Miriam Allred (00:10) Hey everybody, welcome back to the lab. It's Miriam Allred, your host. Great to be here. Hope everyone's having an awesome week. Today in the lab, I am joined by Dave Dworschak, the head of growth at Phoebe. Dave, welcome to the show. Dave Dworschak (Phoebe) (00:22) Yeah, hey, Miriam, thanks for having me. Miriam Allred (00:25) Let's get right into it. You and I have been catching up before we jump on, but I say this a lot, but a new face kind of new name in the industry, yourself and Phoebe. So let's start with an introduction. Do want to talk about your background and tell kind of your personal story? Dave Dworschak (Phoebe) (00:39) Yeah, sure. So I have been, you know, I'm so I'm Dave. Nice to meet everybody. I've been in and around kind of health care in some capacity more or less my whole career. Right out of college, which I don't have a technical or healthcare background, transparently, but right out of college, I helped start a company that was in the early days of electronic medical records for acute care hospitals and providers. So we basically helped people migrate off a paper into EHRs for the first time. I actually thought that would be like a fun little project at a college, maybe six months, a year, something like that. And ended up sticking around and helping grow that business from the first employee in 2010 or about seven years later, were roughly 200 people moving hospitals and moving data for hospitals in various capacities all across the country. Really interesting experience there. It was more of a services business, but I did have the chance to build a basically BPO offshore in the Philippines through that. So traveled back and forth and got to do kind of the onshore offshore healthcare operating model, which was very fun. And then I left that in 2000 and let's see, 18. and I started my first software company. Transparently did not know a lot about technology at the time going into that, but had a couple of ideas with a good friend of mine who was a traveling nurse and ⁓ just dove right in to figure out how to do a startup. And we ended up over the course of around five years building a platform for traveling healthcare workers that are kind of moving around from hospital to hospital, helped build a centralized digital wall or a clearing house, so to say, of their credentials. and paperwork so they're not having to get re-credentialed every time they move from job to job every few months. And then on the employment side we had a product for staffing companies to both manage compliance as well as like self-serve or provide the healthcare workers a self-serve mechanism for getting through like the job search recruitment compliance etc. So over the course of that like five-year run we I had several hundred thousand users, several hundred companies on the like staffing company side of the fence using the product and eventually sold into a private equity backed very large traveling nurse staffing firm. And at the end of the day, moved merged my team of about 60 people into their team of 500 or so and had quite an interesting experience working both on the early stage startup is as well as the, what it looks like to merge in with a very large enterprise org doing, most one of the highest gross staffing ⁓ operations on the healthcare side in the country. Miriam Allred (03:15) Very cool, very interesting. A wide array, but mostly healthcare based, which is neat. And there's a lot of crossover. Obviously healthcare is vast and there's a lot of different facets, but relevant to home care and leading to where you are. You're based in Orlando. Are you a Floridian or where are you from originally? Dave Dworschak (Phoebe) (03:32) Yeah, I'm in North Florida, actually in Jacksonville, so close, but I grew up in Florida. I was here most of my life. Maybe not super interesting for everybody, but I moved out of Florida 10, 11 years ago. For the last 10 or 11 years, I lived in several places, Louisiana, Colorado, several stents overseas. have a couple of young kids and as tends to happen, moved back closer to family a couple years ago from Colorado into Jacksonville. So I'm here in Florida and then I know we'll talk a little bit about Phoebe today as well, the company that I'm working on, but ⁓ I'm here in Florida, but most of our team and company is based out of New York. Miriam Allred (04:14) So Phoebe, kind of a newer name in the industry, but popping up more and more everywhere. And I've just gotten acquainted with you over the last several months and have been really impressed with you and the team. so talk a little bit about Phoebe, overview, who you guys are, what you're up to, what you're building, what problems you're solving, just kind of like high level overview. then we're going to talk more in depth today about AI and about the problems that you guys are already quite literally solving. Dave Dworschak (Phoebe) (04:40) Yeah, sure. you know, at a high level, like Phoebe, we build AI teammates for home care and home health businesses. We've been working on this since really late 2024. Justin, our founder and CEO, was also kind of a second time founder, has built and sold software companies in the past. And, him and I met and he was working on a thesis in this space and we spent a good amount of time, we'll say, ideating and talking to people and talking to healthcare workers and caregivers and folks all across the market, not just on the home care home health side, but also in SNFs and other assisted living type facilities and post acute space. And we really zeroed in on the idea for Phoebe when we met the home care scheduler who And really, even before that, had tons of workers we were chatting with on some of our early products, caregivers, healthcare workers, and they were always saying, my job of keeping myself scheduled is pretty difficult, but you should go talk to the people that are scheduling me. And ultimately ended up landing on what we do today, which is building AI scheduling assistance for home care and home health businesses, and have had that up and running now for a little more than six months. Miriam Allred (05:54) Amazing. you say, yeah, AI assistance. we're talking about the role of the scheduler. The scheduler has primary functions and maybe secondary functions. both know every home care business, like a snowflake runs and operates a little bit differently. go one layer deeper, some of just high-level tasks of the scheduler that you guys are specifically plugging into today. Dave Dworschak (Phoebe) (06:16) Yeah, totally. As great software companies should do, we talk to people who are doing the scheduling work every day and that's how we decide what we're going to build and what we're going to help with first. And the core activities we're focused on today are number one, the scheduling outreach where you've got new clients or you have caregivers going on vacation or you've got call outs, which is actually kind of the first problem we started to tackle. And the scheduler needs to reach out to 10, 15, 20, 30, 40, sometimes 50 caregivers in order to try to get a shift filled. Phoebe helps with that by going into the EHR scheduling platform of the home care business and extracting all the information, you know, the scheduler would typically be reading in order to have a conversation with the caregiver via phone via text about a shift or case opportunity that's open. And then rather than the scheduler picking up the phone and say calling 15 people back to back over the course of 45 minutes, Phoebe's voice and SMS texting AI assistant can simultaneously reach out to the caregivers and have conversations all at once. So that's kind of the first thing we launched and also the thing that's most exciting for folks, especially because schedulers spend so much time making calls and sending text messages that don't lead to people taking shifts. So you've got Phoebe, this AI assistant that's really being directed by you as a scheduler that is going out and doing all the dirty work for you and then bringing you back to people who are interested and qualified and ready to talk. So that's the kind of core thing. And then we've got a few other activities in and around scheduling and schedule coordination where we're doing day before shift confirmations automatically and getting explicit confirmation from caregivers about their shifts that they're going to be at their shift or shifts for the following day. It's very flexible the way we set that up for clients. Phoebe persistently follows up, follows kind of the tone approach culture that the agency would typically take to communicate with their staff. So learns from that and then executes in the same fashion that the scheduler or coordinator would. And then on the backside of shifts, we are also monitoring for miss clock ins and clock outs. So faster than a scheduler would typically even learn that somebody has missed their clock in or clock out. Phoebe has already texted the caregiver, called the caregiver, had a conversation with them, tried to troubleshoot, figure out what's going on. Hopefully just get them to clock in or out. But otherwise escalates it up to the scheduling team on those edge cases that the AI can't quite handle. Miriam Allred (08:35) and it's text and voice based, correct? Dave Dworschak (Phoebe) (08:38) It is, yeah. And it's dynamic based on kind of the situation where if it's urgent, it might jump right to phone, or if it's not so urgent, it might just do text, or if it's kind of urgent, but not necessarily as something that needs to be resolved in the next 30 minutes, it'll do a combination of both. Miriam Allred (08:52) Yeah, and for any either skeptics or AI enthusiasts, you guys have like a model right baked in onto your website. And so people can actually go and try it themselves, which is where I started. And then, you know, you all showed me how it works. But it's incredible. Like, I want to we're to get back into like these specifics that you just shared. But I want to like zoom out a little bit because AI is moving so fast. And I know that's cliche. Everyone's reading it in the headlines like it's everywhere all the time. And I get that. And it really is. But the speed at which it's improving is amazing and also important because before we jumped on I was telling you like, you know, I was at Care Switch over a year ago now and the improvements that you all have seen in the last three to six months, like it's moving so fast. And so what voice was a year ago, two years ago is night and day from where voice is at today. And so I guess my just kind of like open-ended question is like, yeah, what improvements and advancements have you seen recently? that are getting you guys excited about like where this is really going and how quickly we're gonna get there. Dave Dworschak (Phoebe) (09:55) Yeah, that's a great question. And I would say over the last six to 12 months, we've gone, especially in the voice AI, but this is just like LLMs in general. We've gone from interesting and useful to reliable and effective. what previously was interesting where you could have a voice AI make a call six months, a year ago, it worked. It wasn't quite. ⁓ quite there. ⁓ but we've gone from the point of like, you know, this is clearly AI to like, this is a very natural sounding conversation that, many caregivers, even though they're aware that the call they're receiving is, from an AI or from Phoebe, that they forget and just have casual conversations. So we've gone from like, kind of like outbound, just like a voice that can call and say something a full, AI that can have a natural sounding conversation, follow instructions reliably. handle exceptions, handle edge cases. So I would say overarching theme would be, we've gone from interesting to quite reliable. And maybe the caveat I would say is there are two ingredients to make sure that it works well. You've got to have someone that understands the business logic that is going into the conversations, really understands the kind of edge cases and business and kind of culture of the caregiver or whatever the situation is. And then, still pretty important in our opinion to have somebody, you know, very technical expertise to kind of lead this where you can go out today. And I encourage everybody to, even as a non-technical person or non-engineer, you can go out and do tons of awesome stuff with AI. But then getting into the point where you're kind of layering, layering AI on top of your business and your software that is not just say providing you information in the background, but it's like real time running apps. aspects of your business. ⁓ It's that like sweet spot combination of the right technical team plus the right business logic combined that makes things work. Miriam Allred (11:48) Yeah, really, really well said. few months ago or a year ago, was like, was like flashy, cool idea, conceptual, know, futuristic. But now it's like you talk to an AI and you don't know it's an AI. Like, in my opinion, like, that's the bar is you don't actually know that you're talking to an AI. And this might be an overshare, but I was ordering at Panda Express like a couple of weeks ago. And it was an AI and it was very clearly an AI, but I was like, I'm gonna throw some edge cases at it, gluten free, dairy free, how much sugar? I was like, let's just get after a little bit. But it's amazing. You feed this thing so much information, it can handle the edge cases. And I bring that up because in home care, it is so nuanced. We're dealing with people's life. Every single caregiver, every single shift, every single client, there's so many nuances and... And that's actually where AI thrives when it gets good is like it can handle the edge cases. It can handle the complexity. It's just a matter of, training it, building it up and the technology being able to support that. Dave Dworschak (Phoebe) (12:51) Yep, 100 % agree. Yeah, sorry, I've also encountered some of the voice AI ordering systems lately. It's quite interesting. Miriam Allred (12:53) I want to talk about... It's quite interesting, but it's like it's here, you know, it's like, you you encounter it every day, every week, but very quickly that is going to be like standard practice. And so rather than the tension or the friction up against it, it's like, no, we just embrace it and it's going to get better. Dave Dworschak (Phoebe) (13:21) Yeah, totally. Miriam Allred (13:23) I want to hear kind of like industry pulse check, what you're hearing from operators. So obviously listening to this is a bunch of owner operators, some of which are using AI every day, some of which are like tinkering with it, some of which are maybe holding out to see what's coming. you just told me your calendar is booked out almost a month right now, like present day. And so you are talking to owners, operators every single day. I'm curious what you're hearing from them, like average consumption. tools that they're using, fears and frustrations, give me just like a sneak peek of like what you're hearing from lot of operators today. Dave Dworschak (Phoebe) (14:00) Yeah, great, great question. It's interesting, know, I do honestly still talk to some people who, feels like a surprise to me, but like just don't know that voice AI is even possible. And that's becoming like less and less of our calls, especially because, you know, when people land on our website, we give an opportunity for people to just like. chat with the AI. So a lot of people have actually done that before they talked to me. But it's all the way from the spectrum of people who don't know that it's possible. And then we have some more sophisticated operators that we talked to that have many of which have actually tried kind of building some of the stuff that we deploy for agencies. And they do it through a combination of different off the shelf platforms and automation backends like Zapier or Make. And those folks are coming and saying, look, we've gotten the AI to be able to make a call, but we can't really do anything with it. And that's, where the nuance and challenge comes in of having to really understand how to integrate it with your systems beyond what's available kind of off the shelf. But overall, I would say most people we talk to are classic use of like. very normal using chatGPT not just to like help them with business but also to analyze data and you know look for caregiver trends and spot new client marketing strategies. So you know most people are you know using it which is you know getting it's getting to the point where people are comfortable enough in their own daily use where they're saying like look I think I could actually start to put some of this kind of AI out in the front lines. A lot of our customers or even prospects leads we're talking to, they're really interested in AI, but very skeptical to put it in front of their clients, which we listened to that very early on where Phoebe right now is very focused on the agency to caregiver communication loop. We are starting to have many requests from clients to be like, look, this is awesome. I'm comfortable with Phoebe reaching out to my caregivers at this point. So like, let's bring in the client loop. But on average, I would say the kind of... crawl walk run strategy is like first I'm gonna use it internally, I'm getting comfortable, then I'm like, well maybe I'll have this start to help with some of my caregiver facing operations and then the, know, okay, I feel good with that, so I'm gonna start to bring it into the client side. Miriam Allred (16:04) What are some of the common misconceptions? Like you said, it still does surprise some people. What are people maybe wrong about when they come to you? Dave Dworschak (Phoebe) (16:15) You know, thing, I mean, the first thing that comes to mind is there is so much AI out there at this point. There are tons of consulting companies and development shops and, just general AI industry agnostic tools that are trying to figure out what markets they're going to sell into. And, both folks. especially folks coming in from outside of home care, home health, the healthcare ecosystem overall, they kind of come in with the promise of, can just plug the AI into the system and it's just gonna work and it's gonna do work while you sleep and it's gonna do things in the background and you don't have to really pay attention to it. So people come in, I would say skeptical overall of like, look, I haven't seen any AI that I would be comfortable just letting it run. So one thing that... is pretty commonly misunderstood coming into our calls is people don't realize there's a middle ground between like, do I have to just like give the AI my work versus like, I have to not use AI at all. And, one thing that we're trying to, be very heavily intentional about is like, there's a very healthy middle where you can come in and like start to layer AI into your operations, but still have full control over, what it's doing, how it's working, et cetera. like biggest, biggest thing I would say is like, it's not a, it's not a, question of like, do I do it or do I not do it? It's more of like, how can I like start to weave AI into my day-to-day operations so that I don't kind of fall behind? And if I look kind of back at... Maybe three months, six months ago when we were earlier on, ⁓ a lot of folks were more of the mindset of like, look, I'm not, I'm going to watch and see what the industry is going to do. I'm going to watch and see what other people's are doing. Like I understand that there's some value here, but like I'm not going to be on the early adopter side and having come from a. software background, you know, in the healthcare staffing industry where we were building software that was quite new and a different approach to the business where it took years, I think, for people to get from the curve of like the new thing we were doing to that becoming industry standard where we've actually gotten there much faster this time around. And I think that is because you've got the consumer layer of like chat, GPT and anthropic and Gemini, where people are actually seeing this make an impact on their personal life. So people are becoming much more willing, much faster, I think, than than previous periods. in history to adopt new technology. Miriam Allred (18:32) You talked a minute ago and I've seen this as well, people building out on their own because like you just said, there's a lot of just like consumer facing tools. You can start building your own GPTs. Like there's just so much available to the end user and a lot of innovative like forward thinking, maybe tech savvy owners are just out there creating, which I think is great. Address the HIPAA concern there though, because that is a pitfall that not enough people are thinking about. in healthcare, let alone home care. Dave Dworschak (Phoebe) (18:59) Yeah, it's quite complex from a HIPAA perspective to know what you're doing. And you should not just be dumping PHI and PII into chatGPS ⁓ People are doing that, unfortunately, but that is not the path you should be going. think complexity-wise from a security standpoint, it still takes qualified. security engineers to make sure that you're handling things correctly. It also means that if you're going to be working with, the foundational models like, you know, OpenAI and Anthropic and Gemini, which, every AI company like Phoebe is, working, with them and on top of their stack in some level, you have to have the kind of enterprise relationships, you have to have BAA signed, and then you have to make sure you're continuously monitoring, everything from a security infrastructure. And you just can't really get that as an indie builder and you can but it becomes quite expensive because you generally just like a couple examples you sign a BAA with kind of a phone service provider like Twilio or OpenAI you're committing to minimum spends that are frankly like unreasonable minimum spends for individual companies but when you're building software as a whole and you've got kind of many customers and are able to distribute the cost of things like that across many it becomes very cost effective so at this point the kind of security layer for the average consumer is possible, don't get me wrong, there are lots of people doing it the right way, but it's difficult and expensive. Miriam Allred (20:28) And I bring that up not as like a scare tactic, but it's just like, there's a lot of operators out there doing that. And I just like fear for that and what that can turn into. And maybe there's not a lot of like checks and balances in place for that today, but doesn't mean that there won't be very soon. And so I just bring that up as like a something that everybody needs to be aware of and be mindful because yeah, we're all just like dumping stuff in there and you know, where it goes and how that works is, you know, even you and I are like experts when it comes to like that compliance. But Dave Dworschak (Phoebe) (20:30) Right. Miriam Allred (20:56) It's just something that people need to be aware of at this early stage. I like what you said a minute ago about we're in this like mid ground stage, especially in home care where AI is a teammate and it fits inside of existing workflows. It's not a replacement. I think that is a common misconception partly because companies are coming on the scene and saying, hey, we can just automate everything end to end. But I don't think that's where we should be today. We are very much like in this mid ground where we're transitioning. And so today AI is best suited to like plug in to pieces of workflows. ⁓ Why is that? You know, like I guess there is the ability for like voice just to take over and do it all. But I guess like from your perspective and what you're seeing and hearing, why do we have to go through this transition period and why is it best suited to fit into existing workflows? Dave Dworschak (Phoebe) (21:49) Yeah, great questions and something we spend a lot of time talking to prospects and customers about just because they have often before they meet us, tried to been sold on what we consider a false promise of like, just let us plug this in over the next 60 days and automate out the scheduler. And we don't really see it that way for several reasons. From an industry perspective, which is maybe the most interesting, You know, there's data that lives inside of your systems, know, your HRs, your scheduling systems. ⁓ And then there's also the critical layer that lives in the heads of the schedulers and the coordinators and the owners and the people that are actually doing this work and chatting with the clients and caregivers every day. So, coming in and saying, Hey, we're just going to plug in and, leverage the data from your system to know exactly what to do. It's, it's, it's, we think a pretty big mistake just because there's so much critical, we'll say tribal knowledge about the business and the operations and the community of people on the caregiver and client side that, the AI is going to be communicating with that, that the systems just don't know. So, you know, having that say AI as a co-pilot perspective enables AI to, you know, use the data at its disposal to make some, say, initial recommendations on how to approach a situation. ⁓ revert back to the kind of human who's been doing this work for years to say, on or off am I on this list of recommendations? And then give them the ability to coach the... ⁓ the next step forward, both from like a now perspective, like, am I, who am I going to have the AI or Phoebe call right now? But then also, like, what can Phoebe learn for next time so that, the tribal knowledge that's in, one of seven coordinators heads, they don't have to go to that person next time they can, then kind of figure out what what did this person do last time by asking the AI? Yeah. Miriam Allred (23:28) really good points. This might be an impossible question, but my mind, you know me, I love to like quantify, you know, like how much lives in the system versus how much lives in the scheduler's heads. And that varies like every, again, every operation, every scheduler is different. But if you're talking to schedulers, like if you had to put numbers to it, how much of the scheduling function is actually documented in a software versus how much is actually just like ambiguously living in the heads of the people. Dave Dworschak (Phoebe) (24:01) Yeah, that's a great question. And not something I've spent a lot of time trying to quantify, but I'll make my best, just. qualitative guess at this based on kind of all of our customer use and you know all of the conversations Phoebe has is you know usually the initial recommendations that Will say are suggested by default say by the HR like at best. They're like 50 % of the way there Because you have number one Just stuff that's not getting documented even these you know schedulers and coordinators like even if they're a 10 out of 10 at documenting kind of in the preferences and profiles They just don't have time you're having so many conversations like you're calling a caregiver and caregivers saying like, you no, I can't take the shift, but also I'm going to be out of town next week. And it's like nice that the scheduler would go write that down. But the reality is, is like they have 35 more calls to make. ⁓ So number one, like even if you've got just great, you know, folks doing the documentation, a lot of that just doesn't ever make it to the system. But then even in the cases where it does make it to the system, it's very limited in the way that say the, the old way of like kind of preference matching and kind of storing a caregiver's availability and preferences or clients preferences. The old way is just like has a dog, likes dogs ⁓ versus the kind of new AI forward way, which is what currently lives in the heads of the schedulers, which is like, okay, they're okay with dogs, but they're not gonna go into a house with a dog that's over 50 pounds or under 50 pounds, depending upon kind of the preference of the caregiver. ⁓ And there's just really not a place for like existing like one-to-one matching technology in these like. existing or current scheduling systems to do anything with that data. So that's that layer that lives on top of it. And the scheduler side or the coordinator side is you have that base level information and then you have like what's actually important. And maybe one more quick example, know caregiver says they're willing to go in a five mile radius of their home address, but they actually won't go to this neighborhood or that neighborhood. And there's not really an easy way to match that the way that these like systems deal with matching. So that information maybe it lives in a free text field, maybe it lives in a note, but it largely lives in kind of the heads of the team where, we take that, AI as a teammate, not a replacement approach because, number one, we've, you know, we don't have that information at our disposal, but number two, like we can get that information at our disposal by working as a co-pilot alongside the scheduler and absorbing that information over time, not just from kind of the input the scheduler is giving, but also from all the conversations that Phoebe goes out and is having with caregivers day in and day out. Miriam Allred (26:27) Yeah, this is so good, Dave. I agree that I want to like kind of walk back my thinking on that a little bit is the ratio that like what lives in the software versus what lives in the head is not reflective of the quality of the scheduler because I would actually say if a, you know, executives look at their schedulers, the best schedulers likely have way too much in their heads, which makes them a good scheduler. Dave Dworschak (Phoebe) (26:43) understand. Right. Miriam Allred (26:55) And you said a minute ago, you even if they're 10 out of 10 in documentation, the volume is just so great. They can't get a hundred percent of the conversations of the documentation in the software. There's just no possible way. And, and so I actually think like the best schedulers likely have 70, 80 % in their head and, a good amount in the software as well. But like, there's just so many moving parts and so many conversations and so many nuances, like the likes dogs that the radius from their house, like Those are just things that they know about their people and the best schedulers just know everything about all of the people in their purview, which makes them great. so round about, I think this is why AI has to be this mid layer right now is because just operationally, we are still trying to figure out what lives in the software, what lives in the heads, where does everything live best and how does it all coexist? And then how does AI like split the gap in all of these different senses? Dave Dworschak (Phoebe) (27:52) Right, no, 100%. And it's so interesting to think about. maybe something we talked about earlier of like the pace of adoption and people who are maybe still sitting on the sidelines and you know, one benefit, even if you're just dipping your toe into the water is, just like a scheduler on day one when they start, they have very little baseline knowledge of, know, what goes into the clients and the caregivers and it takes them, you know, two months, three months, four months, five months, six months to really like really get to the point where they can just rattle off like, Hey, who should I call in this situation? the AI is the same way. the AI is learning based off of the interactions that's happening with the staff, with the schedulers, with the caregivers, eventually with the clients. So that early adoption for say AI as your co-pilot, AI as your teammate, it's not just helping your schedulers be more effective, but it's also taking that tribal knowledge layer that takes months to build up inside of a scheduler's head and then they go on and get a different job and all that knowledge walks out the door. You start using. this technology now and if you set it up the kind of approach that the approach that we're taking with it, it's you know the AI is from day one able to take that tribal knowledge layer and build it up over time and actually encode it into the software which is ⁓ really only possible for the first time like right now. Miriam Allred (29:08) Exactly, because everything happens through the phones, phone calls and texts. And put AI in that communication already, the time savings, the cost savings, the efficiency, is the AI is just absorbing every single communication that's happening, whereas that was not in existence several years ago. And that is just like the biggest unlock is just like AI is now a part of that communication. And there's way less manual documentation because it can just absorb all that information. aggregate it, place it, process it, systematize it. Like it's just so capable of helping with all of that documentation. ⁓ Let's talk more about these specific workflows of a scheduler. Like I wanna just kind of like unpack each of them a little bit more so you can paint in like what's literally possible. Let's start with just the outreach. And you guys have some incredible case studies already on your website. I was watching and reading and just like refreshing on those. Dave Dworschak (Phoebe) (29:46) Definitely. Miriam Allred (30:06) There's all these phone calls. Well, yeah, where should we start with the outreach specifically is like there's a missed shift or like a call out happening and there's like this window of time when you need to reach out to so many caregivers. what exactly is happening? Say a call out is like made known. What is the workflow for Phoebe and the office and the scheduler together? Dave Dworschak (Phoebe) (30:29) Yeah, great, great question. So, you know, the current way, you know, industry status quo, which is, you know, everyone's doing the best they can is you get a call out. Let's say you've got 90 minutes to two hours to fill the shift because that's all the notice you received. It's, on average with, what we're getting reported back from some of our customers, which we have some of this, you know, in one of our recent case studies, it's going to take a scheduler coordinator and seven minutes to make a call to one caregiver. And that's not just. ⁓ physically on the phone, like figuring out who to dial, dialing them, leaving voicemails, making notes in the kind of... scheduling system as a result of the call. And it's like seven minutes kind of on average from one customer we've done quite a lot of data tracking with. And then on average, you're calling maybe 15, 20 caregivers in order to get a shift covered. if you're talking seven minutes a call, 20 different caregivers, I've now put myself in a fast math situation, but we'll call that 140 minutes of calling. So with Phoebe, number one, Miriam Allred (31:22) Yeah. An hour and half to two hours. Yeah. Dave Dworschak (Phoebe) (31:33) Scheduler asked Phoebe for help, say, hey, Phoebe, I want you to help me with the shift. And Phoebe goes into that moment and goes into the EHR scheduling system, looks at the care plan, looks at the ADLs, IADLs, preferences, past visit information, past notes, ⁓ information about who's seen the client before. And we extract a knowledge base of everything that the. scheduler might actually be going in and reading and it takes them, five, 10 minutes to figure out like, okay, who am going to call? The AI goes in and extracts that information just in a few seconds and builds up a nice knowledge base of everything it needs to know about the shift. And then, based on the data in the system, we'll make that, list of recommended caregivers where your typical scheduler workflow might be like, okay, this person looks good. I'm going to spend seven minutes calling them. And then they hang up the phone, no result. Okay, now I'm going to spend a few minutes finding my next person and really upfront. Then Phoebe provides you with with that baseline set of recommendations gives the scheduler the ability to spend, I would say on average, a minute, two minutes, maybe three minutes, kind of finessing the list based on all sorts of ways the scheduler can search, sort, and filter through the information and the recommendations Phoebe's providing. And then rather than making those 140 minutes of phone calls. Phoebe just has that, you know, three to five minute conversation with every caregiver at the same time. So that scheduler then can kind of move on to the next fire for the next five minutes while they wait for Phoebe to do its thing. Phoebe will let them know like, you know, hey, Miriam, I found someone to take the shift. Click here, check out the results. And when you open that up, you can play back a recording of the call. We generate a one to two sentence summary of the conversation so the scheduler can see exactly what happened very quickly. And then we also provide the full transcript in case there's anything that kind of needs to go back through. With that, the scheduler is brought back into the loop. So scheduler directs Phoebe to do the outreach, of make sure that Phoebe's not reaching out to caregivers that it's gonna waste their time or it's kind of a bad batch to reach out to. Phoebe does the work, brings the results back, and then ultimately brings the decision back to the scheduler. Again, just because you might have two, three, four people who in two or three minutes ultimately end up saying, hey, I want the shift. We bring that decision back to the scheduler. Scheduler's the one who's gonna know who's the right person for that client. And when they approve them, Phoebe then does all the follow-up work where very often, like it's a scheduler's dream to, they get a shift filled. They let all the 14 other people they left voicemails know that the shift is now closed. But the reality is, they, even if they mean to or intend to, or it's part of the policy, it just doesn't happen most of the time. So with Phoebe, it's going to automatically do that once someone's been assigned to the shift. That way you're closing the loop with the caregivers, not having to field callbacks from folks about something that's no longer open. But then it's also going to really pull for more information from that caregiver about like well when else should we call you like this shift is open but what else do you want so that way we can start to layer that information and over time to the recommendations where you know day one we're really just basing the recommendations on what we know from the EHR or from the scheduling system and Day 30, day 60, day 90, it's getting smarter and smarter and smarter because we're able to layer in all the information that we're collecting from those conversations as well as the directives from the scheduler to improve the recommendations over time. like short recap of that is, right now it's an hour, two hours, 30 minutes, kind of depending on the situation to do all the outreach and Phoebe handles that in just a few minutes. And globally across our customers we're, we're you know, targeting of kind of five minute fill rate. We see it a little bit higher, higher than that now, but well under 10 minutes ⁓ on average to get a shift filled from basically time they call out to full replacement. Miriam Allred (35:12) Insane, insane. And it's live, it's real. It's like, this is not hypothetical. Like this is active live today. Agencies are using this. If you're not using it, you're already behind the eight ball. Lots of questions. My first question is, you talk about like the old way of matching, like this to that, this to that with client caregiver based off their profiles and preferences. Is there a layer of transparency for the scheduler to see into that initial like discovery? phase where you just initiate Phoebe, she goes in and she's just like reading and absorbing and like putting together her assessment of matches. there, is there transparency into that so that a scheduler again with all of their base like head knowledge, can they go in and like manipulate that stage? Dave Dworschak (Phoebe) (35:58) They can. Yeah. So we, we, we provide like clear insights into like exactly what the match criteria are. ⁓ you know, because we're basing it off of initially the information that's coming directly from the scheduling platform and the caregiver profiles and client profiles. It's, fairly straightforward, but it's interesting how nuanced it becomes over time because, know, imagine, ⁓ you're a scheduler and if you had all the time in the world, you would just build these like novels of information on a caregiver and a novel of information on the client. And you'd be able to like, every time you go to schedule a client, you just like, read through these, you know, two, three, four page bios of the caregivers to figure out like exactly who I should send. But that's not practical as a person. you know, Phoebe then over time, we're building up those like long form, like very kind of detailed qualitative, overviews of like what people want, when they want it, how they want it. And the AI can read through that very quickly. I'm sure most people listening have experienced this at this point where you can, you know, feed a document or what have you to chat to BT and something that would take you, you know, an hour to read or an hour to listen to. can get a summary back in. in 30 seconds and imagine layering that on top of all of your caregiver profiles at the same time when you're trying to find the match with the client. So we're transparent in the reasons we're making the recommendations. And then we also give the caregiver the ability to say like, okay, well, I see that, you know, this person was recommended because they live within three miles of the shift and it's a faster drive for them than, somebody who is two and a half miles away, but on the other side of the bridge. But then the scheduler knows like, well, despite the fact that this is faster for this person, I know for sure this person's not gonna go work on this neighborhood or this street. that's where that human layer comes into the matches as well. Miriam Allred (37:36) Exactly. And this is just validating what we were saying earlier is like there's still so much in the head and so you can't just rely on the EMR data. It has to be a combination with the scheduler in the loop involved in like manipulating all of that criteria and things. ⁓ Talk about like QA like quality assurance that this goes smoothly every single time because that is the concern of like a one off. You know it does well but like at scale every single call every single text every single match like what I guess checks and balances do you all have internally as a software company to make sure that this is effective and accurate like time and time again. Dave Dworschak (Phoebe) (38:13) Yeah, great question and a super important one. with, you know, quality on these calls, like a few things we're doing. Number one, like we set kind of the AI up to be like very task oriented. Like this is the set of information, you know, ⁓ this is like the set of rules and guardrails you're kind of bound to live by and act by. ⁓ And if, anything comes up outside of the number one, like your knowledge of the client or caregiver situation, or number two, outside of the, the job you were given to do, ⁓ we don't try to make it up. I think a lot of AI out there, the guardrails is quite, they're quite difficult to execute. So if you're not giving very specific kind of like lanes for the AI to stay in, like you could end up having, you know, the conversation go off in a totally different direction for 30 minutes and make some promises that you never should have made. ⁓ But with very tight guardrails, you know, the way we control QA is like, number one, we don't try to box people into conversations. So if something comes up, for example, caregiver asks a question, Phoebe doesn't know the answer. It's not going to try to kind of skip over it. persuade the person to take the shift anyways, it's gonna just bring the human back into the loop ⁓ and say like, look, I don't have an answer for that. I'm gonna bring so and so in from the agency, I'll get them on the phone and you can continue the conversation with them. So number one, like quality assurance mechanism is like, don't make things up, have tight guard rails and then bring the human into the phone and don't try to trap the person into an AI call that's gonna lead nowhere. I think that's... You we talked a bit earlier about misconceptions. ⁓ know, a lot of people come in and think like, robocall, like they're used to the typical like robocall where like you're just mashing zero over and over and over again to try to be human. Like I do that even right now. ⁓ So it's very different than that because like it's very, you know, task oriented and knowledge backed to say like, this is what I know. This is the task I'm trying to do. If it falls outside of the bound of that, I'm just going to revert back to the human layer that like still remains a critical component to be in the loop. ⁓ Miriam Allred (39:46) Heh. Yeah. Dave Dworschak (Phoebe) (40:07) And then, know, other QA mechanisms we actually provide for every single text, every single call. It's all visible in Phoebe at all times. So there's never, never operating in the dark. Even every voicemail is like transcribed, recorded, available to be listened back to by the company. And in those cases, there's a kind of interesting way to think about it, but like thumbs up, thumbs down rating for every single piece of communication that ever happens. So if there's ever something where Phoebe's done something that's not within the bounds of the agency's policy, policies or something that even they would like to do a little bit differently from a culture approach perspective. You just like say Phoebe like thumbs down, it'll pop open a box. You can explain why you didn't like that and then Phoebe will learn. So that's like we very heavily encourage folks to use that especially early on. We also were meeting with customers reviewing their data, reviewing the results, reviewing the outcomes, taking their direct feedback and kind of coaching or guiding or stylizing the way Phoebe's behaving. ultimately the transparency and then the guardrails are the keys. Miriam Allred (41:08) Yeah, and we've all been on the flip side of guardrails gone wrong or the lack thereof. And I think that's where people also run into that issue of building on their own. They can't quite perfect the guardrails, but that's what you all focus on is the guardrails. And I think that's smart of if it hits a wall or if it hits a point where it can't go further, it doesn't have the knowledge necessary, it's just default back to the scheduler, default back to the human. don't guess because AI will. It will literally just like, has a mind of its own and we've all experienced that, but it's telling it not to have a mind of its own when it gets to a certain point, like default to the human and that alone is just like so important. Talk about like data storage and memory. So you were saying a minute ago, ⁓ the beauty of the AI too is it's absorbing and learning like the communication. Dave Dworschak (Phoebe) (41:37) Yeah. Right. Miriam Allred (42:01) patterns of say the caregivers, like their response rate, their response time, the time of day that they're responding. Like there's all this memory that it's accumulating. How is the data stored and kind of explain how it is learning about the schedulers, but also the preferences and communication styles of say the caregivers as well. Dave Dworschak (Phoebe) (42:22) Yeah, it's this is like such a fun technical challenge for our team to be working on. have like some fantastic founding engineers and, you know, early team members that like live and breathe this stuff every day. So it's very fun to talk about. Phoebe. Imagine kind of the human workflow where like the average scheduler still is doing this. They maybe have a notebook or an Excel file or Google Sheets or kind of whatever their workflow is. And as they're like thinking about important things they want to remember about the caregiver, they're writing it down and kind of keeping, we'll say, a diary of each caregiver situation. You could think of that as like the base layer for this, where Phoebe is with every call, every text, it's basically looking for, OK, here are these like 50 kind of key preferences that are important to matching care. with clients and making great relationships. So it's constantly like anytime any of those things are coming up, we're extracting that and putting it into, we'll call it like the core memory of the caregiver. And you can almost think of it like it is not. technically stored this way. But if you just have like a piece of paper with everything that was important, always written down and every time something changed about what something somebody liked or wanted, you kind of scratch out the old thing and update it with the new thing. And then if you break that down a little bit further, ⁓ you've got the concept of, like long term memory. Like I had an incident with a dog that was very large. I'm no longer comfortable going into a house with dogs over 60 pounds, but I'll happily go into house with a small dog. You have that kind of long term, like qualitative, very detailed But like that's something that could be committed to that caregiver memory for forever and then you have more of the short-term things like Availability preferences like I am gonna be on vacation next week. So That's kind of committed to the short-term memory. So if you think about it like that piece of paper reference It's kind of at the top of the page It's the most important thing that needs to be seen based on the existing or current recommendations or current time frame and then you you also have the more behavioral layer which frankly today is very rarely looked at outside of, kind of the approach we're taking, you know, here at Phoebe is ⁓ it's Tuesday. ⁓ over the last 30 days, how many times did Dave pick up the phone on Tuesday? And does he actually text back faster than he calls? Does he sometimes text back faster if it's Saturday versus Wednesday? being able to actually take those response times and not just say like, what do people want and what do people like and what do people have experience and skills with, but like, what do people actually do? ⁓ And that's the very interesting layer that then gets brought into the kind of memory layer of the caregivers of like, I know that Dave is qualified. know that on paper he says he wants to work on Tuesday. I know that zero times in the last 90 days has he picked up a same day shift. So I'm probably not going to put Dave at the top of the list. Miriam Allred (45:06) And this is next level. Like this is where I almost get chills. Like we're so nerdy over this type of stuff. Like this is so good. ⁓ finding out the patterns, the communication patterns of like caregivers at scale because it's true. They're working long hours. Their response times are slow. Some people are phone based. Some people are text based. Like there's just so many nuances to the communication layer. And if AI can... understand that and then also do the outreach in the way that's most conducive to the individual. Like, it's our game to lose. Like that is how this industry evolves from what it is to where it's going. Dave Dworschak (Phoebe) (45:44) And it's admittedly it's it's A very complex thing to solve. I don't want to like paint the false promise of you come in on day one and day two, Phoebe just knows everything about your caregivers. And it's like acting totally different than it did on day one. It takes time to build up these memories. And even for us to like incorporate the memory layer in for new clients, it's not an instant process, but it's something that starts working on day one. And then more and more you work with it, the better and better it becomes. it's, it's a very interesting challenge. It's also changing over time. mean, just as a reference point for folks, if there's any say like chat to be T or, you know, Anthropic or Claude users out here, you can actually go into your settings and look at what memory the AI has on you. So it might give a good like reference point to kind of think about how we approach this. Miriam Allred (46:29) Yeah, and I personally like that word, like learning, like the way a human learns, the AI has to learn. It has to spend time with your people for 60, 90, 180 days to be able to get like an accurate reading of what that person is like. Like you just said, it's not overnight because it quite literally has to learn patterns and behaviors over a set period of time. But I think that's the cool thing with it. And that's the thing with home care, like you and I both. Dave Dworschak (Phoebe) (46:49) Right. Miriam Allred (46:53) Home care is home care, but every agency operates a little bit differently. Every set of caregivers, every set of clients, every office team, there's just little cultural nuances, regional-based nuances to the different people involved. And so AI has to just like quite literally spend time with your people to be best suited to help your people in your processes. Dave Dworschak (Phoebe) (47:15) 100 % and maybe like one more thing to tag on that is, know, there are things that like. schedulers today, you know, wish they had more time to spend with their caregivers so they could learn more and, you know, form better relationships, both on the caregiver and client side, also thinking about growth. And there are so many things that schedulers have to do that kind of prevent them from doing that. But then there are also tons of things schedulers, you know, and coordinators, they just don't do because it's very expensive for a human to sit there and call like a list of 200 people who used to work for you last year, but kind of fell off the radar and fell off the grid. It's very expensive to call them because most of them don't answer. And it's very expensive to try to figure out like, are their updated preferences? Are they looking for work? And you start to thinking about like, how do you bring kind of AI into this early on to learn more about your caregivers and learn more about your clients. And it's not just the learning there, but it's also the fact that, you know, we can have an AI or have Phoebe go out and reach out to people that you would never call because historically it was just too expensive to call them. And it's much more cost effective to do this at scale. So it's not, it's not just about like the Phoebe doing the things that you do today, but it's also about doing the things that you would do if it was less expensive or if you had more people to do the work. Miriam Allred (48:27) Great point, I'm glad you brought that up. Because the concern might be, Phoebe will learn what your people already know, which is here's like a team that always responds, that always picks up shifts, and that doesn't like help the capacity issue. But like you're saying, it will actually do reach out to the BCDEF team, you know, that's like non-existent or non-responsive or hasn't been active for months and months. It can go out. And that's the cool thing is like you can build workflows for this specifically. It's like, we haven't talked to this person in six months. They shouldn't be looped into like present day communication. They need like a certain set of communication tailored to them and their circumstances. And so it can be molded to all these different groups of people and built specifically for like their circumstances. Dave Dworschak (Phoebe) (49:14) 100%. Miriam Allred (49:16) This is so good. I want you to address just quickly, we spent a lot of time about outreach. Like that has a lot of different forms that it can take in a lot of different ways. But the other two that I just want you to like touch on briefly is like shift confirmation, like quite literally the AI just helping with shift confirmation. And then also miss clock ins and clock outs. When we talk about LTC and government payers, like the importance of shift confirmation and miss clock in and clock outs that alone is tedious and nobody wants to do that manually and we shouldn't be doing that manually anymore with AI. Just explain just high level kind of those two workflows and the unlocks. Dave Dworschak (Phoebe) (49:49) Yeah, so with SHIFT Confirmations, it's interesting because we're like 50-50 with our customers where like before Phoebe, they just didn't do it. They maybe would do it for holidays, weekends, you know. different situations, like the consistent day-to-day shift confirmations is not necessarily a pattern of every agency. But then we have agencies that are like literally have a person that does this all day every day before Phoebe and especially situations where you have that say heavy 24 seven, caregiver to caregiver handoff where it's just imperative that somebody shows up. So with the shift confirmations, Phoebe's going in and checking your schedule every day, automatically sending a text message to everybody about their shift or shifts for the following day. ⁓ Caregivers can engage with that, which makes it quite different from like a reminder that would come from like an EHR. That's just a one way thing that most caregivers just block the number. ⁓ It's quite different when it comes from Phoebe because Phoebe can say like, just want to check in, know, Dave, are you going to be at your shift in this neighborhood at this time tomorrow? ⁓ And caregiver can say like, yeah, but can you like remind me what the care plan is again? I haven't seen this client in while. ⁓ And like Phoebe can actually engage with that. And then on the agency side, like for the owner operator, you know, side of the conversation, the scheduler can Phoebe naturally collects the response to that. So it's not like this robotic, like you have to press one if you're going to be there or two if you're not going to be there. You just respond just like you would if you were kind of chatting with your scheduler and Phoebe will process that, ask questions if it has questions about the response you gave. ⁓ And then you have a nice clean dashboard that'll, you from a practical level, the scheduler might log into their Phoebe dashboard at noon every day after Phoebe's done a couple of outreaches to anyone who, know, number one, hopefully they respond on the first one. But if not, Phoebe follows up until it gets a response. So you might log in at noon every day and see, okay, I've got 150 caregivers on my schedule for tomorrow. These 115 have confirmed their shifts and it's just a nice clean like confirmation status all the way down the line. And these we'll call it 32 or skipped because Phoebe's gonna learn and not contact caregivers who see the same clients every day. If people have a consistent pattern of showing up, they have the same recurring schedule, those people aren't gonna get that kind of text every day because it's something that they're just doing day in and day out. can override that and have Phoebe do it for everybody, but most of the time the kind of skip list is utilized. ⁓ And then you might have these. Miriam Allred (52:08) I'm glad you're addressing that because that was going to be my question. Can it be toggled on and off for different individuals? Because yes, there's like the very consistent people that have been with you for years and it doesn't make sense. So it sounds like it can be customized by individual. Dave Dworschak (Phoebe) (52:14) Yeah. It can be customized by individual, also by schedule. We have clients who say we only want to do it for new employees Monday to Friday, but on the weekends we want to do it for everybody. We don't want to do it for even... new employees except for on holidays. So there's, all sorts of configurations you can do. And that's like a, you can real time just kind of say like this week I'm going to change my schedule or this week I'm going to change my cadence. And the end result of that is, very high compliance. We see, 90, 95, you know, hope, you know, oftentimes honestly, a hundred percent compliance with the process, as long as it's implemented well with the caregivers. And we work with you to make sure like the implementation plan is solid. Uh, and then, the flip side of that is it, it uncovers your issues. So if you have you know, three or four or five caregivers that respond back and say, look, I didn't know I was supposed to work. I can't be there or look, my shift, you told me it was at 10. Now this is saying it's at eight. So any of those issues that typically you wouldn't learn about until the person doesn't clock in or doesn't clock out, Phoebe surfacing those early. And then that rolls right into that scheduling assistant to help get a shift covered if you need it. ⁓ I can roll right into like quick explanation on the clock and clock out if you want, but. Miriam Allred (53:28) Yeah, do it. Yeah, go for it. Go for it. Dave Dworschak (Phoebe) (53:30) Yeah, so that one's also interesting, just like the shift confirmations. Phoebe can basically do the work and you obviously have the edge cases of like it let's say it used to take somebody half a day to sit there and like do the shift confirmations all day like Phoebe is eliminating we'll say 3.8 hours of that work like right out of the gate. It's the same thing with the missed clock ins and clock outs where typically agencies have you know somebody monitoring their inbox for notifications about somebody who didn't clock in didn't clock out you know with Phoebe by the time that notification hits because we're not waiting on the information to come back from the EHR Phoebe's going out and proactively like using the EHR to check to make make sure people are clocked in, checking the schedule, checking the system. And by the time you would get that notification, Phoebe's already texted the caregiver. If they don't pick up, Phoebe's gonna call the caregiver ⁓ and it can troubleshoot with them. And we do the same thing on the clock out, except for we call first, because oftentimes, right, you missed your clock out, caregiver's already driving, we're not gonna text them, because they're probably not gonna text back. So Phoebe will call on kind of the backend of the shift. And then depending upon... your pay or max, whether you're like all private pay or, doing, you know, waivers and Medicaid, or what your state UVV rules are. Phoebe can go into, your system and actually update the clock out record or clock in record. So for example, caregiver, you know, picks up the phone and says like, man, I'm already in the car. left five minutes ago. and you know, if it's an EVV situation where the geofence is critical Phoebe, you will direct them like, please turn around and go clock out. or if it's situation where, it's private pay, you don't need the EVV. record. ⁓ Phoebe can say like, no problem. I've documented the fact that we're talking right now. I have this, you on record and I will go into the system and just update your clock out record for you. So that spins off into all sorts of like interesting nuance use cases where people are now texting Phoebe like, hey, clock me in because I'm like, I don't have signal right now. All I have is the ability to send them SMS. I don't have any, ⁓ you know, any Wi Fi to send, especially in like rural situations. So we're getting into interesting use cases where you can literally just just like, hey Phoebe, clock me in. ⁓ And Phoebe will just do it. Miriam Allred (55:33) This is so cool. This is so, cool. we talk about home care being reactive. Like there's just a layer of home care that is and maybe always will be reactive. But this is how we get on the offensive. This is how we become proactive. It's through this type of technology that is real time and is keeping everybody up to date on their toes, forward thinking, like proactive rather than reactive. Like so much is reactive because there's just like volume and nuance and phone calls and documentation. We just like can't keep up. Like the margins are too thin. Like we just can't keep up. But this really, I think is how we're gonna like get on the offensive and how we break into just like a new playing field in home care. Quite literally, I think this is just like the start of an industry that will look very differently in just a couple of years. In our last couple of minutes, I wanna ask you like three like pretty pointed questions. One is I think a lot of Owners are excited. They're on board. They're interested. They've got the buzz like they want in The hang up is the office team and the nature of this office team is some of these administrators have been in health care for a long time ⁓ They might be a little tech averse and might be fearful of you know, rewriting their workflows how are you easing the burden and Mitigating like the change management with the office team Dave Dworschak (Phoebe) (56:55) Yeah, great question. I mean, if you're thinking about like the like individual, you know, execution layer for a scheduler coordinator, etc. like those folks like. There's obviously some general like, you know, what is this going to do to my job perspective? But once people start using it, they really realize like, we're not here to replace you. We're here to help your agency, like double your hours. And we don't want you to have to manage another employee. We want you to not have to make all the calls you don't want to have to make. like, if we think about like what AI is replacing, it's not replacing people. It's replacing constraints and removing constraints and giving you speed and giving you accuracy, giving you consistency, giving you documentation. So like when we're talking about like the hesitancy from office staff, whether it's from a, individual contributor level or whether you're talking about your leaderships, you know, we're not coming in and saying like AI is going to replace your leadership or your culture. It's going to give you more time to focus on things like that. And then, you know, from a actual like owner listening to this, like thinking about making a decision, ⁓ you know, or thinking about exploring this, like we maybe take a different approach than a lot of software companies. We... This is not our first run at building software and we are very interested in working with people who are getting value out of what we build ⁓ and that like frankly want to talk and hang out with us on calls like this. And if you're not happy, if you're not getting results, like there's no financial penalty to cancel. There's no long-term contract. You know, we want you to be here because you want to be here. And we just let the software, you know, think for itself. So if you're sitting on the sidelines thinking like, I can't make a commitment. Like, you know, if it's a bad decision, I'm going to be stuck in this contract. Like most software companies. that's probably true, but we really take the approach of like we want the software to speak for itself. If you're not happy, like ⁓ you're free to go. ⁓ No financial penalty and we'll actually let you take all the data with you. Miriam Allred (58:42) Yeah, I want to just add my two cents to that is like ⁓ I talked to a lot of voters. I feel like they want to like protect their office team and they almost put this. They almost put up this barrier themselves, I think, actually, because I agree when I talk to schedulers and administrators and you explain what this does to their job, it actually removes all of the manual burden that they're tired of doing. And so I think. I was curious what you were going to say to that because I've experienced that firsthand where I think the owner subconsciously is like protecting their team and is like, this is going to be a huge overhaul and I don't think my team is going to want it. And I think this is going to scare away my schedulers. I'm like, no, this is like, and then you put it in front of them and they're like, I can't imagine my life without this. Like they start using it and they're like, there is no going back. Like I've literally experienced that firsthand. it's like my two cents to the owners listening to this is like, don't put that barrier. on your, don't put that barrier up around your team, like expose them to this, get them excited about this. Like this is the future of their workflows. Expose it to them now early and let their minds like run on the possibilities of this. And then you build something with them and it's like revolutionizing of your operations, like quite literally. And so ⁓ I agree, like too many owners are yeah, like that first line of defense. And I get that they want to protect their team, but don't. Don't do that. Dave Dworschak (Phoebe) (1:00:10) 100 % and just like quick comment like, you we talk to owners all the time and they're like, I don't know how my scheduling team is gonna feel like I got, got to position this to them. And I'm always like, just can I talk to them? Let's have another call. Like, let me show them this. And I would, there's been very few situations where the next step after that scheduler call has not resulted in them using Phoebe just because they see it, they feel it. They're like, man, this is great. And you know, the... positioning we're taking to is like, Phoebe is very easy to learn. Like it sounds complex. AI is doing all these things for you, but there's not a lot of things to click. It's like pretty straightforward. And admittedly like, what I'll call it like self-driving software where you're not actually going through and clicking buttons and, know, creating things and approving things. Like you're just kind of chatting with Phoebe and then Phoebe's using the software for you. And admittedly we're like in the middle ground of that too, where like before we get to the kind of full self-driving kind of interface where you're able to just chat with Phoebe like a teammate. Like we've got to do some traditional software things, but at the end of the day, like our goal is like we come in, we use your tools, we use your software, we use your infrastructure. Phoebe learns your procedures, learns your policies, learns your culture, and we just work the way you work. ⁓ And you don't have to come in and have this heavy burden that typically with software is like I've got to like rip out this tool and replace it with that tool. Like we're not trying to rip and replace anything. We're just here to like help you use your tools better and make you click less things every day. You Miriam Allred (1:01:34) Add to, yeah, add to and supplement everything that's already happening. And we didn't even talk about like ratio, like a scheduler to roster ratio. But I think when we just talk about scale, it's just making every scheduler more efficient. You said it like, you know, one to two X, we're talking like maybe like three to four X, like just the volume that one scheduler is able to handle with this level of technology. And that's what we're all about is like efficiency and home care in every industry. It's just how do we maximize. and make every single person more efficient to cut costs and that's exactly what this can do. Dave Dworschak (Phoebe) (1:02:06) 100 % and you're not just like the schedulers not just say able to now handle 3,000 hours of care instead of 1,500 on a weekly basis But they're they're also doing it in a mechanism where like the 90 % of the phone calls they used to hate They don't have to do those anymore So it's quite nice for the scheduling perspective Miriam Allred (1:02:25) job satisfaction is going up and burnout is going down and turnover is the silent killer in this industry at the caregiver level and even in the office level. And so if you're reducing that burnout, reducing that turnover and reducing those turnover costs, like it's a win for everybody. Dave, I'm so excited for you guys. I'm just this talking about this, just lights me up. I'm excited for you guys. I'm excited for the agencies that are adopting this. I'm excited for like where this industry is going. Again, I just get like chill thinking about this because we're at like the brink of of the beginning of what this is about to do to this industry. And I don't say that lightly. I genuinely mean that. That when you start experiencing this technology firsthand and seeing what it can do, your customers have said it, the agencies I work with have said it, it's just like, there's no turning back. Once you're exposed to this, there is no going back. so ⁓ thank you for joining me in the lab. guess my last question is just best place for people to learn more about Phoebe. What's the website, your email, or how do want people to contact you? I know you're booked out. And I know there maybe is a wait list almost at this point, but how can people reach out to you and get in contact? Dave Dworschak (Phoebe) (1:03:27) Yeah, so you can find us online. bit of an odd domain, but phoebe.work. So it's easy to remember, but like phoebe.work is where you can find us online. ⁓ You can email me also dave@phoebe.work So pretty straightforward there. We have a nice like easy process for you to like get it, on our calendar on the website. There's case studies, there's videos, there's a ton of information for you, but ⁓ directly reach out, you know, dave@phoebe.work or you can head to the website, fill out the form. And you know, if you do that, we kind of based on the information you're putting in the forum, like ⁓ it's a pretty clear, can we get you implemented right now or would you be on kind of more of a wait list for as we expand to, different EHRs and scheduling systems, but ⁓ reach out. We are all humans despite the fact that we're an AI company. If you're reaching out to our company, there's zero AI layer between me and you. So if you're getting an email back from me or from Justin or from anybody on our team, ⁓ it is us doing that. Miriam Allred (1:04:27) Good disclaimer there. Dave is not an AI. I'm talking to a real human across from me. So Dave, this has been awesome. Dave Dworschak (Phoebe) (1:04:32) The video AI stuff is getting pretty crazy. I don't know how many years away we are, but at some point I might not be here talking to you. Miriam Allred (1:04:40) Yeah, we're not there yet. I still have a job today. People say that, what about AI on your podcast? I'm like, I mean, we're a ways out, but we're real humans having this conversation, which is what makes it so good. So Dave, thanks again for joining me. This is super exciting, onward and upward for both of us. We'll stay in touch. Thanks everyone for listening. We'll go ahead and wrap here. Dave Dworschak (Phoebe) (1:04:57) Yeah. Thanks, Miriam.