The AI Front Desk: Patient Access, Escalation, and Practice Readiness with Jakub Grajcar

Patient access problems are often invisible until calls go unanswered, staff become overwhelmed, and patients start looking elsewhere for care.

Chris Hutchins talks with Jakub Grajcar, Head of Marketing at Talkie.ai and host of Scaling Practice Management, about where an AI medical receptionist can help and where human judgment must take over. Their conversation covers routine scheduling and appointment questions, after-hours coverage, escalation pathways, EHR integration, least-privilege access, and the operational preparation required before a voice agent goes live.

Jakub also explains what practice leaders should ask during a vendor demonstration, how to tell whether automation is actually removing work instead of moving it somewhere else, and why clear communication with both staff and patients matters during implementation.

In this episode:

• The patient calls practices may not realize they missed
• When an AI receptionist should resolve, route, or escalate
• After-hours coverage and emergency boundaries
• EHR integration, privacy, and access controls
• How to evaluate a voice AI demonstration
• Preparing staff and patients for a new front-door experience

Connect with Jakub and Talkie.ai:

• Talkie.ai: https://talkie.ai/
• Jakub on LinkedIn: https://www.linkedin.com/in/jakub-grajcar/
• Scaling Practice Management: https://talkie.ai/podcast-scaling-practice-management/

Jakub’s current role and podcast are confirmed on Talkie.ai’s official site. Scaling Practice Management.

Transcript

Copy Transcript Copied

SPEAKER_01 0:00

Welcome back to the Signal Room. I'm Chris Hush. Quick thing before we start. I wrote a book. It's called Beneath the Signal, and it's about the human work behind trusted data and responsible AI and healthcare. If that's your job, go search Beneath the Signal on Amazon now and pick up your copy. Okay, here's our episode.

SPEAKER_02 0:16

About a week ago, I was the guest on my guest show. Had a fantastic conversation, and I'm very excited getting to know my new friend Jakob Drakechar. He's been doing some really interesting things in the outpatient world. Jakob is the head of marketing at Take AI, and he builds conversational voice agents for medical practices, amongst other things. And we're going to delve into all things outpatient care and coordination and scheduling, back office operations and agents, not the kind that you are hoping that will book you on the light shows, but much more importantly, to help you get the appointment that you need to see your providers. Jakum, so thank you so much for agreeing to come on the show. And I'm really excited for folks to get to know you a little bit this morning.

SPEAKER_00 0:59

Thank you. Thank you, Chris. I'm so stoked to be on the on the other side of the equation, as you mentioned. Typically I host Talkie's show, Scaling Practice Management, but I'm really happy to be a guest as well. Really enjoyed our conversation previously. And yeah, super excited to talk, AI receptionists, front desk, and all that good stuff today.

SPEAKER_02 1:16

Fantastic. Before we get into the the real challenging and rewarding work that you're doing, tell me a little bit about your background, uh really who you are and what Talkie AI actually does. Right.

SPEAKER_00 1:26

So my name is Eku Greizar. I'm the head of marketing at Talkie AI. I have been with them for almost two years now. Uh so joined in, I believe it was late 2024. I'm head of marketing for them, been a marketer for most all of my career, and in tech for all of my career in some way, shape, or form. I started at a software development company. I was this, you know, just one-man army marketer, and then it turned into uh a content marketing team of 10 quite quickly. So that was a very interesting first chapter of my career. Then at some point I switched to a company that was doing click-up implementation, so in the project management world. That company has gone interesting places in terms of their use of AI. Uh, but that's you know, uh that's a little bit on the on the margins of this conversation. And now I'm at talkie. At talkie, we help over 300,000 unique patients every month with uh conversations that our AI receptionists are having with them. So that's scheduling conversations, they want to refill the prescription, uh, they want to cancel, they want to reschedule, they just want to ask about their appointment. Our talkie agents uh help with all of that, and we help over 300 practices already and scaling all the time. So it's uh it's a really exciting space to be in, definitely, you know, healthcare being one, and then also layer AI on top. No week is boring, honestly. So many announcements all the time. The models are getting upgraded all the time, so it's it's definitely been a trip, but a pleasant one. Love my team and really, really enjoy that I get to talk about this every day.

SPEAKER_02 2:53

I'm really excited about it. I think the interesting scenario that we are seeing across the board because a lot of shifts that have gone on, the pandemic certainly had an impact on it. But you know, we we have a serious shortage of resources, certainly in the nursing space, but increasingly in the general practitioner space, uh the cost of going to school, amongst other things, the constant stress and the gaps that you are currently focused on, these things have been opened up over a number of years, and we're really having a hard time with patient access. So we've spent a lot of money in this space over the years, but almost nothing when it comes to dealing with the phone call that's actually going to get a patient in to be seen. From where you sit, what is the patient access, the phones, the scheduling, the call center? Why is that the most under-discussed failure in really the whole system? It's it really tends to be a signal quite a failure.

SPEAKER_00 3:43

One reason might be that it's to some extent invisible, right? So, you know, you've got there isn't a very strong difference between your, you know, your phone is ringing off the hook and you're handling as many conversations as you have, and then you don't even know, some practices don't even know that while all the lines are busy, you know, 30 other patients, 50 other patients were trying to reach you at the same time, right? If it's the Monday peak or if it's right after a holiday. So I think there's a lot of uh occupation with trying to address you know the conversations that are happening, that the ones that are falling by the wayside, and you know, some of the statistics show that this might be even 50% of calls that are being missed, right? This is you know, as this is just accepted as a status quo. And we're trying to change that, right? We're trying to say that, hey, you can pick up all of the calls, maybe not in person, but with the technology that's currently available, you can at least have the AI receptionist pick up the call, depending on what the issue is. You know, the worst case scenario is we'll have the you know a transcript of the call, a summary of the conversation, and you can follow up on it. And in many cases, the scheduling conversation, rescheduling conversation, it can just happen end to end and keep your keep your front desk uh, you know, allow them to talk with the patient who's in the room, uh, who's uh tr who's at the front desk physically and trying to get help, right? So because at the end of the day, we're trying to unburden the medical teams and we're trying to address this problem. As you mentioned, there might not be enough attention on it, but definitely a company like ours has our full attention on it right now.

SPEAKER_02 5:09

So you mentioned over 300,000 conversations with patients on a monthly basis you're doing. What does that volume let you see about how patients actually try to get care that a single practice administrator would never see from inside the office?

SPEAKER_00 5:23

Yeah, so I think what's interesting is that a lot of these conversations, because typically when we talk with a practice, right, and they're looking at introducing an AI receptionist, their mind is drawn to the most complex conversations, you know? What if it's uh what if it's twins, for example, right? And you can't actually disambiguate uh between one patient and another based on their date of birth, right? They pull out all of these different edge cases of what's happening in the conversations. At the end of the day, though, a lot of the calls coming in are really, really routine and repetitive. You know, do you have parking? Can I pay by credit card? Could you remind me about my appointment? Especially people, you know, uh older patients, right? They might not be able to look that up easily on the patient portal, so they just call and they just want to get that information. And as much as we would like all of these conversations to be with a human and nice and chill and you know, with uh with a level of empathy and patience that every patient deserves, that's not currently possible with the with the amount of burden that's on the system. So we're taking those conversations. Just to be clear, we can take a scheduling conversation too, and even complex scheduling conversations. But I just think, you know, if you look at the data, such a huge chunk of these conversations are repetitive and really ripe for automation. And that's what we're seeing in the data, you know, as we're analyzing the, as you mentioned, 300,000 unique patients that speak with uh with talkie every month.

SPEAKER_02 6:43

Right. So I think that one of the questions that I have probably a lot of listeners would have talk about where this is really effective. So in a front office environment, it tends to be pretty active, and I would say probably there's a bit of noise. Sometimes there's even an element of chaos. But I'm curious to hear about what your experience has been in terms of when you're implementing things, what are some things that actually really causes more friction? And then how do you actually address that with your AI? Where does an AI voice agent genuinely earnest keep? And where does it kind of quietly break?

SPEAKER_00 7:16

Great question. So I think one of the big impacts that talkie has, and it it makes Front desk teams a little bit nervous at first, is that there's a few ways to implement it, right? But one of the ways to implement talkie is as the first line of defense. And if it's the first line of defense, then to a large extent, the phone stops ringing. And that's something that pay that practices actually get need to get used to, right? The calls get picked up by talkie first, and then they get kind of triaged at that stage, right? If it's something that the AI receptionist can resolve end-to-end, it just does, you know. I want to schedule a follow-up appointment, or I want to get some details about my my upcoming visit. That happens without the phone physically ever ringing. And if the phone actually does ring, that means that something got transferred to the human team. It seems to be a patient who is in distress, it seems to be an emergency, it seems to be, you know, according to the AI's assessment. A more complex case where, you know, if the AI doesn't have yet the answer to that particular scenario, it doesn't try to guess, it doesn't hallucinate, it just passes it off to the to the human front desk. So I think that is why that is part of what changes and where the AI, as you mentioned, earns its keep. I think another part of it is I was speaking with one of our clients. That was actually one of the first client interviews I did as I joined Talkie. And what she said is, you know, even if there are conversations that need to be followed up on, right? So, you know, let's say a patient just leaves a message with the talkie receptionist saying that I would like the doctor to call me because I have some questions, right? And there might be, you know, dozens of these conversations in the system, but at least you can view them, you can see the transcript, you can play the conversation back, you have an AI summary of these conversations, and you can get back to them in a prioritized manner. The front desk team can take that as a sort of backlog, if you will, and there's some sort of information about what these calls were about, instead of, you know, patient X from phone number Y just called at that, you know, at that time of day. Maybe they called three times, but did they call three times because it's it was a real emergency and we need to call them back AACP? Or are they just impatient and they called three times because you know they had the time and the will to do it? So I think that additional visibility on top of automating some of the processes end-to-end, that's another impact and another way that talk earns its key.

SPEAKER_02 9:31

Yeah, it's an interesting challenge that I think one of the areas that I was working on prior to launching my own business was within within a health system that's largely in the raw parts of the country. And you know, when the phone rings at the emergency room or whatever line you've actually set up to deal with transfers, you really have a very narrow window to be able to help them because the next available location from a hospital standpoint may be two hours away. So you really have to have a good handle on what your capacity is. You have to really balance the scheduling aspects of it yourself in terms of understanding what the nature of the practice actually is from an operational standpoint, but then also the ability to see with reasonably good uh visibility what is likely to be the next available bed. How long is it going to be before you do have a bed available? I've thought for the longest time. It's really good that healthcare is not trying to run a hotel because we're not great. Measuring capacity proactively. So it gets us to a point of trust, I think. Because we all in general conversations, the first few seconds in any conversation really set the tone for us in terms of are we comfortable? Do we feel like we're being heard? All those things. So talk a little bit about this and how you guys have worked through this to really address an LA feels right at the very beginning of a contract, excuse me, contact.

SPEAKER_00 10:51

Yeah, yeah. That greeting is really, really important. And actually that's you know part of how we almost enter into a little bit of a consultative role for the practices practices that we work with. Because when you're able to actually set that uh, you know, in a in a fashion where it's consistent, that the AI receptionist is going to consistently greet the patients uh in a given way, then you get to ask yourself these questions. Okay, so what is the perfect greeting like? What does what what should it entail? One important piece of it in the case of an AI receptionist is that it should introduce itself as AI. You know, that's important. In you know, in some areas it's also a compliance uh requirement, right? So typically our AI would introduce uh itself as this is Jessica, the AI assistant. Right after that, it would say, if this is an emergency, please dial 911 to make sure that actual emergencies that mean immediate attention, then it redirects that to the 911 line. After that, it might say, I can help you with with uh scheduling an appointment, with with a prescription refill, just directing them, explaining to them also, because that's something that's not always clear with this kind of new technology, how broad is the uh you know the selection of different uh things you can get done with talkie, different issues that you can actually get done through the AI. So that's all part of the greeting. Of course, we use you know the newest voice models to make it sound warm and you know, like it's talking with a smile, etc., uh, to put them at ease. Of course, you know, there's different kinds of reactions, but we try to make it as positive as possible at this uh beginning of of the conversation, while also making it clear that this is an AI, this is not a replacement for the 911 line, etc. So that's that's more or less how it looks. And then some practices like to tweak it to their particular uh needs. You know, maybe they have uh a different number for for the med spot that they have, you know, uh associated with their primary care facility. So we do also customize it per practice, which I think is really important part of the equation.

SPEAKER_02 12:46

Yeah, so you're not waiting for a specific point in the conversation, you're leaving with the introduction. I think that's probably the thing that I hear most often that isn't happening. You know, you're halfway through, and at some point there's something evident where if you deviate from the expected path during the conversation, it just keeps going full steam ahead.

SPEAKER_03 13:07

Yeah.

SPEAKER_02 13:07

Like we didn't say anything. Like if you mentioned the weather, for example. It's not not really designed to handle the weather. Talk a little bit about how you address some of those anomalies that might occur in that very first interaction.

SPEAKER_00 13:18

Yeah. So there's a set of guardrails within the talkie product, and the day the way this would manifest typically is if anything is mentioned that doesn't fit the the types of scenarios that talkie is uh is programmed for, or you know, what's what's predicted as possible scenarios in the prompts, then it might try to redirect you, initially saying, you know, I'm here to to help you with and reiterate, right? With scheduling, with questions about your appointments, if you want to reschedule, if you want to, if you want to ask some questions, I can help you with that. Uh for anything else that's not, you know, that's not the right line. If it keeps happening, then it would just redirect, right? Again, if there is if the AI hits on something that it can't yet help with, and honestly, that you know, that area is getting smaller and smaller every day, but it's never going to be zero, right? Uh then it redirects to the to the human staff to still make sure that the patient is is supported and their issue is is resolved. So that would be no. And it and it's important also to have an architecture of you know sub-agents within there, right? So uh so that the particular parts of the task of scheduling, which really breaks down into you know a series of subtasks, like making sure that we get the right uh appointment, right? We negoti negotiate the the right time and date for the appointment versus the availability of the doctors, etc. So we break it down, but the guardrails are you know across the product to make sure that we don't devolve into providing a cheesecake recipe or talking about the weather or whichever other conversations. Because I mean, yeah, people do do that with AI products. They speak very differently sometimes with an AI than than they would with a human. We predict this is going to change as time goes on, and the conversation is going to be so natural that people aren't going to, you know, try to do these tricks of, hey, can I get it to do something that's not kind of predicted in the programming? And it'll become much more of a part of daily life, you know?

SPEAKER_02 15:03

You you touched on what I consider an escalation pathway, uh picking up on the patient that's calling in appears to be confused or frightened. Uh where are those trigger points where the AI knows I need to hand this off to a human being? And how do you make sure that the system generally recognizes that there's a need to do that versus continuing to try to optimize the a solution that really the only answer is to have a human being in the loop?

SPEAKER_00 15:30

Yeah, I think the most important overarching philosophy philosophy here is to just err on the side of caution, right? If there are any indications that this might be a conversation that requires a human, that isn't a typical scheduling conversation. If severe pain is mentioned to some extent or if there's any confusion and that gets uh gets routed. I guess what I would emphasize here is that we do know that it's better for the phone to you know to actually ring and to for the human to actually pick it up and be able to help, than for us to show you know superior automation rates in in our product, right? Because at the end of the day, this is uh this is a sensitive industry that we're in. Uh some of the situations uh can be even life or death, right? So it's very important to uh to route that. And in terms of what would trigger that sort of escalation, these are the typical things that would also trigger it on a you know on a human side if if there was a human on the line. So like I mentioned uh dimensions of of severe pain, or if there just isn't a scenario that this uh this conversation fits into, it's better to err on the side of caution and to send it to the to the human front desk, who are freed up to actually handle these conversations with the appropriate you know, time and care and empathy and patience because uh the routine conversations and some of the more complex ones are taken over by the AI receptionists.

SPEAKER_02 16:47

Yeah. One of the interesting things is the ability to not only have the initial dialogue, but they're also reading and writing into the records. We have this concept of privileged access historically in the healthcare sector, and particularly when we're talking about access to data of any kind, but the concept of least privilege mean in this context when there is an agent has the ability to write to a patient's chart or their appointment.

SPEAKER_00 17:13

Yeah, I think you know, within this context, the most important thing for a product like ours is to basically stay in our lane, right? So the most important data that we take from from the EHR and write into the EHR actually isn't the most sensitive type of data you would you would imagine, right? It's the doctor's availability, it's the patient's uh you know, preference for an appointment, and even then that's just within the the the context of one of one conversation, right? Uh so we only use the privileges that we absolutely need from the within the APIs of the EHRs that we that we integrate with. So, you know, don't take any data wholesale, of course, is you know, is is one of the one of the guidelines. So, I mean, as you mentioned, the the most important pieces of information, of course, there's there's personally identifiable information in there. There has to be for the conversation to work, right? Name, surname, date of birth, right? Uh some of these uh we asked about the insurance because depending on the insurance, the routing of the of the appointments is is different, right? But we always ask ourselves like what's the what's the minimum of information that we need for a particular type of conversation because we take that security part very seriously. We get audited on this, we're SOC2 type two certified as well. So uh we have to do a lot of uh confessing about the security standards that that we follow. So yeah, for any product in our space, I think that's extremely important.

SPEAKER_02 18:34

You you just touched on the PHI, PII thing. And I'm curious, I'm sure you have clients that have on-call kinds of support for emerging care or or other things. Talk a little bit about how you can manage the after hours kind of liabilities that can come when agents actually operating and at two o'clock in the morning, there's no one in the staff that's actually present. How do you deal with that and talk about the accountability factor? If it gets something wrong, what does governance actually look like in that scenario? And what does a practice actually need to have in place before any of this goes live?

SPEAKER_00 19:07

Yeah, okay. So that's a few questions in one, right? But I'm glad you're mentioning the the after hours scenario because honestly, this is often where we start with uh with talk implementations, right? Because many practices just uh come to the conclusion that, okay, you know, during the day we can sort of manage, or you know, we'll we'll sort of manage for now, but we have nobody and nothing in place for the after-hours calls or you know, the weekend calls when we're just closed, we're out of business, right? So the talk agent starts there. I had a conversation with a client, actually, he shared this on one of our webinars, where the on-call provider, in this case, the on-call provider was available, right? They were starting to get conversations that they really shouldn't be getting because people were you know calling after hours, getting to the on-call provider that's supposed to be there for medical emergencies and trying to schedule an appointment, right? Because they, you know, they got through either way. So you might as well help me with this. And that's not what we want to be happening, right? We don't want the on-call provider to be having those conversations. So that's so that's an important layer that we add here, I think, yeah of you know being able to intercept these conversations, actually get the scheduling done in those cases through AI, and the on-call provider can can remain available for for for these uh these sorts of conversations. You know, and if there actually isn't anyone to to take the call, again, we do mention 911, right, every time that the conversation starts for those reasons, because some of those cases should go just just to know to 911. And in the other cases, then you know, still better than leaving just uh you know just a raw voicemail, we take the message of the patient through the AI and we leave it in the in the EHR uh so that it's it's easily viewable. Within the HR, you can see kind of a summary of what the m of the message that was left. One important factor though is outside of the EHR, we do believe in close EHR integration, but there is also a talking platform you can access for kind of deeper insight. You can see a full transcript of the conversation and a recording of the conversation. So if you want to even you know check the you know the tone of voice that was used in the conversation, you can also do that. So we try to provide as much transparency as possible, and you know, that's that's the way that we handle those sorts of situations. I hope that answers the question.

SPEAKER_02 21:14

Yeah, it does. I think that's one of the things that's always brought up whenever I've had conversations with folks around that the call center kind of capability and dealing with the off-peak stuff. Because emergency medicine, for example, their skeleton crew during the off-peak hours, and the types of things that are going to occur in those hours is pretty unpredictable. So you've got to have the technology that's actually able to discern things in a way that it's hitting the escalation at the earliest point that it needs to be hit. I think it's remarkable. One of the things we've talked about historically, and I think you mentioned this in our recent conversations as well, is the litmus test that we apply to what solutions we want to use. And it really comes down to is what we're about to implement going to give time. Back, or is it adding more time?

SPEAKER_03 22:02

Yeah.

SPEAKER_02 22:02

I'd love to hear your perspective. Does this actually mitigate some of that? Or is it just repositioning it, moving the work into another bucket for someone else to deal with? How do you guys think about that? I'd love to know what your approach to that scenario is.

SPEAKER_00 22:16

Exactly. My life changed the day I went on LinkedIn and I saw this quote. And I wish I could attribute it, but that got lost to time in my memory that AI should remove work and not just move work around. And that's kind of the assumption that we've had. To me, this comes back to the EHRs, actually, you know, the direct integration with the EHRs because some similar products in this space are not directly integrated. And the best they can actually do for you, and you know, and a lot of kind of homebrew AIs also start like that, is they they drop it in a spreadsheet, right? And you can then go into that spreadsheet, move that data, or start building some sort of automation to move that data from the spreadsheet, then into your EHR, right? But it's a lot of manual work, and to me, that's moving work, right? That's not removing work, which is why we've always focused on the patient crew, sorry, the practice crew, the the front desk team staying within the EHR. And the work that Talkie does shows up in the EHR, clearly mark that it was performed by Talkie. So, for example, I was recording a product demo just an hour before this, right? Had the conversation, I was pretending to be a patient and I wanted to schedule a follow-up appointment. Then I was showing it on this on screen, like in Athena 1. You could see it on the calendar that it was booked, but it also had this adnotation, it had this note with it that said, this was scheduled by the talkie AI agent, right? And if interested, you can go back to the talkie platform, as I mentioned, for deeper insight to actually view that full conversation. But the most important but we didn't want to you know pollute the EHR with that full context, you know, dump the transcript in there. That's too much for a busy practice team member to go through. What's important though is that if it's a scheduling conversation, we automate that end-to-end. If the patient calls, books an appointment, it shows up in the schedule right after the conversation ends, and that's really important for the practice teams to actually remove that work. That conversation does not need a human right now. And I and I wish actually more practices knew that this is actually fully end-to-end automatable already today, because I don't think all practices or enough practices are aware that this is where we are right now, and it's only going forward, and as I mentioned, the pace that it's going forward at for our product and many others is actually quite quite crazy.

SPEAKER_02 24:28

So I want to delve into the monitoring and the types of metrics that you are looking for that indicate that you've achieved the objective. Essentially, what is it that you're looking for that would indicate that you're giving people back valuable time during their daytime shifts?

SPEAKER_00 24:43

Yeah. I think uh a practice talking with a vendor like ours, if we want to, if you want to put our feet to the fire, you should ask about the automation rate. You should be asking all the vendors in this in this space about their automation rates. Workflow by workflow, this might be a different automation rate. You know, you could see a much higher automation rate for simple conversations like canceling an appointment versus more complex conversations like scheduling an appointment. Because in terms of canceling, you don't have to talk about a replacement appointment slot. You just have to say that you're canceling, right? And that's that's it, that's the end of the conversation, right? So I think that's one of the most important metrics of like, okay, of the X amount of conversations that come in that pertain to canceling or requesting information or uh refilling a prescription, what percentage of those are you actually able to automate end-to-end? Of course, for all the rest, you still get the benefit of, you know, the call gets initially picked up, it gets triaged, it gets transferred as needed, right? Uh but you know, the full to full end-to-end to automation rate is what we're focusing on very much as we improve our product because we think this is this is where the real value lies for a product such as ours.

SPEAKER_02 25:48

I want to pivot just a little bit to your what your role is doing, because I think it's important.

SPEAKER_00 25:54

Okay.

SPEAKER_02 25:54

Since you're marketing this capability for a living, it makes you unusually well placed to answer what you know this kind of question, I think. What should a practice ask a voice AI vendor to demo that the demo itself might be designed to keep them from asking? What should they be thinking about and what should they ask the person who's delivering the demo? Because oftentimes the demos are designed in a way to almost keep them from asking the tough questions. I'm sure you've seen this. I've certainly seen it as well.

SPEAKER_00 26:22

So, yes, of course we want to show the happy path, right? The conversation in which the uh the appointment gets scheduled and it's you know it's fairly straightforward and you know the job gets done. Of course, and you know, uh we also communicate about that on our website, but I think talking with a vendor similar to ours, you want to ask about you know the the edge case or corner case scenarios, right? So, yeah, as I mentioned, ask us about twins. We've done we've done twins a few times. Yeah. I mentioned them because you know, some of the I mentioned the disambiguation by you know by date of birth, you know, one of the ways of patient verification. And then it turns out that you know there's a ton of twins in the world, and we need to find other ways to disambiguate, right? So I think what's the what the best approach would be to actually bring honestly your full complexity of the scheduling of you know the different kinds and types of cases that you that you get and and bring them to us and ask about them, you know, just just ask about a more complex scenario. That's what these demo conversations or you know free consultations are for, right? To actually see how this works for your practice. And at this point, I would love to also mention that you know there's various kinds of approaches, and I think it's really important for our product such as ours. This is our approach, of course, not the only one that's possible, right? Is to actually implement the product in a consultative kind of way. So I call it uh collaborative continuous customization. At least I tried to call it that on our website, and then people told me it's too complicated. And now I barely said it, and I guess they might have a little bit of uh of right in that. Uh, but what I mean by it though is that you know there's a fair bit of ramp up. You know, we could launch the product after just a couple of weeks, probably feasibly once we hook into your system, but it pays off to actually have the conversations and get the complexity out of the head of your team and actually into the product, right? So if Dr. X never takes the follow-up visits on Tuesdays, right? And that's something that the team just knows, but it's not reflected in any way in your EHR, well, you know, the talkie AI is a member of that team. So it should also know about that. And if there isn't, if it isn't written down in any way, then we need to write it down. And there's a ton of those types of questions, you know, gaps in between appointment slots, for example, all of this complexities of scheduling that actually need to get written down. It's almost like a service in its own right, where at least we help the practices actually dive into the complexity of their of their scheduling and the way their front desk works as they're setting up the tool, right? And I think that's really, really important. Because for some other vendors, they do have more of their own workflow that's you know very polished and very effective, but then the practice has to kind of adapt to that and change the way they work to that. And to me, that's that that again is kind of moving work instead of removing work, right? It's uh a bigger lift to actually adapt and actually customize to each practice. But I think you know, in the long run, this will help AI adoption just in general. You know, when companies actually sit down with the practice with the team, dive deep into it, customize the product, maybe even build out a little bit of new functionality, but to make sure that for for your practice, the only practice that you will care about, the product works as close to perfect as possible. So I think that's that's a really, really important aspect of this of this whole scenario.

SPEAKER_02 29:36

Right. In terms of how a practice can prepare, I want to delete that just a little bit, because I think oftentimes there's pressure to move, and I'm sure you see virtually every kind of uh scenario possible in terms of the readiness that may or may not be there for practice to start working on an implementation with you and your colleagues. So I think there's this black box kind of mentality sometimes, and there's this hope and belief that's that's a magic wand of some sort that all of a sudden your data is clean and ready for prime time. What are some of the things that practices should be thinking about right now in terms of positioning themselves themselves to be a great implementation part over the yeah?

SPEAKER_00 30:15

It's funny because it kind of brings me back to to our conversation on scaling practice management. The first the first thing I'm gonna talk about is clean data, right? We talked about this data, I I I kept calling it data hygiene piece of it. Yeah. Uh where, you know, well, first of all, make sure that your EHR reflects reality, right? If there is stuff that's not in there in any way, shape, or form, I mean, I would I would start there. Because there's the you know the written rules and the unwritten rules of of the way your front desk runs, okay? So first of all, make sure that the data there is is clean, is systematized. But the other thing honestly goes back to to my previous point of write down how you're scheduling, how your front desk actually works. Uh, all of these unwritten rules that are just in the heads of the team. I think that's a very important part of the preparation. And on top of that, I mean this is maybe part of already the implementation process of AI, but there's a huge piece of this that is about educating the team and educating the patients for this kind of implementation. I'm happy to say that since joining Talkie, I had, you know, I've run client interviews, even had a couple clients on the show on scaling practice management, talking about the implementation. And the change management piece is something they keep coming back to. You know, let the patients know two weeks ahead of time, via text, via email, on the website, an AI receptionist is coming. We will be still working out the kinks even when it launches. We ask for your patience, but this is for your own good, so your team has more time for you for your actual needs. That is so important. And we do try to help with that. We provide templates for these emails, you know, copy you can use, uh, materials, you know, posters that you can you can hang in your practice to prepare people for that change because AI has a reputation and it's not a hundred percent a good one. You know, I'm not gonna I'm not gonna lie to you about that. I do think, personal you know, bias aside, right? But I do think it's it's a force for good at the end of the day, but we do need to get people used to to the change. Any sort of change. I had a conversation with one of our one of our clients, Lisa Horst from from ESD Pediatric Group. She said that any change that you're implementing has this kind of 90-day ramp up, you know, and for the first 30 days, everybody's going, or a lot of people are going to actually complain just because they like the status quo. They don't want the additional burden of the change. 60 days in, still some complaints. 90 days in, she says, the the majority of these complaints are gone. It was a very similar situation with our AI receptionist implementation. So just a little bit of patience is also called for here. This is a new technology. We're still figuring out, you know, the how far it can go, right? Within healthcare, we're being, of course, very cautious about where we let AI in versus where we don't. I do think a company like ours, like Talkie in the AI receptionist space, we're in a pretty safe space. It's it's mostly administrative, right? So we're I think this is one of the layers where it makes the most sense to start implementing AI. But with any sort of change like that, the you know, preparing people for that change and actually having a united front of this is a positive change, and the front desk team is saying that, and the leadership of the practice are saying that, and the communication is saying that, you know, on and off the record any in any conversations, or even if a conversation gets routed to you know from the AI agent to the human front desk, we provide these scripts for the human front desk to then say, but hey, have you tried resolving this with uh with the talkie agent? Or, you know, it might have a name like Jessica or Zoe, right? Have you tried talking about this with Jessica? Because I'm not always available, but Jessica can always help you with that. You no longer have to actually go through me and wait for me to be available and wait on hold. So try that next time. You know, when both sides are supporting each other, this can be such a positive change for a practice. Uh but there needs to be this collaboration. That's a really important part of it.

SPEAKER_02 34:07

I love how you're handling these things. I think the the piece that strikes me from a marketing standpoint is a really big opportunity for a practice. Uh is that the patients are pretty well aware of the challenges that exist in terms of getting somebody on the other end of the phone who can help.

SPEAKER_00 34:23

Yeah.

SPEAKER_02 34:23

Getting into to be seen for a routine visit versus emerging care. How how do you encourage folks to be thinking about the messaging up front? I think that's a really big opportunity for a practice as they're announcing, hey, listen, you're going to be hearing from our AI agents really messaging up front. This is what it's intended to do. These are the pain points we're going to try to eliminate. I'm not saying that's the nature of the dialogue, but the message itself. I think that's a piece that I'm really curious how you'd encourage a practice to operate in in terms of preliminarily teasing the whole concept that's coming.

SPEAKER_00 34:57

Yeah, thanks for asking that. I would say, I mean, so there are some of these obvious benefits, like no more waiting on hold, for example. I just I just mentioned that, right? Or available exactly. I mean, for some, that's you know, that's that's you know, you don't have to say any more than that. So so the 24-7 availability, and I think you know, one objection to that might be, well, I'll just go through the patient portal, right? I would say not everybody likes to do that. And the benefit of a of a solution like talkie is that you don't have to know how to navigate a patient portal. You can actually go back to the old way of doing scheduling with this new technology. You can just speak. You can just speak your preferences, you can just talk about your, you know, anything that's concerning you about this visit. And again, the AI is set up in a way where if it's doable, it'll do it. If it's not doable, it'll reroute it, right? And I think another thing here is as well, is that AI has this patience advantage, you know? Some people are more introverted than others. I personally, when I call a doctor, I kind of feel like I'm intruding on the front desk time sometimes, right? Because I know also because of the nature of my job, how busy they are, etc. So I don't want to take up too much time. I maybe am not going to ask all the questions I typically would if I knew they had all the time in the world. Well, guess what? AI does have all the time in the world. You can ask about anything you like, right? And you don't have to worry that there are other patients on the on the line waiting. That AI is just for you. And I think that's something that, you know, not a lot of the patients, the, you know, the ultimate end users of it are actually looking at is that this can actually give you more opportunity to uh you know to interface with your with your favorite practice in a way that actually suits you at a time that suits you, right? I think honestly, just being able to call and get your things done and do it via voice, maybe when you're you know, you know, via uh an earbud or a headset when you're driving a car, just just call, just get your thing uh get your issue resolved, just scheduling your appointment. Uh, you know, you can't do that uh you know through a patient portal, but you can easily do that through an AA receptionist like talking. So the benefits are actually multiple. Again, still some kings to iron out, you know, for for some edge cases here, but there's already so much benefit that can be uh that can be reaped here. And I hope that, for example, through through this podcast episode, more uh more practices are going to just give it a try, honestly.

SPEAKER_02 37:14

I love that. I think it's an interesting time that we're living in and definitely. I think there's a a fair amount of fear and apprehension, uh particularly among the the workers that are running a practice. Just if if you wouldn't mind, just giving your perspective on how we address some of the fears that people have. We're actually working inside of the practice. Um, because the technology that you're introducing is a brilliant use case, honestly, because the pain points that people deal with on the front line, there are many of them. And you probably couldn't automate enough stuff to for those people to be super comfortable just to do what they're trying to accomplish, let alone the fear of being replaced. I don't think it's a and that's kind of a scenario. I don't think it's realistic at all. So how do you think about that? And and what will you tell, what would you tell somebody working in a practice right now that's concerned that this capability actually could displace them?

SPEAKER_00 38:03

That's definitely one of the major learnings that I had since I joined Talkie, you know, in terms of our messaging. Because initially coming in, I was quite naive thinking, okay, so we'll position ourselves, you know, as maybe a bit of a cost-effectiveness kind of thing, right? Where, okay, if you implement talkie, you'll be able to reduce your front desk team by half, you know. And and like you mentioned, that's not at all the kind of conversation we're having here. I mean, talking can have excellent ROI with the with the work that it does, right? But not as a replacement, as a supplement. I was just working on a case study, not published yet, but it might be by the time that this recording goes out, where we actually know calculated how much they're paying for the price of the subscription, how much of that actually goes to the you know the the the scheduling conversations, just the scheduling. And on just the scheduling, you know, calculating the worth of those conversations, the ROI was over 10x. And that's you know, just on the slice of of that conversation. Uh so I think you know, more so because coming back to the way this interfaces with the with the practice team, I think the most important thing to realize is that this is not supposed to be replacement. Talkie is supposed to be your assistant as well, as much as the patients, right? You know, I think the practices that do try it and the teams that do try it end up being very grateful that talkie is leaving these notes in the system for them. You know, that the conversations already happened. You can see the transcript of the conversation, as I mentioned, you can revisit it and address it, you know, in a prioritized way, as I mentioned before. The teams really uh really appreciate that. And there's such a large chunk of the front desk team's work that is so far from being automated. There's so much that you have to actually do physically at the uh you know at the at the practice, right? People are coming physically to the front desk, asking questions, they need to have their insurance card scanned or you know their their documents scanned. You're still helping with all of that. And there's also the the very human work of, you know, somebody's in distress and they need somebody to actually physically come up to them and ask if they're okay, ask if they need a drink of water, you know, depending on what what kind of emergency they might have. That is the kind of work that we're not intending to automate, but we want to make room for that work because it's difficult to make that work, to do that work when the phone is constantly ringing, right? So we want to remove that from the equation. And we love the front desk teams that we work with, and we really want to support them with an AI product like talkie. And as you mentioned, I want to emphasize it again the replacing the front desk conversation, it's it's a story that's easy to latch on to, it carries well, it can go a little bit viral, you know, on social media, but it's not the way it actually plays out, and it's not gonna be.

SPEAKER_02 40:42

You know, what strikes me is the fact that some of the things that you're addressing, they're not even actually seeing any of this a lot of the activity that's being missed, but they are being held accountable for waiting times, they're held accountable for call abandonment metrics, things like that. Yeah. Talk about that angle feed in mind for a little bit. A lot of what people are getting relief from are things that in a perfect world, they would actually be getting the calls and addressing it much, much sooner. But when people have to wait, there's a limit to their patience before they're like, you know what, I'm out. This is creating a mechanism that actually engages early on, they're not waiting for a whole lot of time. So I think there's also this patient retention uh lever that you guys pull as well.

SPEAKER_00 41:20

Yeah, patient retention and honestly patient acquisition, right? Because if you're looking at your practice from a business standpoint, right, you still want to make a you know a financial case here, then I think a strong case is, you know, if we take practice A that has an AI receptionist and practice B that doesn't, then you know, whoever calls practice A, the call gets picked up, it's a new patient, new money coming in potentially for your practice. They can you know start their patient chart right then and there, and they can schedule the first appointment right then and there, right? For practice B, this is only true if somebody's at the practice physically and they're able to pick up the phone. And either of those can be untrue at any point in time, and then you're just missing those patients. If you're practice B, you there's so many calls that you're missing, particularly you know on a Monday morning. New patients that are trying to become your patients and and and you know, join your uh your roster of patients, they don't take into account that this is the busiest time of the week and you know, and they don't and they're going to wait for the follow-up appointments to go through, and then they're going to make sure they get through and you know, and become your patient. If they don't reach you, they just call the next practice and the next practice after that. And the statistics are, you know, they they're willing to wait a minute or two maximum on hold before they just move on to the to the other practice. I mean, at the end of the day, we just want to get you know, get help, get support. There's options, there's there's competition on the market, and that's you know, of course, to some extent good, right? That that patients have have options. So, yeah, I think you know, in terms of actually being able to capture that demand, if you will, you know, from a marketer's point of view, to improve patient acquisition, and as you mentioned, also, you know, retention, because for the follow-up visits, the A receptionist applies all the same. I think it I think it's there's a very strong case here for just being able to spur growth and to increase the growth of your practice.

SPEAKER_02 43:03

Phenomenal. I think that there's so much opportunity there.

SPEAKER_00 43:07

Yeah.

SPEAKER_02 43:08

This is actually going to provide. And the efficiency conversations are usually front and center. But there's a oftentimes there's this whole aspect like you just mentioned. There are blind spots that they have no idea are going on until they're already being dinged for it from a performance standpoint. So I think that's an important part of what you guys are bridging.

unknown 43:27

Yeah.

SPEAKER_00 43:27

Unbelievable. And this isn't even you know, we're not even sorry to cut in, but we're not even fully there when it comes to to the technology. You know, uh some of the stuff that's being built out here uh right now is is for example memory of the AI receptionist, right? I think you know, a lot of practices and patience, a lot of people can just imagine the future where when you call your practice, it's you know, it responds to you in a way like it's your own personal assistant at that practice. You know, that previously this kind of convers this kind of a kind of appointment was booked, or for example. It knows that you can never do Thursdays, right? Because you have a standing commitment of some sort. So it just won't suggest Thursday appointments to you. Little touches like that. I mean, if anybody's used Claude or ChatGPT for any you know longer stretch of time, it does you know build up a memory, it does adapt to the way that you interface with it, right? And I think these AI products, uh, you know, talking included, they're going in that direction of being able to have even more, you know, some additional benefits on top of being able to do the work of a human receptionist. The you know the patient's part of the equation, I as in the uh the AI can be more patient and more available than a human, I already mentioned, right? But this you know, memory and being able to be your assistant at the practice, uh, you know, that's coming. And I think that the future is actually really, really bright for this kind of space. And I do hope that uh more practices are going to give this a try and actually talk with, for example, our team building talkie.

SPEAKER_02 44:56

We're cleaning the conversation. I'd love for you to just put your uh crystal ball right out front and center. You're obviously doing some things right now that are not in your live environments, so maybe not asking you to give away a bunch of secrets, but looking ahead the next three to five years, what are some things that you feel are a possibility that people may not believe are possible today? Because I think where we were before AI started to emerge, we really didn't know how effective it could be and what some of the options could be. And even today, if you look ahead, even for another week or two, there's likely to be something that pops up that no one believed was possible. But here's some of the things that you think we should be uh looking forward to from the advancement of a perspective.

SPEAKER_00 45:38

That's so interesting that you mentioned that because just this week Chat GPT dropped the GPT live model. I don't know if you saw the demo of that, but it's so interesting. It's such an interesting new voice model where there's a lot more possibility to interrupt the AI and have a more natural flowing conversation. And I think that's one of the things that's on the horizon. Three years is definitely too long of a timeline for that, you know. That's coming in the next, you know, six to twelve months, I feel. So, because right now, you know, one thing that is true of a product similar to ours is there's a little bit of turn taking, right? You do have to most of the time wait for the agent to stop speaking so you can respond to it. You know, it's it's pretty smooth and pretty flexible and pretty fast, um, but there is still that that part, but that's being solved already. And the other, and you know, again, faster than we would have realized because now the AI is building the AI, right? That you know, one model is coding the next. And that's you know, that honestly makes my head spin when I think about it. So uh and the other thing that I would mention, even more sort of futuristic, I'm seeing this from from one of the one of the guests that I just had on scaling practice management. I asked a similar question, what's coming, and they said that think about what we're doing right now, which is a lot of the time through text and voice, and now think about the vision side of the equation, because that's also coming, you know, multimodal models that are going to be able to not just hear you but also see what's happening. I think this is more so, you know, for what's happening in in the doctor's exam room, for example. That's gonna be an interesting privacy conversation, definitely of you know, do you allow the AI to also not just hear what's being said, but also see what's being done and kind of advise on that and create a note from that? So that's gonna be also its own interesting conversation there. But that's something that I would put in the minds of the of the watchers and listeners as well. That, you know, this addition whole additional layer is also coming. So the you know, a lot of people are talking about the the AI bubble, right? That this is going to kind of slow down at some point, that we're going to, you know, hit a bit of a you know a rougher spot with this, maybe at some point in the in the future. But for right now, there's still so much unexplored and so much on the horizon that I don't think we're I don't think we're getting there yet, to be honest.

SPEAKER_02 47:53

It's mind-blowing, honestly. I think uh even a year ago for me, just playing around with some of these models, I couldn't get it to give me a decent PowerPoint. Honestly, I'm giving it all content, asking for it to arrange it and make something flashy, and it does give me these really primitive looking renderings. Fast forward, it wasn't even two months after that, my last experience, where I started to see really groundbreaking capability being unleashed that makes my newsletter, for example. I had a really basic cover for it because that's the best I could come up with using the tools. Now it looks like I've employed a team of graphic design experts and visualization experts. It's just amazing how fast this stuff is evolving. Jakob, this has been an amazing conversation, and I'm certain that listeners are gonna want to understand more about what you guys are thinking about what you're about to do. How can people find you and how can they follow along and just be aware of some of the great work you guys are doing?

SPEAKER_00 48:49

Yeah, so we're doing a lot marketing-wise. I don't want to drop too many CTAs, but definitely check out talkie.ai, t-a-l-k-i-e.ai, uh to find out more about uh you know what we're building in our product, it's gonna be. I mean, we we put a lot of effort into our site to keep it up to date and keep it very informative. Uh, of course, the most important button there is book a demo button. I honestly think if you've never had a demo with an AI company, then now it's time to at least educate yourself on that. And a good way to educate yourself is to book a demo with with a company such as such as ours. So that's one place. I'm also quite active on LinkedIn. I must I mean borderline addicted, but let's not let's not delve there. So you can find me on on LinkedIn as well. And we also have quite a live YouTube channel on the YouTube channel of Talkie, which I hope you can link to in the show notes. Uh, you will also be able to find Scaling Practice Management, the podcast that I'm so proud to host for Talkie. It's been such a great, great series to drive, met so many cool people through that, yourself, of course, included, right? So I really, really recommend that. If you give scaling practice management a watch or a listen, please get back to me with your thoughts. If you think you could be a guest, by all means, let's have that conversation. Because yeah, I mean, recording these episodes, having podcast conversations, I mean on both ends, it is one of my favorite things to do. So those would be would be some of the main things to to check. And you know, who knows by the time where when this airs, who knows what else we'll have out there. So yeah, just stay in touch, follow us on LinkedIn as well, just as talkie and not just uh just me personally, and definitely we'll find some some cool stuff to do together, I'm sure.

SPEAKER_02 50:26

Fantastic. I'm stepping away from this with an appreciation for how they're approaching things for me. The way that we're using the capability really needs to be aimed at solving the challenges that human beings are actually wrestling with. So I love the focus on the human aspects of this. Uh there's clearly a passion that you speak with, and I really sincerely appreciate it because I think there's a lot of apprehension when it comes to technologies. And what you guys are doing is remarkable because it really is personalizing the care in the way that it's learning the patient what they need, you know, these memory aspects of it. It's beautiful. Uh Jakob, thank you for trading seats with me. And for my audience, as uh Jakob mentioned, you'll find a ton of information in the show notes. Jakob, thank you so much.

SPEAKER_00 51:13

Thank you, Chris. It was a is a it was a pleasure. Great being on the show, and have a great day, everyone.

SPEAKER_01 51:20

Healthcare has spent five years getting AI in the door. Pilots, ambient documentation, predictive models, triage tools. The technology is here. The oversight is not. I'm Chris Hutchins. I've spent thirty years inside health systems. The hard part is almost never the model. It's the meeting after the model. It's who owns the alert. It's whether anyone will say out loud that the data isn't ready. I wrote a book about that work. It's called Beneath the Signal. The human work behind trusted data and responsible AI and healthcare. It's for the people who have to make this actually function. It's the CMIOs, the data leaders, the ones who get called at three in the morning. Search Beneath the Signal on Amazon and get your copy today. And you'll be back here on the SIM room.