Home Care U - What We Learned About AI in Home Care in 2024 and What’s Coming Next

Episode Date: December 23, 2024

In this episode, we dive into the AI breakthroughs that shaped home care in 2024 and explore what lies ahead for 2025. Dan, Spencer, and Victoria from Careswitch, share their perspectives on how gener...ative AI has impacted the industry, from technological advancements to real-world results seen by home care owners using tools like Careswitch. Tune in to learn about emerging trends, practical tips for leveraging AI effectively, and the exciting innovations on the horizon for 2025.Enjoying the show? Send me a text and let me know!Learn more about Careswitch at: careswitch.comConnect with the host on LinkedIn: Miriam Allred This episode was produced by parkerkane.co

Transcript
Discussion (0)
Starting point is 00:00:00 Welcome to Home Care U. It's great to be back with everybody. This is the final episode of 2024. It's been an incredible year. Thank you all for being here. Thank you all for listening and learning alongside of us this year. It's been a blast. To close out the year, we're going to do something a little bit different. I've decided that we should wrap up with an episode by the team at CareSwitch. So today I'm joined by an episode by the team at CareSwitch. So today I'm joined by three of my colleagues here at CareSwitch. As you all know, a big focus for us in 2024 has been bringing the latest and greatest AI technology to home care providers. So we've learned a lot, we've built a lot, we've had hundreds of conversations,
Starting point is 00:00:38 and we've landed on some pretty remarkable results. And we want to share that with all of you on this show today. So today we're going to download on essentially what we've learned about AI in home care and in general in 2024. And then we're going to wrap up the episode by talking about some things that we're looking forward to going into 2025. This technology is evolving extremely quickly and we're learning about it every single day. And so there's things on the horizon that we're looking forward to. So we thought this would be kind of a fun, fruitful, interesting, more casual conversation today. So I'm really looking forward to it. Today, I'm joined by Dan Ogren, Victoria Brown, and Spencer Strombone. I didn't ask you how to
Starting point is 00:01:18 pronounce your name before we jumped on Spencer, so I didn't butcher that. But let's start with introductions. I want each of you, because this is kind of a new platform for you all to be on, I want to give you a couple of minutes to introduce yourself. So Dan, why don't we start with you if you want to just share name, role, a little bit about your experience in home care, maybe personally and professionally, and then how long you've been tinkering with AI, and then we'll go Victoria and Spencer. Yeah, sounds good. Dan Ogren, head of design here at CareSwitch. I'll try to keep it short, but my background goes a little bit further back with Spencer. I started my career at IBM in 2014. That's where I met Spence. And we decided to start a company together called Repora. We founded that company. And it's a home care agency management system,
Starting point is 00:02:02 more of a traditional home care agency management system, you know, something we're not building today. But we sold into small and medium sized businesses. And we did that for about four years. After my time at Repora, I worked at a company called Frog. I worked there for about two years, worked on a lot of different enterprise projects, mostly oil and gas, agriculture and financial services. Did a lot of design and building there. And then in 2020, I met Ilya, and he had known about my time at Repora,
Starting point is 00:02:33 shared a lot of the same, you know, vision and ideas around home care. We both shared a lot of the same perspective on the opportunity and, you know, how home care can really be elevated and brought up in healthcare as a whole. And so we bonded over that. I decided to join CareSwitch in 2020. So the year I met Ilya and I've been there ever since. So four years. I think the last thing was to talk about AI. Yeah. So I've been tinkering with AI for about two years,
Starting point is 00:03:03 pretty much since it came out, when it started to really make headlines. And I've been tinkering with AI for about two years, pretty much since it came out, you know, when it started to really make headlines. And I've just spent a lot of time using it day to day. It's a daily driver. I definitely use it for a lot of my work. A lot of my work is around design and, you know, building the product for you all and working with Spence and the engineering team on that. So it definitely helps me, you know, organize my thoughts, keep things moving. And it's been really exciting and fun to learn and be on the journey as we all know. I don't, we don't know where it's going, but it's going to be, it's going to be fun.
Starting point is 00:03:32 So that's me. Thank you. Awesome. Thanks, Dan. Shout out to all the OG Repora listeners listening to this. I bump into them. They're out there. I know they exist.
Starting point is 00:03:40 So shout out to all of you. You're hearing from Dan and Spencer, some of the OGs behind the Repora platform. So cool. Thanks for that, Dan. Victoria, you want to jump in and then Spencer? Great. Yeah. Hi, everyone. Victoria Brown, Enterprise Account Executive here on the CareSwitch team. I've been around the, I'd say, caregiving industry, you know, essentially from leaving college. So my very first professional role after graduation, I ran a volunteer program for a hospice. So I recruited, trained, and managed all the volunteers, spent a great deal of time connecting them with caregivers and, you know, kind of a team approach to supporting families going through, you know, the hospice program. And, you know, obviously a really challenging
Starting point is 00:04:20 time for them. So learned, you know, just how challenging the role is of caring for people emotionally, physically, and how incredible those people are. So I really loved, you know, the industry and kind of providing that care, keeping people at home. So then, of course, transitioning into software, my first role was for a company called In The Know Caregiver Training. I assume many of you listening are familiar there with the great Linda Leakley. So we then merged with Home Care Pulse, which is now Activated Insight. So I was on that team. Of course, I met Miriam at Home Care Pulse. And all told, I probably spent, you know, maybe 10,000 hours at this point, you know, on one-on-one calls with home care owners,
Starting point is 00:05:01 administrators, schedulers, billers, learning, you know, kind of every facet of what you all go through. So really have a passion for it and very, very excited to be here at CareSwitch. I'm almost at a year now. So I joined in January and it's been an incredible year. The advancements that Dan, Spencer and team have made this year are really phenomenal. AI as a whole, the improvements and just overall journey that we've taken with the technology is incredible. So for me personally, same as Dan, probably two years now, as soon as I really learned about OpenAI, immediately curious and immediately started kind of getting acquainted with what was possible. At that time, my daughter was in middle school. So
Starting point is 00:05:45 she's now a sophomore in high school. But obvious kind of first steps were homework help. If anyone's gone, you know, back through eighth grade math, after you've forgotten it all, chat TBT really bailed me out some some study sessions. So that was my kind of foray into AI. And now I talk to a robot all day. So it's really incredible. Awesome. Thanks. Great. That's great. Spencer, last but not least. Hello, everybody. I'm Spencer Strombodny or Strombone. There we go. There we go. I've been involved in computers pretty much my entire life. Ever since I've been a little kid,
Starting point is 00:06:21 I've been tinkering with them. Professionally, I've been working in software now for about 15 years, developing software for home care for about nine years now with Dan, as he said earlier, for CareSwitch for four years now. And with AI specifically, generative AI, about two years. First became interested in AI probably back in university. I mean, it is the holy grail of computing. It's something that everyone that works with computers wants to use one day. I would be lying if I were to say that I was a specialist or it was a special interest of mine, but super cool tech. Started paying a little bit closer attention in 2018, 2019. Somebody at a dinner party showed me GPT-2.
Starting point is 00:07:05 That's before GPT-3 and before 4 and whatever we're using now. And at the time, I thought, wow, this is a really cool party trick. We're getting really close to AI. Maybe you want to start paying attention soon. And then 2020 comes around, GPT-3 is released, and now I'm the one going around at dinner parties showing everybody AI and being the evangelist. And yeah, it's been great.
Starting point is 00:07:27 Awesome, awesome. Well, I'm glad we actually ended with that question of like kind of base understanding and usage of AI today because here we are sitting in the expert seats, educating and teaching. But I think like Dan said, it's probably a primary driver for all of us at this point in time,
Starting point is 00:07:41 but we've spent a ton of time using it the last two years. So that's, I think, why we feel empowered to have this conversation and lead out on some of the conversations around AI. So let's start kind of generic and zoom out a little bit on just the advancements and technology that we've seen over the last year. Dan, do you want to start us out with just some of the trends and breakthroughs that you've observed with generative AI in general over the past year? Like what stood out to you? What have been some of the trends, things that we've kept an eye on? Yeah, absolutely. So being a content creator myself with a background in design, I actually, you know, studied graphic design in school. And a lot of the, you know, the initial versions of
Starting point is 00:08:20 the AI that we're using today, whether it be, you know, chat GPT, Dolly, mid journey, things like that, you know, the big breakthroughs are really in content creation, right? So if you're a copywriter, a designer, a product person, and you need to generate assets, visual assets, or copy for material that you're working on, whether it be marketing or sales, or, or even internal assets, you know, AI has really stepped up and been a big support. And I know I'm preaching to the choir here internally that we're all using this day-to-day to generate assets and content.
Starting point is 00:08:51 So I think the big one is really just a lot of the ability to generate assets quickly and kind of iterate on them. And I think the other piece to that is the ability to generate the first draft. AI is not going to be perfect, but people are able to kind of tweak and continue to work with it to get it into the output to be in some state that they're satisfied with.
Starting point is 00:09:13 So I think the productivity on the creative side, the design side, you know, the copywriting side, all of that has been really boosted by AI. And I know there's a lot of fear around that of being replaced. But I would really suggest people that feel that way to think about it as more of a boost to your productivity, less of a replacement of you as a person doing that job. So that was a really big, exciting one. I think that's probably the most obvious. You see a lot of memes, you see a lot of stuff on the internet generated by AI, very common. I think the next thing we're starting to see really is this, this idea of no code or low code AI. So people that are non-technical, the ability to make an app, make a website, just using plain language and text
Starting point is 00:09:58 to talk into the computer and it generate and build you, you know, real software, which is super exciting because, you know, not everyone, you know, real software, which is super exciting, because, you know, not everyone, you know, went to school like Spencer did to learn, you know, computer science and things like that. But, you know, for people like myself, when I'm trying to show Spencer what I might be asking him to do, or talking with an engineer, it gives me the ability to, you know, really spin things up and proof of concept and prototype things out without writing a single line of code. So that's really cool. Super exciting. And I think we're very early kind of on that track. More specific
Starting point is 00:10:31 to healthcare, you're just seeing a lot of companies, you know, come out with different tools that are really good at, you know, reading and understanding documentation. We know that, you know, the medical and healthcare space can be very intense in terms of documentation and technical information. So, you know, people are starting to really take AI into the healthcare space and use it to speed things up and be more accurate in a lot of the documentation and things that are being collected. So I wanted to tie in a point specific to healthcare because we are in the home care space that, you know, it is making its way into healthcare. And it's really exciting to see that. And then I think the last big trend that I was thinking about was just this idea that people are actually coming around to AI. They're warming
Starting point is 00:11:16 up to it. You know, it starts out as, you know, a party trick, as Spencer called out, at a dinner party. And now businesses that are, you know, serious about productivity gains and efficiency gains and accuracy gains are really, you know, looking at this and taking a serious look at it to bring it and considering whether or not they should bring AI into their business and use it on a day-to-day basis. So, you know, a lot of the people, I'm sure Victoria can attest to this, a lot of the people that we're talking to in prospects and even our customers before they were our customers, they were using AI on the side with the other piece of software they were using to run their business. you know, day to day making grocery lists and doing math for their kids in eighth grade. But we're starting to see that. And that's really exciting because I think, you know, that's the direction this is all headed is AI is going to really infiltrate business and more
Starting point is 00:12:15 specifically healthcare and then, you know, now home care. Awesome. Yeah, that was a great place for us to start kind of laying the foundation of some of the things that we've seen. That was awesome. And we'll talk a little bit more about like adoption, especially adoption and home care. Like what are people saying? What are people thinking? Where is that headed? Victoria will share more on that. Spencer, I want you to take us maybe one layer deeper than what, where Dan has kind of started and talk about specific advancements. He was talking about like assets to code, to building websites. What are some of the, like the technology components at play here that people are using that are starting to create kind of those efficiency gains? What would you kind of
Starting point is 00:12:50 call out in that regard? Well, a number of things. We've been dealing with text for quite a while now where we have LLMs that can generate out anything that you'd like. But at the same time, we're starting to experiment around with different modalities. Not everyone deals with text day to day. So things like video, text to video, video generation, picture generation, lots of different modalities to explore. And we're really just at the beginning of this journey. So what people have been experiencing so far is really just the tip of the iceberg. I'm mostly excited for what comes next. But in terms of the advancements this year specifically, generally we've just seen increasing capabilities of the models themselves.
Starting point is 00:13:34 So at the beginning of this, I mentioned that I first started using GPT-2 or first saw it. Then GPT-3 came out, GPT-4, and a lot of people might be asking, so what? What's the difference between these? But I look back on some of the messages that were sent with GPT-2, and they're laughable at this point. I pulled up some conversational history I had with GPT-3 earlier today, and it's embarrassing. It's really difficult to look at these tools and the responses that they were returning at the time and thinking I was going around at dinner parties, you know, telling people that this is it. The general capability
Starting point is 00:14:08 of the models this year have improved in a way that I don't think many people have been paying attention to. It's not exactly like you can really put a thumb on the pulse of what's happening and really kind of describe how these things are getting better other than just saying, I used it today and it provided me an answer that it wouldn't have been able to more than a year ago. So just the underlying model capabilities themselves are rapidly increasing at a rate that I've personally never seen in any other piece of technology. Can you expand on that a little bit? Because people may be wondering, is it just general, the volume of usage is increasing, and so therefore they're getting better just by pure usage?
Starting point is 00:14:49 Or how are the models actually getting better? In fact, the models don't get better at all while you use them. That's one of the big complaints. Doesn't remember, doesn't learn from its past mistakes. It's always a fresh new conversation every time you open up that chatbot and you begin talking with it. But where it does really improve is in the training. And this is where companies like OpenAI,
Starting point is 00:15:10 Meta with their Lama models, Google themselves are starting to catch up here, and even Axe, formerly known as Twitter, have their own model that they've been training. And it's more of a question of the amount of data that's going into the models at training time and how much compute resources are you devoting to it. In the very beginning, GPT-2 was trained on just a handful of computers.
Starting point is 00:15:32 You know, consumer technology that you or I could, with a large enough credit card, go out and buy off of Amazon right now and start training our own GPT-2 model. GPT-3 required a computer that was significantly larger than that. We're talking about orders around 10,000 computers or so combined. Now, of course, these aren't consumer grade computers anymore. We're talking about clusters of GPUs and a bunch of nerdy stuff. But just for your own kind of mental model, just imagine 10,000 computers stitched together with wires all working together to train this new AI. Then we're up at GPT-4. GPT-4 took 24,000 computers stitched together with wires, all working together to train this new AI. Then we're up at GPT-4. GPT-4 took 24,000 of these computers, and we're now approaching even larger
Starting point is 00:16:13 scales. We've got a new supercluster that was built in Memphis, Tennessee earlier this year. It's owned by XAI, formerly known as Twitter. And they're currently in the process of training the newest state-of-the-art model right now, which will be four times as large as the current by XAI, formerly known as Twitter. And they're currently in the process of training the newest state-of-the-art model right now, which will be four times as large as the current state-of-the-art models are, which would be GBT4, Google's Gemini, Meta's Llama, etc. And we're looking at 100,000 GPUs stitched together. So that is where this progress is being extracted from, is literally the amount of resources, funding, data that you put into these models going in.
Starting point is 00:16:48 You get a model that comes back out that's far larger than the previous models, but far more capable. So just to clarify, I think a common misunderstanding is that these models are being improved based off our usage. But if I understand you right, the training is totally separate. They're not using our usage to train the models. The training is a totally separate process. Is that accurate?
Starting point is 00:17:09 Absolutely. Yeah. So you can imagine somebody within these organizations spends a considerable amount of time going around the internet trying to find as much data as they can, whether it be Wikipedia articles, blogs from recipe websites from your grandmother, anything, any piece of text that they can get their hands on the more text the better and of course there's an element to making sure that the text is clean and that you're not completely putting in nonsense
Starting point is 00:17:35 into the AI and things like that but we're operating mostly off of philosophies of scale at this point, the more data the better and we also see a trend at the same time of AI companies that are creating similarly sized models. And some of those models are better than the other models of the same size. And that's mainly to do with the quality of the data
Starting point is 00:17:57 that they're actually turning it on. But when you are using that AI, it is the same model that it was the last chat conversation you had with it. It will be the exact same model the next time. And every time you open up that chatbot, it is literally reintroduced to you from scratch. It's given your name. It's given your profile.
Starting point is 00:18:20 It gains access to who you're working for, who your clients are, who your employees are, where's payroll this week, but that's on every single new conversation from scratch. We don't necessarily have this long term memory yet, as part of the AI systems. I think I think we're getting into the weeds here. But I actually think this is important. There's a lot of misconception around how this technology works. I think a lot of people think that their usage is a part of the training and it's learning based off what we're feeding it. But I just want to like kind of draw that distinction here that that is separate. But I also like letting you run on just like the computer concept, how many computers are strung together to train these
Starting point is 00:18:58 models. Like it's, it is pretty mind blowing, but I hope that like kind of paints the picture of what's at play behind the scenes here. Earlier, you were using the word like modality. The primary modality for us to interface with AI is text. You were talking about like text to video, text to audio, like these different modalities. Can you explain a little bit further what you mean? Like today, it's text heavy. Is that working well? And we're getting ahead of ourselves, but like kind of what's coming and how will what's coming be more beneficial than the text modality? I think text was just most convenient for these companies to go out, collect a bunch of data for and then train their AI models on such that they produce text. But that's not the only thing that
Starting point is 00:19:40 AI models can produce. It is essentially an algorithm that takes a bunch of information and figures out how to recreate it when you ask a question or you ask it to generate something. If it's been trained on text, it's going to generate text. It's been trained on YouTube videos, it's going to generate YouTube videos. It's trained on artist paintings, images, things like that. It's going to generate images and artist paintings. If you train it on driving data, it will learn how to drive, which is what Tesla is currently doing. So what you train the AI system on is generally what the AI is going to be able to generate. And currently, we're all about text. We use computers all day. We're sitting in front of a screen. Text is convenient. There's a lot of it, but maybe not the most efficient way
Starting point is 00:20:26 to interact with the computer. It's kind of how we've been doing it so far, but we're starting to move towards a direction where there's more than just text. You can speak to it. It can listen to you. It can show you pictures. It can generate how-to videos for you on the fly. You can't find it on YouTube. It certainly can do it for you. And we're moving in that direction very quickly. Yeah, super exciting. We'll talk a little bit more about like voice at the end, because I think that's the thing that a lot of us are excited about in the new year is just like the advancements with voice and how much more maybe natural that will be for a lot of us than text. I know I kind of cut you off there and took us in a couple of different directions.
Starting point is 00:21:01 Anything else that you wanted to call out on just advancements that we've seen this past year? I would say that I would forgive everyone for not being up to date with what's going on in AI, because every week something changes. Within the last 10 days, OpenAI has released a new product one after the other every day. They call it the 12 days of OpenAI. So it's really, there's a lot of information out there to pay attention to. But I would say if you were to pay attention to one thing and one thing only, it is what's going to happen with the newest model from X, formerly known as Twitter, Grok. There's been a hypothesis that has been debated throughout the year of whether or not these models continue doing better the bigger they get. And up until this point, it's really just been conjecture,
Starting point is 00:21:46 people arguing. There hasn't really been a new data point. No one's tried it yet. This is the model that's going to be substantially bigger than everybody else's model, and it will be the signal to us all of what's to come. If the model performs far, far better than the models that we have now,
Starting point is 00:22:04 then I can confidently say that we're on a pathway to AGI, artificial general intelligence, where the world is going to look far different than it does today in a few years. And if not, then we're probably going to have to temper expectations a little bit and try to figure out what can we do, what's the most we can get out of the models and the technology that we've got now? But we're in this like wait and see moment where maybe we're going to have artificial general intelligence here in the next two years. Maybe what we have now is pretty much as good as it's going to get, but it's slightly a waiting game right now. And I would say pay attention to what happens with Grok version three. I think you've got me on the edge of my seat.
Starting point is 00:22:44 I don't know about everybody else listening to this. I think we're all nerds here and really into this, but it's like, we are. Like, look how far we've come in the last two years. The next two years could be exponential, which is both terrifying and exhilarating. But I think that's what you're calling out here is like, we've learned so much,
Starting point is 00:23:00 but we're still on the edge of our seat. And what's coming in the next two years could quite literally change, you literally change the way that we approach our everyday lives. A really good question. Before we keep going here, a good question came in from Wade. Wade, great to have you with us. He was talking about in chat GPT, you can build your own GPTs.
Starting point is 00:23:16 And he was asking, what about when you create your own GPT? It seems like it's remembering, like you're kind of training or building your own GPT and it's remembering that. Can you explain how that process works when you build your own? It's remembering in the sense that you're telling it who you are and what you want it, how you want it to behave. But that part is static. You know, it's as you're having a conversation with that GPT, as you're getting back and forth with it. It's not going to learn from your interactions. It's not going to pick up on mistakes that it's made. It's not going to learn further information about you.
Starting point is 00:23:51 It'll remember that it's been told to behave a certain way whatever you end up typing into that system prompt. But it's not going to learn beyond that experience unless you tell it to. Now there are some features within ChatGPT right now. For example, this concept of preferences and memory where you can tell it specifically as you're having a conversation with it, hey, remember that I like dogs, or remember that I'm allergic to cats.
Starting point is 00:24:18 And then it'll write it down in a note that it can reference later. But when I say remember, I mean that it has a core memory where it doesn't need to reach out to an external source to go look at a note that's been written by somebody or, you know, last week, Spencer told me to write down that he was allergic to cats. So now I remember that he's allergic to cats. I mean, literally learn from the model and kind of a training mechanism, the way that they train the AI from scratch to begin with. Yeah, good, good explanation. I think the keyword that stuck to me is like static, like you give it this set of information or context that's static, but then the conversation
Starting point is 00:24:54 is still like live in real time. So there's some distinctions there. And to give you just a quick example of that, the GPT and creating your own custom GPT is very close to how it works when you're using these AI products on either a care switch or any other web platform. That is essentially how we teach the AI who you are before a conversation starts. We take your name, we take your job title, we take the last five things you've done. And we essentially we tell the GPT about it in a very static way. And until we tell it additional things about you, it's not going to learn anything else about you. Yeah, it's not like proactive. It's not like running forward. It's using what it has. Let's talk about adoption and conversation and what we're hearing from people. Victoria,
Starting point is 00:25:41 you have talked to hundreds, if not thousands of owners and operators and administrators this past year about how they're thinking and feeling or using or the lack thereof with AI. I want to ask you, are home care owners and administrators warming up to AI? Are they still hesitant? Are they ready in 2025? Or what's kind of the landscape in home care? Yeah, great question. I would actually say, in one word answer, are they warming up? Yes. But I'd break it down more into like three buckets. You have this, you know, kind of group A, still highly skeptical, very, you know, fearful in many ways of the unknown. And that kind of closer to the beginning of the year was probably the majority of people I talked to, or maybe at least half.
Starting point is 00:26:27 Then you kind of have this middle bucket. I found a lot of these over the course of the year when I ask about, you know, what are you doing with AI? You get the, you know, I had it write a nice, you know, Mother's Day card for me or, you know, a nice poem for my wife or something like that. Very, very just surface level, you know, I think to Spence's point, kind of like a party trick, right? Just seeing what it could conjure on its own from kind of making a content perspective, but sporadic use. And then the third bucket really are kind of daily users. And within the third bucket, you have the daily users, as in I replaced my Google search with a chat GPT search.
Starting point is 00:27:08 A smaller segment of those, I would say, you know, the people on this call, myself. I, you know, I think Dan made a comment earlier about, you know, not being a software engineer, but he can do a lot on his side before he, you know, instructs the engineers with GPT. I actually coded something for the first time in my entire life this year. So I majored in psychology. I'm a people person. I deal with people. I'm not understand technology, but not a diehard. So I created a cost savings calculator just to show people with some of the advancements what that could translate to in dollars for their businesses. And this is something maybe in former roles that I would have had to ask for and wait months for and have an engineering team take on this big project to make this tool. And in an afternoon,
Starting point is 00:27:56 I coded and tweaked and designed my own really dynamic cost savings calculator. So a smaller segment of those. And what's happened over the course of the year is, you know, the spread on these buckets has changed, where fewer and fewer people are in that first group of just real terror at the concept of, you know, the machines are thinking this is a sci-fi movie that doesn't end well. You know, that segment has decreased. Right now, I'd say most people still fall into bucket B, where they are getting some daily usage, not digging super deep, not doing a lot of deep analysis. For home care owners, practically what that means is I'm having GPT write a thank you
Starting point is 00:28:39 letter or a welcome letter to a family, create a job post, translate something, a lot of that, you know, very surface level, not quite deep analytics yet. And then yes, the third bucket, definitely increasing of people who are using, you know, GPT mostly for like business analytics and, you know, uploading financial statements and having summarizations and, you know, really then being able to put pieces together from maybe different segments of their business. You know, compare this report from my billing manager to this report from my care coordinator. You know, where should they be working together?
Starting point is 00:29:17 Just, you know, maybe kind of uncovering the unknown. So definitely a wide array of application and adoption, but more and more we see daily users and less and less I see people being, I hate to use the word afraid, but there is and has been a fear that seems to be decreasing. What advice would you give to people that are hesitant or on the fence? Like how should they get their hands dirty? How should they dip their toe in? What are some of the initial steps people should be taking to at least get exposed? Yeah, great question. And I would just be very direct. Rip the bandaid off. Go to your app store, download ChatGPT, you know, get it on your phone. And in those moments where you normally would go to Google and do a Google search, ask GPT.
Starting point is 00:30:05 You're in Google all day, every day. You already have accepted for the most part that your phone is listening. And if you talk about something in conversation, you're going to get an ad for it in just a few minutes. Like the horses are out of the barn. So we may as well, you know, optimize and utilize as much as we can. So I would literally say rip the Band-Aid off, get GPT app on your phone, have it just replace those simple daily searches and start to get comfortable with relying on it as a resource and having it, you know, just as a second opinion,
Starting point is 00:30:36 even, you know, on some of this content creation, even if you today are a fabulous copywriter, and you love the way you write something, I would still run it through GPT. Hey, how could I make this better? No one's perfect. You know, you could still give yourself that little edge. And it's such a small, you know, time requirement. It's so easy to do that even getting just incremental gains and daily tasks is worth the effort. I would take it one step further. I don't think anyone can hide from AI in 2025. You got a smartphone in your hand, you're using it. We're all Gemini, Meta, it's everywhere. iOS, the latest model, it's here and it's here to stay. So if you're trying to run and hide,
Starting point is 00:31:18 just know that in the new year, you're not going to be able to. It's basically impossible. Let's talk about CareSwitch for a couple of minutes, Dan. We have made a lot of progress on specific proprietary functions and features inside of our platform that real home care owners, administrators are using today and seeing results with. Can you share kind of what the breakthroughs that we have had have been? And then we'll talk about some of the results that those owners are getting. But specifically, what have we built that is not just cool, but is also driving real gains for people? Yeah, yeah, it's it's been a lot. We've been busy this past year. So I'll do my best to keep it short. I think the three that I'm most excited about this past year that we've worked through are the AI driven forms, our AI shift review, and our AI reporting. Just to give some
Starting point is 00:32:07 background on all three of those so you guys can all understand how those work. The first one is the AI-driven forms. As we've been building the product, we've met with a lot of customers and prospects in terms of how do we want to shape this product around AI. And one of the big insights was just this information gathering. A lot of home care agencies are spending a lot of their time quite literally just trying to get information from people, whether that be a caregiver or a potential employee, a client, a family member, an insurance provider, a payer, whomever, they're trying to get information and document that information into the system. And so zeroing in on that whole entire process, when you get all the information,
Starting point is 00:32:47 then the information, once it's inside your business, is being passed around to different departments and different people for different purposes, right? And so the idea behind our AI forms is we really zeroed in first and foremost on the care plan. And the first big piece of information you get, well, not the first piece of information, but the first piece that we had introduced into the product to gather a lot of free form data around clients.
Starting point is 00:33:13 So the way our system works is you can set, you can build basically a care plan template in your workspace, and you can define all the sections within your care plan and how you want the information gathered. And then when you go to perform a care or create a care plan, you basically have a conversation with our AI assistant. We call him Looper or we call it Looper. And we can build those forms for you. Now, to build on that.
Starting point is 00:33:38 So that's where we started. And we kind of taking a step back, notice this pipeline of information that's passing from people outside of your business into your business. And we know that the process of home care where a lot of times there's sales discussions, marketing discussions into the assessment, the family room visit,
Starting point is 00:33:56 and then the care plan is generated from that. So after we launched the care plan, we went all the way back to the beginning of the funnel and the beginning process of what it means to take a client into your business and make them, you know, someone who's receiving services from your company. And we started with the intake. So very similar to the care plan, you can define what are the sections of information you want to gather when you're having, you know, that first call or that initial meeting with a potential client or family member. And then the same exact
Starting point is 00:34:25 kind of pattern applies to the assessment. When you go out to perform an assessment for a client, you know, what are the sections and what are the information you want to gather when you're with these people and get that information into the system? I should note that, you know, all of this is, these are not checkboxes, these are not dropdowns, these are not form inputs. This is like a free form document you're generating. And it's very customizable and very unique and specific to the person the form is for. So once we built all three of these documents and kind of stood them up side by side, it goes intake assessment to care plan. When a customer of ours gets a new client on the phone, they're able to have a conversation,
Starting point is 00:35:07 take those notes and dump them into our system. The AI will take those notes, reformat that into the template based on the owner's preferred preference in how these things are structured, how these documents are structured. And then from there, it's a one click to generate the initial draft of the assessment. And it's another click to generate the initial draft of the assessment. And it's another click to generate the initial draft of the care plan. Because a lot of the information you're collecting on that first sales call is really the basics. What's your name? What problems do you have? And how can we help you solve those problems? And so that's the foundation of figuring out more and more information when you go deeper into the assessment. So the salesperson might collect some high level surface, you know, problems. Maybe there's hazards in the home or, you know, mom has dementia.
Starting point is 00:35:51 But then you go a click deeper when you're in the home, you're having the assessment, you're actually looking at the environment, you're actually looking at the client or the potential client, you know, with the family to their face. And so you're getting more and more of that information. So we think about this and we talk about this internally is really just snowballing information, you know, the first touch collects a little bit, and every touch that follows sequentially after that is a little bit more information. And like Spencer talked about earlier, you know, it's not training, it's just more data. So the more data that we can get in, the more efficiently we can help you, you know, follow your process and the templates and the forms that you like, but also
Starting point is 00:36:30 basically spin those up in a matter of seconds and have those documents ready to go. So that was one I can keep going, Miriam, or we can, we can stop there. If you want to talk about AI forms, I'm happy to do that. No, that was great. I think you broke that down really well. I think the only thing that I would add is like one of my biggest ahas has been how useful this is for nurses. You know, nurses love what they do because they get to spend time with people and helping people. But at the end of the day, they're spending hours on documentation and that's inevitable. But we're seeing that AI can significantly reduce the amount of work that they have to do on documentation, which is phenomenal. So I'm glad you kind of ran long on that. And I think
Starting point is 00:37:08 that painted the picture of how this works for real people. Let's go through the other two really quickly. And then we'll keep going. This is great. I'll try to keep it quick. So the next one that I'm very excited about that we got through this year was AI shift review, AI shift review, you know, we know home care agencies, one of the biggest time sucks in the back office is just getting shift documentation back to the office, looking at it and confirming that it's right. And part of that is not just saying, you know, they fill out all the right information. We understand and we know that with different payers and different preferences of all different types of parties involved in someone's care, there's a lot of complexity and rules. And, you know, if this happens, then we need to do that. And we need to make sure that the
Starting point is 00:37:48 billing and the payroll is going out the door on time. So the goal of this feature that we released this past year was really to just speed up the review process. What we allow you to do is we allow you to create rules custom to your business within your workspace that tell our system what you're looking for on a shift. And so let's say you want to know if someone's clocking in or clocking out early or late, right? You can write a simple rule that says, you know, block this shift from being billing or payroll approved, you know, if someone's clocking in more than 12 minutes early, for example. And these rules can be really dynamic. It's not just a simple, you know, if someone's clocking in more than 12 minutes early, for example. And these rules can be really dynamic. It's not just a simple, you know, 12 minutes early, it can also take into account other attributes of people that are involved in that shift, such as tags on the
Starting point is 00:38:35 caregiver profile or criteria about that caregiver or criteria about the client. And so it really takes into account a lot of different properties to make a decision as to whether or not the shift should be blocked from billing or payroll. And the way that it works is if, you know, if no exceptions are caught on a shift, our system will just automatically approve any shift that meets that goes through the rules filter and doesn't get, you know, dinged with an exception. So our goal here is to really, and we've proven this, is that we can really reduce the amount of time spent looking at shifts because we're just giving, we're just kind of taking the most critical things based on the things you care about
Starting point is 00:39:15 and bringing those to the surface and just auto-approving basically everything else that met the criteria. And it's already on a timesheet. It's already on an invoice. It's ready to go. Of course, you can pull that stuff back, but that's kind of the general idea here is we really want to automate and have Looper be the first, Looper, our AI assistant, be the first thing to look at the care that was documented and be the initial review so that, you know, people can spend a lot less time being the first person to look at the shift documentation when it comes in the door. Can you tease how beneficial this will be for something like Medicaid? As everybody knows, there's all of these like requirements and restrictions around Medicaid. And when you think about an AI filter that can review shifts from a Medicaid lens, like, yeah, that's really exciting. So
Starting point is 00:40:05 can you just speak to like what that will look like? Yeah, for sure. And it gets really crazy when you have, you know, a multi payer scenario or a multi client shift, right? You can have, you know, up to three payers or even more payers on a shift or on a on a client service. And so all those payers in a Medicaid world, I know Medicaid doesn't typically have multiple payers, but with all these combinations of payers, you're going to be able to have rules for all your different payers that are very custom to those payers. And so the AI will look at the shift through the lens of each of those payers. You know, maybe payer one is VA, maybe payer two is LTC, and maybe payer three is private pay, where they'll just cover
Starting point is 00:40:46 the remaining bill if there's, you know, excess charges, it will look at the shift through those lenses. And it's kind of a waterfall approach where it'll basically take the rules from that first payer and apply them down through the stack of payers within the system. So again, very dynamic, it can be very, you be very configurable to those different rules and all that different complexity that people have to manage to really speed up the efficiency of billing specifically in that Medicaid scenario. But I guess to be more specific about the Medicaid one is getting the claims perfectly configured and right the first time. So you're not getting rejected, you're not getting any information sent back, and you're getting approved and paid
Starting point is 00:41:27 in a timely manner. Yeah, that was great. I know this is a little bit conceptual for us to talk through. But I think I hope this is just illustrating to people like the potential of like a rule based system, you essentially like, write the rules, then the AI acts as the filter reviews the shifts. And then someone in the office then takes the next step, like what needs to be acted upon or what is perfect and can be passed through. Like that alone is so powerful. And I love that we've spent so much time on it this past year, and we're coming up with these new use cases like Medicaid, VA, billing, like taking it to the next level because of what's possible. And just to really emphasize, so Miriam, so people don't misunderstand, you know, these rules are not limited to what software we provide or what UI we provide to create the rule. The rule can be, it's a string of hurdle or it's not difficult to interact with setting
Starting point is 00:42:27 this up. It's plain language. This is what I care about for this person in this circumstance. And you can say it just as that. So I want to make that clear. Essentially like infinitely customizable. Right. But also I'd add like we're going to support on that.
Starting point is 00:42:40 It's not on the agencies to figure all of this out. It's like we're going to do the due diligence to write these rules with you, make sure they're perfect, and then learn as we go. Yeah, go ahead on the last one. Let's finish. Okay, last one. I'll be quick. So last one, I'm sure Victoria can attest to this. You know, the number one question we get is like, can Looper, our AI assistant, report on this data, data XYZ, right? And so that's like a very common question we get from our customers and prospects is, I look at my data in a specificZ, right? And so that's like a very common question we get from our customers and prospects is, I look at my data in a specific way. And I'm looking for a very specific insight to my business, because that's how I like to think about things. That's how things have worked for me
Starting point is 00:43:15 in the past. And that's, that's what I care about, right. And so in traditional agency management systems, you're going to really have kind of two parts of the system. You're going to have like the classic dashboard, which is hand coded by the software engineers that work at that company. You're going to get all these tiles and you might be able to like pick what tiles go on your dashboard. But, you know, if that tile that you're looking for doesn't exist, you're not going to have it. Right. And you put in a request and you hope they build it. The other part is the querying and like the debt the data harvesting so the ability to go in configure what data you want and pull a report right literally like a spreadsheet or a file
Starting point is 00:43:55 out of the system so then you can take that and go into excel and you know write a formula and manipulate it and do whatever you want maybe combine it with other data which is very common very common. We see it with a lot of management teams. They'll take all these different spreadsheets from their departments and teams and try to make a super spreadsheet by hand and hope that the data is all accurate and right. And so with our recent AI enhancements that Spencer worked on specifically, that's completely different. It's, it is all through the chat interface, but it allows you to skip the pulling of data and get right to the insight. You can just, you don't have to scan through lines and lines of spreadsheets. You can just ask a question. If you're looking for something very specific, you can just ask for that thing and it will return to the best of its ability, right? There's always room for error here.
Starting point is 00:44:48 The second piece is just the ability to query for data. So if you do want to pull a report, you know, in traditional systems of old, you would have to, you know, figure all configure all the faceted searches in the right way and get all the data, tell the system what you want, and hopefully it will build the spreadsheet that you want. But there are limitations to that. But not with AI. With AI, you can say, I want to pull this data from over here. I want to pull that data from over there. Combine them.
Starting point is 00:45:15 And maybe I don't know what insight I'm looking for. Just what insight should I be looking for at this point in time? And so you could even ask that question. So that's, you know, you can basically pull any data from within your workspace, combine it together and get a report. You could pull that out if you wanted. And again, we know some people still are diehard Excel users, so they're going to do their data manipulation outside, but you would still be able to continue a conversation within CareSwitch and manipulate the data and have conversations about that
Starting point is 00:45:45 data to gain insights. The last piece, kind of like touching on that earlier point about merging data, maybe you have another piece of software that runs alongside CareSwitch in your software stack, right? And they have reports and they generate reports for you. CareSwitch is not limited to the data within the workspace. If you want to run a report in some other software and drop it in and compare it or contrast it to what's in CareSwitch or merge it and do some manipulation or calculations on that data, you can drop that in and basically introduce new data into the system and it will be able to handle that and give you the insights and whatever you need to get from that.
Starting point is 00:46:26 So lots and lots here. There's like no one way to do it, which is really cool. And it's, you know, we're seeing people do a lot of different things in a lot of different ways, again, because everyone looks at their data and how they run their business differently. And we understand that. And that's where AI is really actually really shines when it's not prescriptive. It's specific to you. Yeah, that was awesome. I think this lights all of us up just the possibilities here. I think the paradigm shift for me and maybe for some of us
Starting point is 00:46:56 this year has been less about reports and more about insights. And that's where AI thrives is I think in home care, we've talked a lot about data, and we've come a long way. But I think what we're actually after is the insights behind the data. Yes, numbers on a graph or in a spreadsheet are great. But what does that data mean? How do I extract answers and insights from that data? And that's where AI, like you've talked about, is thriving is like, it has all the information, it has all the data, now prompt it or ask it questions and gather the insights that you're looking for. And then, you know, be proactive,
Starting point is 00:47:31 ask it what you don't know or what you aren't thinking about, and it can help you like proactively look at your data from different lenses. So a lot more on that in 2025 for us. One last thing on that, you know, whenever owners ask me, you know, do you have a report for
Starting point is 00:47:45 that? Right? Where is this report that I always pull from this other system? I'll ask them, you know, what are you looking for on that report when you run it? Because a lot of times, it's they're like eyeball scanning it, and they're looking for the number or numbers at the bottom, which tell them something because they've seen this report before. And they, they can kind of like they have a feel for their business, right? I'm just saying, what we're saying is maybe you don't need all the lines. We just tell you the number and then you can have a conversation about that number. And if you want it to, you know, check its work, like how did you get this number? All those lines of data are in our database. We have access to them and we can tell you, you know, how we did
Starting point is 00:48:23 that calculation. So we're trying to just cut through the noise and get to the point so you can kind of move on. Yeah. I hope everyone listening to this is like letting their minds run on this. I think this is a big paradigm shift for home care in the coming year or years is just how we think about the data, the information, the documentation in our business, less so like reactive and stagnant and more so forward proactive, like insight driven. I think that's going to be kind of a big push in the future. Victoria, you work closely with a lot of our customers and are talking to these prospects and you've seen firsthand the results people have driven from some of this technology
Starting point is 00:49:01 that we built that Dan has talked about. What have been some of your takeaways or ahas from people actually using this CareSwitch system daily? Yeah, this is my favorite part for sure. Because we know real people who just this year had their entire professional life change. So when Dan talks about the AI forms and that intake process, one of the things for me that stands out is the ability to just do all this with voice input. So if I'm doing an intake on the phone, I can hit a microphone, tell it everything I just learned. If I got hired yesterday and I'm not really sure yet what the
Starting point is 00:49:37 administrator wants, I can start a conversation with Looper and Looper will walk me through every single question I need to ask so that I can satisfy the requirements of that template. Then next step, I'm a nurse and I'm going into this home. I don't have to read and analyze this big note from the receptionist or the sales rep. I can create my assessment from their intake node. It copies everything forward into the right locations. I immediately see what information I'm missing. I can vocally input that. Then care planning. Okay, I need to sit down for an hour and think about risks and precautions.
Starting point is 00:50:08 No, I don't. Looper knows all the conditions. Looper knows the home environment. Therefore, Looper can tell us what risks and precautions are. So that process, that intake assessment care planning process, when I ask agency owners and administrators, how long was that taking, just data entry alone, from the time they're done and they're leaving the house from the time they're done and they're leaving the house until the time that everything is in your system and a scheduler
Starting point is 00:50:29 can pick this up and get the perfect caregiver and schedule this? I mean, on average, I would hear four hours. Sometimes like end of the day, I get back to the office. I'm typing for two hours every day. Some agencies would even tell me like maybe tomorrow it could be 24 hours. We have a testimonial on our website, one of our care coordinators. This was a two-hour process for them. So they were incredibly efficient, you know, by kind of old school standards. They're down to 15 minutes. So now also the modalities of input kind of dispense this point earlier. I can type all this in if I like text and I'm a typer and that's how I operate. Vocal input's probably most common. You know, so there are even options there. If I walk into the home and I'm super tech savvy and I'm taking in a tablet, I can hit the microphone and do this question by question live in front of
Starting point is 00:51:14 them. You know, we know some of the best in the business will tell you I'm putting this pen and paper down the day I retire and not a day sooner. Great. We're never going to micromanage you at that level. I'm not going to micromanage you at that level. I'm not going to force you to be a robot. Take your paper notes, get in your car, hit the microphone, read them all in. Or if you have fabulous handwriting, take a picture of those notes, upload the photo, and my system can take it from there. You are done when you're driving away. There is no more looming cloud of data entry following you around all day. And now I have to go and type for two hours. Like that's over. There are real people who had those two
Starting point is 00:51:49 hours every single day eliminated from their work this year. That alone is crazy. The second part, shift review. So when I talk to agency owners, doesn't matter what software you have, right? Anything that exists that I know of, there's still someone at the office who's visually checking every single shift before they're invoiced out, checking for errors, checking for specifics regarding payers so they don't get claims rejected, whatever the case may be. When I asked the average agency, well, how many of those shifts that you're looking at every single day are business as usual? It's Tuesday. They got there on time. They did their tasks. They left on time. It was a normal day. I've heard, you know, I'd say average is 70%.
Starting point is 00:52:31 7 out of 10 are totally fine. Really didn't require any effort on my part. Super efficient agencies, I'm even hearing like 95%. So those shift review rules that Dan spoke to, you know, you're writing in all of the things, just word per word, exactly what you're looking for. Now, if no rules are broken, every hurdle is cleared, those are automatically going on invoices that are automatically generated. Now you're only looking at the ones going wrong. So you shave off 70% plus of the time requirement, but also just the energy now becomes focused on where are the problems, let's put in better policies and procedures to address those. And now as you scale and the agency grows, this menial task isn't scaling with you.
Starting point is 00:53:19 Now, because you're improving and identifying issues and correcting them, the system becomes more and more automated the more you scale. So it becomes more valuable as you go. That for agency owners, and I know we have just a few moments left, but I mean, it's always so many important doing that, right? Like it's not an assistant. It's the owner. It's the admin.
Starting point is 00:53:40 It's a billing manager. To free up that amount of their time and brainpower to put back into your people. You know, obviously turnover is a massive issue. Every agency pretty much in the country is concerned about it. And so if you can dedicate your time to focus on your people and making them happy and making them feel like they're part of a family and not just sitting in front of a screen all day, you know, that really translates into more success for the agency as well. So it's really a ripple effect of, you know, eliminating this menial, never ending, you know, very easily automated
Starting point is 00:54:12 task, you know, really opens up a lot of possibilities. If we had more time, I would let you keep running on this. I think you and I, you and I both love like the facial reactions we get from people when we show them this technology or when they use Looper for the first time or when they get that insight. They're just like literal facial reactions, the oohs and the aahs. It's just so satisfying for people to use this technology, see the results, save the time, drive the outcomes for their business. That never gets old.
Starting point is 00:54:42 And I would love really quickly just to answer Suzette. Thank you so much for your question on the AI discerning like synonyms. It does a really fabulous job with kind of interpreting what you meant. So I'll give a quick example. In my Care Switch demo system, one of my example clients is named Lillian Lewis. So the microphone picked up one day that I said, Lillie, A-N-D, Lewis, as though it were two people. I didn't, I just hit the microphone. I spoke everything in. I sent it into my Looper conversation without double checking myself. You're so welcome, Suzette. And as I did that, Looper actually caught it for me. It said, wait a minute. I think you mean client Lillian Lewis, not Lily and Lewis. So even when the microphone wasn't like
Starting point is 00:55:28 spot on with what my vocal input is, I'm hiding it as best I can, but I have kind of a Southern accent. So someone else in my system is named Al, A-L, and sometimes it reads it as Owl, O-W-L, and Looper will still catch me, Southern lady, and correct me. So the technology is fabulous. And little things like that aren't hiccups in my day to day. Yeah, I would just echo that. It's surprisingly good. I think all of us have been just surprised at how phenomenal it is and how much better it's getting so rapidly.
Starting point is 00:55:57 I know we're almost at time here, but I want to spend just the last couple of minutes talking about what we're excited for for 2025. I want to hear from each of you what you're looking forward to in regards to just AI in general or something specific that we're thinking about in building. Let's have each of you share. Spencer, why don't we start with you and then Dan and then Victoria. I am personally excited for the chatbot experience to go away. I think this is like a really early manifestation of AI. It's the easiest to talk about. It's the easiest sort of way to approach it for someone who hasn't really used it for the first time. It freaks out people a lot less than, you know, the talking AI variety. But I do think
Starting point is 00:56:36 that as AI gets better and better, the ultimate goal is for it to drop out of the spotlight. The perfect working AI is the one that you don't notice. So sort of the methodology that Dan was describing earlier, where you just have a really, really diligent system that's proofreading, checking, looking for errors, trying to figure out if it can pre-detect insights to bubble up for the humans. I think these are the far more useful and innovative uses of AI.
Starting point is 00:57:08 I think the chatbot will always have its own place. You know, I will always have any number of miscellaneous questions that I would want to ask at any given time. But I think the actual features that are going to change my day-to-day are the ones that I'm not going to necessarily notice at the surface, if that makes sense. That feels, yeah, cryptic and kind of advanced. Not over my head, but it's like the more advanced the AI gets, maybe the less prevalent it will be in our day-to-day. It's like it'll actually be working behind the scenes. Yeah, it's like, do you have your own personal Jarvis that you talk to all day? Or do things just automatically happen as if your next actions are being predicted on your behalf by something somewhere?
Starting point is 00:57:52 I'd much rather log into a system in the morning and just have the three things in a notification tray somewhere that's been caught over the last few hours. Out of the hundreds of thousands of things that took place in your business over the last few hours out of the hundreds of thousands of things that took place in your business over the last 24 hours, here's the three things you need to pay attention to as you're drinking your morning coffee versus me logging into the system in the morning and then asking the AI, did anything happen today? That sort of conversational style has a place,
Starting point is 00:58:20 but I'm looking for the real magic. And I think the real magic comes when it starts to do things on your behalf without you being aware that it's happening. How far are we talking here? Are we talking we're going to see some of that in 2025? Or you think we're further out from something like that? I think we'll always have a tendency to say, hey, look, this is the AI that did this just because it's so cool and we're so proud of the things that it generates. But yeah, it's fully within reason that you can have a system that looks through all of this documentation. We never really want to let the AI take action on your behalf. So again, this is mostly for insights. It's going to be quite a while before
Starting point is 00:59:01 the AI can go out there and just do your job while you're sipping mimosas on a beach somewhere. Don't get me wrong. But in terms of bubbling up the actual important insights to you seamlessly as if it had predicted or read your mind, I think that's kind of the true test of human level intelligence or superhuman level intelligence. Can I predict what you want to do next and can I do it for you? I'm not going to lie. When Bill Gates starts talking about three-day work weeks, my ears perk up. I think all of us are looking forward to that. So we're maybe not too far out from that. Dan, why don't you jump in? What are you looking forward to?
Starting point is 00:59:33 I'm going to talk about us. I'm really excited about building on the forum stuff we've done. Like I said, the big idea here is there's just a lot of data and a lot of information about people that you're trying to stay on top of like all day long and that that information changes and so the one the one exciting thing that we're going to be working on is is this this idea of like dynamic care planning so and it starts with the caregiver you know the caregiver is the person who is quite literally the front line of your business. And they're in the home.
Starting point is 01:00:05 They're with the client. They're with the family sometimes. And they see and hear and do things that, you know, are not always, you know, captured. And, you know, I'm not suggesting it's like a Skynet thing here. But we do already do a great job of collecting rich documentation from the caregivers through our native app. And where it stops is, you stops is at the shift review. So the caregiver works the shift, the data comes back to the office, the AI checks it, and then if you need to be notified to look at something that's not quite right, we'll
Starting point is 01:00:37 tell you. But it stops there. And I want to keep going. I want to take that information that's coming back into the office, run it back through the AI and make recommendations on how the care plan can be improved instantly. And so really turning these care plans from like a static document that might get touched, you know, 30, 60, 90 days when reassessments are happening to something that's living and breathing by the shift that is being worked by every caregiver day in and day out. I think we can really speed up the feedback loop in how care is,
Starting point is 01:01:13 all the, you know, these caregivers don't write documentation for nothing. They're supposed to tell you what happened in the home. So how are you taking that information and turning it back into the care plan for the caregiver that's about to show up in two hours for the next shift. And the example that I always like to use is, you know, maybe there's a rug inside of Mary's house that, you know, people keep tripping over, whether it's Mary or the caregivers that are visiting. And maybe there's been two small notes on a task where a caregiver wrote that the rug is kind of problematic in the living room. You know, that would take someone today either getting that piece of paper from the house whenever that comes back, or if you're using a digital system, finding that shift in a sea of
Starting point is 01:01:54 shifts, finding that task, finding that note, that little detail for you to be like, okay, is it worth my time to go back and update this care plan so the next caregiver knows that the rug is truly a hazard in this home how can we make that instantaneous and just through your approval the care plan is updated and how do we surface those little insights to make it really a living document in real time up to date you know to the hour in that person's current condition in their home at all times so that's probably the most the thing that I'm most excited about is like, you know, Looper is not an accident name for our assistant.
Starting point is 01:02:32 We want to loop, you know, pull the care out of the home and loop it back into the care plans. You're providing that really premium experience that really is up to date in anyone's condition that you're providing services to. Love it. Super, super awesome. Victoria, what are you looking forward to?
Starting point is 01:02:50 Yeah, definitely ditto on both the guys. So for me, I also think about within our system. There are things that we know technically are becoming possible that we're obviously testing. Something I appreciate about this team, they don't go all green lights the moment a concept becomes somewhat tangible, right? We know you're running real businesses. So we talk about that intake process. And, you know, I mentioned you get in your car, you hit the microphone, you read it all in, and now you're done.'ll see in 2025 the ability to walk in, hit the microphone, have it just record the entire encounter. It will understand, you know, this is the adult daughter, this is the spouse, this is the client, this is the care coordinator. Catch everything everyone says, you walk out, the care plan's done. I mean, we are like right on the cusp of that.
Starting point is 01:03:41 That's, for a care coordinator, that's, that felt like a sci-fi movie that doesn't feel real to them that they could have just no real time requirement to get that level of documentation done and have it be, you know, really rich and accurate. So that I'm super excited about. Secondly, for me right now today, you know, it's AI driven, but human led. You still have the human in the loop on everything. Speaking about earlier, you know, it's AI driven, but human led. You still have the human in the loop on everything. Speaking about earlier, you know, those three buckets of people, there are still people who are anxious. So we are, you know, having you be the one deciding what actually happens.
Starting point is 01:04:14 So having a little more automation, I think to Spence's point where it's being more proactive and telling you what you should be asking instead of waiting on you to ask, that's exciting. Lastly, though, for me is then the last step is going to be starting to expose some of this technology to caregivers. We all understand in this industry that they are, you know, historically very low tech, you know, slow to adopt technology. I know a lot of you are struggling to get them to kind of use an app anyway. We have offline mode, by the way, so ours is easier and better. But what I'm excited about is their interaction with AI, their ability to say, hey, this is going on with this person. What should
Starting point is 01:04:50 I do right now? We're sitting here. It's quiet. What questions should I ask someone who was born here in 1941? You know, those conversation starters, even obviously everyone on this call, everyone listening knows that, you know, the caregiving staff in this country, a lot of the support we get are from people who maybe don't speak English as a first language. And even if they speak great English, English literacy is quite challenging. So their ability to hit a microphone, speak in all of the care summary and task summary for a shift, and then have that translated and documentation be done for them. I think, you know, getting them over those kinds of barriers will increase adoption of technology as a whole. You know, you're not going to be missing documentation because they didn't know how to spell something and they don't want to be embarrassed, right?
Starting point is 01:05:35 They're just going to say it and you're going to get it. And all of those gaps are going to start to close. So for me, that's super exciting. Something that at the HCAOA conference, Miriam hosted a really great caregiver panel and we got to meet some caregivers and really show them and test some things with them where they spoke in in their native language and then seeing, you know, the output in English. They were astonished and so excited. So the way that everyone feels right now listening to this call, how excited you are about what
Starting point is 01:06:02 this can do for your business, we want to next year see that roll over to your staff and have them be excited to participate, take these steps forward technically, and make their jobs easier as well. Fantastic. Team, this has been really fun. We're going to wrap up here. Thank you for your comments, for your preparation, for the insights that you've shared. I hope everyone has enjoyed listening to this. We've got an engaged group that's still on here with us as we run over an hour. But I hope everyone that listens to this just feels excited about what's next. 2025 is around the corner. The next 12 months, we're going to keep moving hot and fast, and it's going to be really exciting. So lean in, follow along, follow along with us, follow along with the news, tinker with it. Don't be bashful. We're all
Starting point is 01:06:43 learning and figuring this out together. I know some of the conversation today has been conceptual. For those of you that haven't seen the technology that we are building, reach out to us. If you're considering AI, new software, just innovating in 2025. We all have everything we've talked about today is essentially real in some form or another today. And we're really transparent and open and showing this to people. So don't hesitate to reach out to us if anything we've shared today has been interesting to you. Guys, great job. Thanks for being here. We're going to go ahead and wrap up. Like I mentioned, this is the last episode of Home Care U for 2024. I hope everyone has a Merry Christmas and a Happy New Year. And we'll look forward to seeing everyone back in the new
Starting point is 01:07:21 year. So thanks for being here and take care. Bye everybody. Happy holidays. Bye everyone. Thank you. That's a wrap. This podcast was made by the team at CareSwitch, the first AI powered management software for home care agencies. If you want to automate away the menial of your day to day with AI so that you and your team can focus on giving great care, check us out at careswitch.com.

There aren't comments yet for this episode. Click on any sentence in the transcript to leave a comment.