someone. I'm not sure who said it, but AI is not going to take your job, but someone that's using AI will be doing your job.


Alyssa Dennis 30:48

I told you that.


Ryan Seratt 30:49

That's a very true statement. Okay.


Alyssa Dennis 30:51

But yeah, it is very true. And I should give credit where credit's due. I was a part of a webinar, and a lady said that. It just resonated with me, because yeah, AI won't take your job, but Bob, who's using AI to do the job, is probably going to take your job.


Ryan Seratt 31:06

So the next two years- three years will be extremely exciting as AI becomes more and more to the forefront, but we're also using it today. As we've been talking about, we've been using AI for quite a long time, and really, that is even sped up in a lot of tools, services, and which are providing new capabilities. And I think we're also going to show a couple ways that 314e has incorporated that into our tools to make people's lives easier and be more efficient as well.


Alyssa Dennis 31:42

Yeah. So one of the first pieces that we are going to show you from products within 314e is the vector search, the ability to search in more than just a linear fashion. And that's where you're able to... essentially, you're able to utilize this search to peek around corners. That's probably the best way to say it in a layman's term, where it's not just a case of show me blank, but it's going to be show me blank and everything that it applies to, and it vectors off to make those connections. And again, leading to higher productivity, because you're not having to spend as much time searching, you get your results quickly, and higher efficiency, because again, you can do it quickly and move on with your life.


Ryan Seratt 32:27

Yeah, absolutely. And let me show you how we're using vector search in the training tool Jeeves. So Jeeves is a Just-in-Time training tool that is really meant to connect end users with the different pieces of information and training information that's available in your system. People can pull it to themselves in their moment of need when they have a question. So Jeeves is actually integrated deeply into EMRs, but we're going to use the desktop version today for demonstration purposes. So when a user either has a question, maybe they forgot something from class, maybe there was an upgrade, something changed, maybe they just don't remember, they're able to use natural language searches to actually do a vector search. And in this case, what are the steps to make a new auto text, which is a Cerner question. We're going to open that up to all departments, because it is filtering. I'm assigned a role, so it gives me very specific information. And we do a search, and it's pulled up the how to create auto text in its first thing.

So the way the vector search is using vector search is better. It's the next generation of searches, because it's not just searching on keywords or metadata that might be assigned to each one of these objects. So it's pulled up three videos. It'll also pull up text documents, etc. But also notice that auto text is a Cerner feature, but it's also pulled up create smart phrases. That's an epic feature. And the nice thing about the vector searches with the rag, I'm not going to go into the technical specifications. Welcome to get a virtual cup of coffee, and we can talk about it more if you're interested. But what it's doing is it's searching for the words that I typed in, but it's also searching for the intent. And the intent is auto text and smart phrases are the same thing, just from two different companies. So it's pulling up the other, even though I didn't type that in. And I think this is a great example of the vector search. And Alyssa, how about the tools you're working with? What are they doing as far as search goes?


Alyssa Dennis 34:45

Yeah, I have search in both of the products. But the one I really want to showcase today is the Muspell Archive. So this is our FHIR-native healthcare archive platform. And so this is where you're able to take if you sunset a source system, you know, if you've moved from Cerner to Epic, and you need somewhere to put Cerner information, because here's my HIM hat, you got to hang on to that stuff, you got to have it accessible. So what do you do with it instead of leaving it where it natively is sitting? Well, that's when you can move it into this archive platform. And you can have all of this information at your fingertips. And it's really helpful for your care teams, because then within that new EMR that you're working out, they can click a button, boom, they have all the archive information, not just the pieces that you chose to convert. So it just it really helps with that continuity of care aspect. And it also helps from your HIM teams being able to release information in one succinct location. So that being said, as you can imagine, some of the patients that you have that you've archived, maybe they've been with your organization for 30 plus years, that's not uncommon. Well, it can be extremely overwhelming if you're trying to find certain pieces of information within the 1000s and 1000s and 1000s of documents that might exist from that patient's history of care. So that's where we're able to come in here. And very quickly, now, before I get too far, this is demo data, this is fake data, I'm not showing anybody's PHI. So take a deep breath. But what I can do is I can come in here. And I thought I could master talking and typing. But I can go ahead and I can just put in something in natural language. Because again, I don't have to do that whole thinking through how am I going to program my search in a certain way to get the best results. And I'm able to tell it show me any consents from a certain data service. Boom, first consent is from that data service, I can go ahead and pull it up, I can review it to my heart's content. And I can move on. Now, if you'll notice, my search results are more than just that single document, though, because that's where it's using the vector capacity and looking at well, okay, she said consent. So what other consents are there? And in this consultation report, did they talk about informed consent? Because that's where it's able to pull and is trying to be helpful. Again, this is an assistant, this is a tool that's trying to help you do your job. And so it's going to say, hey, what about this? What about this? What about this? So again, coming back to what we've been talking about, we still have to have that human capacity and that human acknowledgement that I understand all of these results are not what I personally need. So I'm still engaged in what the system's providing me. But it made my job a lot easier by being able to come in here, tell it what I want and see it right away, versus if I needed to come out here. And this is definitely very doable. But if I needed to come in here, go to my documents, go ahead and sort, and I can pull up the consent this way as well. So again, looking at that productivity, looking at that efficiency.


Ryan Seratt 37:45

And to sum it up, I think that the vector search allows you to connect with the right information. So like as we were saying earlier, there's so much information, how do you connect to the right ones? And that's been a frustration in the past. Our older tools just weren't able to connect us and a lot of people got frustrated with that. So the Large Language Model, the vector search, all of that allows us to find the right information very easily.


Alyssa Dennis 38:03

Exactly. So now switching gears slightly, let's talk about content extraction, because that's another key component. And that's where a lot of the, quote unquote, magic happens when it comes to AI. Because based on the content that it's able to extract out of the document or the data, you're going to be able to do a lot of manipulation to automate some really routine tasks. So Ryan, what kind of content extraction do you have in Jeeves?


Ryan Seratt 38:32

Let me show you a couple of examples here. We do a lot of content extraction to make things easier and provide information in different ways. So the first one I'll talk about is our chatbot. So users can go into our chatbot, type in a question, hit start conversation. And what it'll do is it'll search the entire catalog and have a dialogue with it and pull out that information. Let me show you an example from using a tip sheet. So I'm going to take a recorded video. So this was actually, we'll grab this OpTime blood products example. And we're going to extract all of the steps that are in this video. So the AI went through this five minute video that we recorded, and it's pulling out each of the steps that someone needs to do. Click on the back arrow, fill out the path. Normally trainers, when they do this, it takes them about 45 minutes to do it. The AI already has all of those steps. I'm just going to move those over to the work area over here on the right. And now what I can do is I can go through and when the video gets to the right spot, I'll just hit screen capture and then add a screenshot to this. So I'm able to take a five minute video and I'm able to create the text, add screenshots and do it in 10 minutes instead of 45 when we do it manually. And trainers and clinical informaticists create a lot of these. So that saves me time and increases my productivity. In addition to that, the videos that are created, we have different AI agents that are going through and doing a lot of things to this video. It's going through and it's actually putting an AI narration that's smoother than the human narration onto this. So it's replacing the voice with an AI narrator that's near human quality, if not human quality. That's changed a lot in the last two years. It's also putting a transcript on this video. It's putting closed captions on this video so that I don't need to do any of that, increasing my productivity.


Alyssa Dennis 40:48

Yeah, I know personally, I absolutely adore the video to tip sheet aspect of Jeeves because in my past lives and even in this life, to be honest with you, I have gone ahead and I've had to make tip sheets and it's an exhausting process. So being able to do it in one full swoop is just, oh, it's so slick. I get excited every time I use it.


Ryan Seratt 41:11

That's people's favorite feature.


Alyssa Dennis 41:15

Now what I'm showing here is content extraction in Dexit. So Dexit is our Document Management System. It has a full fax server capability, cloud-based fax server, but we have a lot of AI that actually can be laid over top of incumbent products. So some of the extraction and some of the assignments of information and the metadata I'm going to show you is actually something we can apply over top of an incumbent Document Management System. So again, even if you're thinking to yourself, well, this is great, but we're never going to change. That's okay. It's more of a case that you can add on instead of take away. So I'm not going to go through the whole document upload because again, when you do the upload, it takes a couple of seconds while the AI reads through your information. So you can see I did this a little earlier this morning, but this is a PDF document, nine page document that I uploaded into the system. And the AI within Dexit is able to go through and read every single page of that document and identify which document types based on document types I programmed into the system. And when I say program, I mean that very loosely as I typed in progress note and hit add. So I want to be careful with that, but really it's able to go through and identify all of the different pages that are affiliated with each document type. And it's able to also extract patient information. Now, again, I want to pause and say, this is not real PHI. So this is not any type of a breach. This is all fake demo data. But one thing I really want to show is, you know, being able to identify document types and being able to identify patients, that's not new. That's been around for quite a while, but usually that functionality relies strictly on OCR. And OCR is essentially barcode technology. That's like when you go to the grocery store and you scan the barcode and it automatically knows how much that item should be. That's OCR. So if you notice though, this document does not have any barcodes. It doesn't have a patient barcode. It doesn't have a document type barcode, but it was still able to identify that this should be a discharge summary. This is tiny. Let me zoom this in a little bit. There should be three pages, one, two, three, and it was able to identify the patient name and match it with the EMPI feed that we have coming into Dexit. Now, one thing that you'll notice that there was no encounter assigned. This is actually something that is on our strategic roadmap where our developers are working on it probably right this second, where they're going to be able to come in and extract information such as the account number, dates of service, and be able to match that again based on the ADT data that we have coming in and be able to extract that encounter information right away. So really what we're envisioning is that Dexit's indexing process, you know, that manual process that you would have to have team members come in and make all of the changes. That's really going to become an automated process where you just simply need somebody to validate and come through and say, yep, yep, that's correct. Move on. And it's really handy when you're dealing with documents that originated from outside your organization. Because again, the barcodes are spectacular if your system recognizes that specific barcode. What my barcode for progress notes is probably isn't the healthcare organization across the street or across town. They're not using that same barcode. So again, it's something where it's really focusing on being a tool to increase your team's efficiency and productivity.

All right. So now the next one is routing rules. So we've extracted that content. What do we do with it? How do we make sure we're able to extract that content? it gets to the right people and how do we do it in a way that we can really again focus on the efficiency. So Ryan do you have any good examples of routing rules in Jeeves?


Ryan Seratt 45:10

We do and for time's sake we're going we'll keep it fairly brief. I think yours are much more advanced than mine are today but so when we're actually when people do searches is notice that we're also filtering by roles and to help make sure that they're getting the information they need. We're also using it here on their on our home screen where we're not only looking at past searches but we're actually suggesting people to take a look at other trainings that are available and based on what they're looking at we're specifically serving them other content that they might not have seen. And on our roadmap we're actually looking at taking a lot of the physician and nursing data on the reports on how people are using their EMRs and then the system will analyze that and push out training based on how they're scoring on the reports from the EMR system. And that's very exciting and you know when we talk about it it's like well haven't you always been able to do that? I have always not been able to do that because I've never had time to do it. My teams have never had time to do it. So what AI is doing is it's taking something that we could have done but it's doing all the analytics and matching it with the content for us and then we can supervise what's being pushed out to people. So that's a very exciting and that's on the roadmap for next quarter.


Alyssa Dennis 46:40

Yeah and then within Dexit again showing the routing capability this is something again that's very very exciting because usually when it comes to paper there's a lot of the tasks that have to happen again and again and again and again. I mean something so simple as you have a physical fax machine in your office and you have to have somebody who gets up from their desk walks over there and goes hey this fax should go over here or here and you know just the simple routing of that information. And that's very very simplistic compared to some of the stuff that we can do within Dexit today. But one of the best examples I can show you is just our routing rules and our routing logic are so strong that it literally becomes a case of if you can dream it, it can do it. So the hardest part is figuring out what you would like to automate. So for instance what I have right here on the screen is just a very very simple rule that I put together and I say that specifically because again I'm not a programmer. So it's something that we've tried to make very very obtainable to the end user because there's nothing worse than having a really cool tool that you don't know how to use. That's very intimidating. But really what this tool is is it's saying that every fax if you identify that the information coming into Dexit is a fax, extract the document type. Again using that logic reading through and identifying if this should be a progress note a consultation etc etc. Once you've identified what the document type should be, go ahead and put it into a certain queue. And so that's where you're able to build all of these queues or essentially work areas that if something is a consultation, send it to this queue because this group of people needs it. Say if there's a discharge summary, send it over to this queue because a different group of people needs it. You know if it's a request for Release of Information, send it to another queue that your ROI team is managing. And at the same time let Dexit send an alert so you don't have to be monitoring it on a you know an hourly basis. Dexit will go ahead and ping you in Teams or Slack or send you an email and let you know hey there's work to be done, come on back over, let's get started. So again looking at that productivity and that efficiency because there's only so many hours in the day and lord knows we have a lot to get done. So I know that we went through some of these products really quick. If you are interested in learning more, by all means like Ryan already offered, you know let's have a virtual cup of coffee, pull a chair and let's have a conversation because you know if you hadn't noticed both Ryan and I are pretty excited about what our products could do so we'd love to talk your ear off. But otherwise really the big pieces that we want to drive home from today is AI is pretty darn great right now and it's only getting better as time progresses. But again remember it is a tool to be used. It's not something that can be used standalone. You still have to have that human component. You have to continue taking responsibility of what AI is producing because that's never going to go away. Right Ryan?


Ryan Seratt 49:44

No I don't think it is even in the future so maybe someday but not someday soon.


Alyssa Dennis 49:52

Exactly. I mean that maybe then we are talking Skynet in 200 years. But for right now what we have to worry about today it's a good tool that we should all be utilizing but you got to use it responsibly. Alright.


Ryan Seratt 50:07

Alright, should we move into the Q&A? If you've got any questions from the presentation, anything that you would like more information on, like us to go more in depth on, please type them into the Q&A and we'll be sure to answer those.


Alyssa Dennis 50:22

Alright, and I do say we have one question already so let me read it out to everybody and it says I'm concerned about the reliance on AI and not enough genuine application from the physician. What measures are in place to ensure there is a good balance of AI and genuine human observation and input? I'm not convinced that physicians won't get lazy and rely totally on AI. It's a good question. Ryan you want to take a shot at it?


Ryan Seratt 50:53

Yeah I think and this goes along with the concept of how are you going to use AI in your organization. AI you're going to need to develop rules of engagement basically. This is how we use AI and obviously there's some training that goes along with that so I've been working with some people on developing kind of those rules of the road that here's our processes. It has to be built in that AI might actually be able to take a look at a scan but then who is reviewing the scan and what do you expect them to do. Is there a checklist that goes along with that so you can make sure that the supervision that is really needed is along the way. I completely agree that some people are just going to say - oh well the AI said that there was - that there was a anomaly right so you don't want to tell someone oh you got cancer because the AI said something. That would be that's what we were trying to avoid. So what's - what rules and processes do you have in place to make sure that the human supervision is there as well?


Alyssa Dennis 52:04

Yep exactly. It's really a cultural aspect within your organization. You know quality assurance programs are wonderful for reviewing but at the same time you do have to have those expectations of how you interact with the AI. You know maybe not the best analogy but in my mind I think of well it's like PTO or time off. You have it available but there's expectations of you might have to ask for that time off within a certain time frame of the day that you want. You know it's there but there's still parameters that are being put in place around it from an organizational policy standpoint and I really think it's no different than that when it comes to AI. You have to set the expectations. You have to set a measurable way of ensuring that those expectations are adhered to and unfortunately you have to come up with consequences when you have those individuals who choose not to adhere.


Ryan Seratt 52:55

Absolutely I agree and I think that you know this is a question that I see quite often is the public AI domains and the - the posting of those. People are in my opinion people are going to use AI because it's going to make their job easier and I've been using Grammarly for example for years because it does help me with my writing and any mistakes. A lot of other people use it's very popular program. That is AI and I'm taking my intellectual property and I'm putting it on a tool that's out on the web and what kind of controls are you going to put in place. So we like we said to really be effective you need to take your Business Intelligence and put it into an AI that's behind your firewalls where people will go out to the public once in my opinion and so and when you do that I think it always comes down to either you can build it yourself or you can work with a partner to bring those tools in but those tools are constantly being upgraded managed and so I think that's something you need to have in your strategy as well and really AI is coming to the forefront. We're seeing co-pilot already rolled out for all the Microsoft tools and Google tools and that's we're going to just see more and more of that in the future. So the concern is yes you have a lot of concerns about putting your business data out on the public ones. I would not do that. I would have policies in place to avoid that as much as possible. Yeah but bring those tools inside where they're behind your firewalls and you can protect that data and people can really get the most use of them that way as well.


Alyssa Dennis 54:43

Yeah exactly and I would be remiss if I didn't mention again HIM first, the same for your PHI. Don't put your business data out there on the world wide web and for all that is important do not put your PHI out there either. That's you know bring as Ryan said bring the tools inside use them and really analyze them study them ensure that you have good benchmarks around them that you've got a good control around them from policies procedure aspects from a monitoring aspect but really make sure that you have that those end spots in place to ensure that nothing leaks out.


Ryan Seratt 55:18

So we also have a question about intellectual property. Who do you believe should own the IP of a generative AI output? Alyssa do you want to take that one or you want me to take it?


Alyssa Dennis 55:36

Nope I want you to take that one because that's that's you know that is such a - an important question nowadays and I think it's one of those where it nobody has a right answer everybody just has the answer and it's I again I'm sorry I told you to take the answer you know - you'd answer it and then I'm doing it anyway. But I mean I really think again it's going to have to be you know from if we're talking from an organization standpoint it's got to be from that organization level from you know what is your strategic goals what is guiding your principles within your organization and write that into your policies because how are you going to handle that that is absolutely going to be a situation that occurs if not today probably by tomorrow.


Ryan Seratt 56:19

Absolutely and for those who don't know I did a webinar on the efficacies of using AI and copyright issues highly suggest if you're interested in taking a look at it and but for intellectual property that's generated with AI when to boil it down very quickly is that you as the content creator are the own that intellectual property and that's what there's even guarantees that Google and OpenAI have put into place to to make sure that that's protected. Now with that being said most of the controversy is coming from on how is that AI trained so if that AI is trained on copyrighted materials then what it's generating is based on those and that's where a lot of the controversy is coming from. The AI companies are going back and they're making sure to kind of clean up those databases to remove intellectual property that they didn't own the rights to when they trained it so that's going on behind the scenes but the everyone as of right now is very firm that if you do create something that it's yours if the AI companies used that - that information to make something they might be liable for it but you as the creator are not.


Alyssa Dennis 57:46

Alright.


Ryan Seratt 57:47

And that's why we talked about it for 45 minutes it's really complicated.


Alyssa Dennis 57:50

Exactly so wonderful questions again don't feel like this is the only time you can ask your questions as you'll see up here you know go ahead and reach out to us you know the contact information was from the invitation to this presentation if you have other questions you know you're walking your dog you're in the shower in the middle of the night you go - oh shoot I should have asked - you know, feel free drop us a line we'd love to continue the conversation like I said Ryan and I are pretty passionate about it so we'd love to keep visiting with you.


Ryan Seratt 58:16

Yeah I think we have time for one last one what are some common misconceptions about AI that you frequently encounter?


Alyssa Dennis 58:23

Mine for sure is the whole it's magic it's going to rule the world it's going to take over everything and you know I think we've kind of laid that to rest over the last hour as we've been discussing so I think the biggest thing that I could urge people to take away is don't be intimidated by AI it's here it is going to be a part of our lives for the foreseeable future and so the important thing is don't be intimidated by it but educate yourself and I know that in and of itself is intimidating but sometimes it's something so simple as you know signing up for a daily blog or a daily newsletter and just start reading about it and the more you read about it the less intimidating it will be it'll let be less buzzwordy and it just becomes more of your daily vernacular and I think that's really really important because we need to get comfortable with using AI and the only way to do that is to understand what AI is and how you can apply it to your daily life.


Ryan Seratt 59:19

Yeah I agree just get in there and use it you're not going to break it don't put any company information on it but you know use the public ones and ask it public questions general knowledge you know even ask it you know what's the weather today and you know and other information other just general questions that you have.


Alyssa Dennis 59:40

So yeah sorry I was going to say one thing I've started doing myself is I actually use it as a search engine instead of Google because sometimes again it's really easy to be like my leg hurts am I dying and then it's interesting to get back you get that again from more of that vector search capability you get more of the stuff that is just pertinent. instead of having to go through page and page and page of the Google searches. Or like me, if it's not on the first page, it don't exist.


Ryan Seratt 1:00:06

Absolutely. And in summary, if you'd like to learn more, we would be glad to set up a session just to have a conversation. Or if you want more information on the tools that we're offering and how we're using AI, we'd be glad to have that conversation as well. Thank you so much for your time today. We know it's valuable. And we appreciate you joining us.


Alyssa Dennis 1:00:30

Yes, thank you, everyone. Take care.