Read the Full Video Transcript

Jason Hafron: Just as awesome and exciting is Dr. Patel from Atlanta Urology and his administrative team. Dr. Patel is—I don't know what I would call you—a maverick, a cowboy. But he has really figured out how to apply AI in practice management, how to improve efficiency, lower overhead, lower cost.

And he is really—we're going to see today—doing some pretty innovative things. A lot of things that you could potentially bring back to your practice on Monday. So Dr. Patel and team, welcome.

Neal Patel: Thank you, Jason. It's always an honor to follow Scott Eggener. I will say that what you're about to hear is someone who is a sucker for gizmo idolatry, but hopefully today what you'll leave with is some great new technologies that you can implement in your practice. I've got a great team here today, who are each going to talk about a different way we've implemented automation and AI at our practice.

A lot of what you're going to see today wasn't even here a year ago, and that's just to show you how quickly this AI is progressing. In many of the other sessions, you've been learning about maybe ChatGPT and so how you can apply that in your practice. But what we're going to do today is take it to the next level and show you how it can actually improve our lives in many different ways.

So I just wanted to introduce Mark Thompson; he's our VP of Rev Cycle. We have Candice Sonko, who's our VP of Clinical Operations, and we have Alicia Davis, who's our VP of ASC Operations and Clinical Trials. All right, so we can go to the next slide. Oh, that would be me. So today what we're going to talk about is obviously the growing administrative burden in healthcare. Obviously, this can lead to costly mistakes.

We're going to talk about why AI and what we call robotic process automation is important for your practice management. And we'll show you some actual implemented technologies that we've done in our practice, as well as the data that's driving it. And then obviously, open it up to questions. So first, I'm going to start off with just a quick Slido just to have the audience see what actually you have implemented in your practices, as far as AI and automation?

So I want to give everyone just a second to do that. So we just want to get a general consensus or an idea of what everyone's practice here is currently using as far as an operational tool, whether it's for practice administration or for helping physicians do their job better.

Okay. Wow, a large number of you have not joined the AI wagon just yet. All right. Well, hopefully this will be very informative for you. So I'm going to turn it over to Mark, our VP of Rev Cycle, and he's going to go over an implementation that we did on the practice administration side in our billing office.

Mark Thompson: Thanks, Neal. We'll kick it off today with a few numbers. So maybe some have seen these numbers before, but the United States spends upwards of $5 trillion on healthcare spending each year. And we love that a large portion of that goes towards actual patient care. But we're at a point now where close to 25% of that amount is going towards administrative tasks like Neal alluded to.

And those are just not necessarily value-added dollars the way that the others are. So we've taken a few steps at our practice to try and cut down on the dollars spent, the time spent, the complexity that all the regulation and administrative burden has heaped on top of healthcare. So yeah, wages and costs continue to go up. We know that reimbursement's not going up at the same clip or at all in some cases.

So there's growing pressure to make sure that we can save every dollar, every hour that we can. And then we'll talk through some integration challenges as well, how we've worked through those at our practice. So especially seeing the results of that Slido poll, about three-quarters of the practices out here are not using RPA or AI tools today in practice management.

I'll start with a 101, just the basics, differences between some of the tools that we're using at the practice. So robotic process automation or RPA, as some might've heard it, it's a simpler form of automation. It gets the same result as some of the generative AI solutions that you've heard about today at this conference, that you're going to hear about from my team.

But we really kicked this off about three years ago using a suite of RPA bots to help us automate some of the most basic tasks in revenue cycle medical billing. The important distinction between RPA and generative AI is that RPA is totally rule-based, so it's being trained by the humans. It's executing exactly as it's been trained to do.

Whereas generative AI and some of the AI tools that Candice and Alicia are going to walk us through that we've implemented more recently in our practice, the bot is training itself. It is getting smarter over time. It is generating its own intelligence. For robotic process automation, it's very crucial that if you're going to consider exploring this option, you need very structured data, you need tighter templates.

Whereas AI, generative AI can use and thrive in more unstructured data. All right. So I will walk us through some of the RPA wins that we've seen in revenue cycle and billing, and really more broadly in our practice administration. I think the immediate benefit that comes to mind when people hear automation, when they hear AI, when they hear RPA is headcount.

How much can we reduce headcount? FTE reduction—you see it in the media, you see it in the headlines, and that's true. There's a huge potential for reducing headcount. We've actually achieved about a 10% reduction in our headcount over the last two or three years using some of these RPA tools. And we really feel like we're just starting to scratch the surface.

But today, I really want to focus on some of the other benefits, maybe the lesser-known benefits that don't hit the headlines when people hear about automation. So a big one for us is scalability, so you front-load the time, you train the bots. You create these automated tools that then allow us to plug and play.

If we were to open a new clinic, hire a host of new doctors, launch a new service line, we've got the engine in place to absorb all the administrative tasks that go along with it. So it's one less thing to worry about when expanding rapidly. The surge capacity goes along with that. If there's ever a backlog at any point in your process, the bot allows you to work 24/7.

The bot works 24/7 to help you catch up quickly, so that there aren't bumps in your cash flow or in your operations. Patient lead time—we actually have some bots that communicate with our patients to let them know about upcoming services. We'll show some of those in a moment here. Maybe they've estimated patient liability before they come in for that service.

And our bot allows us to stay further and further out, so that our patients are having that communication more proactively and with bigger lead time. Reliability—so the way that I sold this to my team when we had some skeptics in the early days about what exactly would this mean? We're bringing in these bots to replace jobs, the anxiety that you'd expect initially ensued.

We like to say that our bots don't take vacations, they don't get sick, they never go on leave. There's that ultimate reliability that once they're in place, once they're trained, once they're ready to go, you can expect a reliable product every single time. They don't have those human interruptions. Accuracy, reconciliation—we'll go through in a second some of our actual, practical applications for bots in revenue cycle.

But what we found is that we're billing more unbilled charges, things that were left behind before, the bot is finding these things. Mistakes that were being missed by our people are being rectified by the bot. And then lastly, we're using bots to create functionality in our native EMR or our billing systems.

We're able to stay on an EMR billing system, probably be more flexible because we can use bots to create those functions that are not there native in the system. All right. So some of the use cases at our practice, we're using the bot to verify eligibility and benefits. The bot is pinging these payer portals.

They're text messaging with the patients prior to the visits, making sure they know not only that they're going to be eligible when they see Dr. X, but also how much they're going to owe when they get there. Our bots are creating claim shells. They're doing coding for some of our services like anesthesia in our ASCs. They're posting payments.

We have them clearing patient credits, teeing up refunds so that the books are clean. When payers are denying our claims, requesting medical records or other very simple administrative-type requests, our bot is actually curating those records, collating those records. And putting together a package so that we can efficiently respond to those denials and correspondence from the insurance.

And creating exception lists, cleaning up the system, doing chart audits. You name it, as long as it's structured data, we've been able to program bots to do a lot of those rote, repetitive tasks.

All right. We're going to just show—this is going to look like mass chaos behind me probably—but we're just going to show a few bots in action, and I can narrate what's going on here. So the top-left screen is showing our eligibility benefits bot. It really does move this fast. It really does; it's running on the cloud. It runs 24 hours a day or until it finishes its work.

The credit resolution bot on the right is teeing up patient overpayments and/or applying those credits to other underpaid claims, so that we don't have people doing data transcription. The bot on the lower left is actually creating claim shells for our anesthesia claims, and actually going into the medical records, reading the time that was spent, putting in those time units, those base units, bringing in overall the patient demographics.

And actually teeing up the claims so that it can go out to the insurance company. And then the last bot on the bottom right is retrieving medical records, collating them in the way that we programmed it to do, and saving all of those, so that when our AR and billing team goes behind, they don't have to do all of that repetitive work.

I guess I'll just throw in before I pass this off here, that the way that we talk about our providers—we want to use all of our providers to the top of their licensure, mid-levels, MAs. We're trying to do the same thing on the administrative side of the house. We don't want to have our people doing rote, repetitive tasks that can easily be automated away.

And when I mentioned that 10% FTE reduction that we've seen at our practice, we've actually been able to redeploy over time as we grow, all of the good folks we have into other more value-added tasks. And I think that that's been a huge sell for our employees and our team to where they don't fear this automation anymore. They see that it actually can boost everybody up along the way.

Neal Patel: And if you think about it, you may be spending hundreds of thousands of dollars on various outside services to do this, and it is less than a tenth of the cost to do in-house, and it's built around your own system.

As we know, we all have different silos of techs, old and new, and it's sometimes hard for all those companies to communicate together, integration costs. Six to 12 months sometimes to implement, these can be created in less than a week.

Speaker 4: Somebody has to write this software, right? Are you guys doing that?

Neal Patel: We have people in-house, some of it is actually not software. There's some no-code solutions that we've been using, but we also have a team of people that do some coding as well.

Mark Thompson: We have an outsourced partner to help us with some of the more complex coding, but yes, it's a blend.

Neal Patel: Yeah. For the sake of time, we'll keep moving, but we'll definitely do questions at the end. So next up is Candice? No. Sorry, Alicia, who's going to talk about the use of AI medical agents.

This has been one of the biggest changes, I think, in AI tech in the last year is the ability to have human-sounding medical agents. So I'm excited for her to talk about how we've implemented that in our practice, and what the implications of this can mean for your practice.

Alicia Davis: Thank you, Dr. Patel. Yeah. So one thing that we continuously struggle with is the number of phone calls we have to make to our patients, whether that's pre-opping them, multiple attempts to get them on the phone, collecting their medical history, and reviewing all of that information. Calling them post-operatively—how are they doing?

Most are doing fine, but you have to call them five times to say, "They're doing fine." So we're definitely exploring solutions that allow us to have conversational AI, interactive AI with our patients. So some of the things that we've done with this are we can educate our patients, creating hotlines for common disease states or conditions like OAB, BPH. Patients can call, and they can talk to AI as long as they'd like.

They don't need to get to the next patient. So having those available for patients, collecting that information. Whether it be just a general pre-anesthesia form of "What medications are you on? What complications have you had in the past?" And feeding that information back to us so a human can review it, make sure that the patient is safely moving forward with care.

Again, just saving the time, avoiding the multiple phone call attempts, really being able to still escalate that critical information. So if a red flag that we program it to recognize comes up, being able to flag that to our team to say, "Hey, a human nurse needs to call this patient." And it can also transfer the patients out to the human nurse if it's something critical on that phone call.

So a lot of fun. I'm going to show you on this next slide, just some quick snippets of that. We don't have a whole lot of time, but this first call you're going to hear is going to be that information collection, just asking the standard questions for a patient.

AI: Go over some important information about your medications. There are certain medications that you'll need to stop taking before your procedure.

I'll guide you through this process. Can you tell me if you're currently taking any medication to treat ADHD?

Speaker 7: No, I'm not.

AI: That helps to rule out a few medications. Next, I'd like to ask, are you taking any high blood pressure medication?

Speaker 7: No.

AI: That's helpful to know. Are you currently taking any herbal supplements?

Alicia Davis: Okay, sorry. Now, this is post-operatively.

AI: It's excellent to hear that you're not having any problems urinating. However, I do want to gently remind you of the importance of staying hydrated, especially during your recovery.

Not drinking enough fluids can sometimes lead to issues like constipation or urinary tract infections. Can you tell me a bit more about what's been making it hard for you to drink as much as you'd like? Is it due to a lack of appetite or maybe some discomfort when you drink?

Speaker 8: I think it's just my mind telling me to give it time to heal. It's not that I'm dehydrated or anything. I'm still drinking plenty.

I'm just not drinking as much as I would probably normally.

AI: It sounds like you're being mindful of your body and giving it time to heal, which is great. It's also reassuring to hear that you're still drinking plenty, even if it's not as much as you normally would.

Sometimes, it's all about finding that balance and listening to your body. Now, let's move on to the next question.

Alicia Davis: So again, just a quick snippet there, but what I like about that second example is that patients can say whatever they want, it's going to pick up on those pieces that are important to care. So I know even when we demoed it, it said, "Do you have a history of falls?" And I was like, "Well, I was walking through the living room, I tripped on a cat."

And I went on this whole ramble about tripping on the cat. At the end of it, she was like, "Well, it sounds like that was an isolated incident. I'm glad to hear that you don't have a common occurrence of falls." So it's being able to pick up that information that's actually relevant, and again, not taking human bandwidth for that.

This is just a quick data slide, and I know Candice has a lot of really fun stuff to show you guys as well. So I won't spend too much time on this, but this is very, very early on in our pilot period. This is without telling our patients what we were doing, without giving them the information of, "Hey, this is something you're going to receive a call from."

670 distinct calls, 250 of those were connected, because again, that's the same thing that happens with our humans—we just have to keep calling. Out of those that were connected, 76% of those patients verified, 75% of those actually completed their phone call with AI. And now, I'll let Candice talk about some medical scribing solutions.

Neal Patel: Yeah. So just to comment on that, we've used that for pre-op intake, post-op calls for all of my patients in the recovery. And on the whole, they all really loved it. Some of the older patients were told by AARP that they should be wary of any phone call that calls and asks them for their name and date of birth.

But in general, if you prepare them ahead of time, it's great. We're also rolling out a version of this for our OAB navigators, prostate cancer navigators, kidney stone, and BPH that we're launching here in the next few months. So I'll turn it over to Candice to talk about our implementation of medical AI scribes.

Candice Sonko: Yeah. So another utilization that we've found very helpful for our efficiencies within our clinics is an AI scribe solution. I think we've actually helped some of our—we've got great reviews from our physicians as well on the results of this, that just balance out their day better. So if we average all our physicians in a day, we see about 38 to 42 patients in a day per clinician.

Of that, you spend about six or seven hours with the patients, but an additional two to three hours or more, depending on the clinician, doing those notes, just depending on the patients' needs and what came out of that appointment. So what we've been able to see with the AI solution is we've definitely created a lot of efficiencies within our workflow.

Not just with our providers, but with the support that the support staff is able to give them with orders, etcetera. It gives physicians more time with our patients, and our patients really enjoy that. I think anytime I have to talk to an unhappy patient, it's always about time. I hear that a lot of times, so I think that's very beneficial to our patients.

We're seeing an increase in our notes, the detail in our notes, which is leading to improved coding. So working closely with our billing team on that, and it's been working really well for us on the back end. And of course, the reduction in time spent on documentation—that's a huge win for our providers. So additional uses we use it for is you can create letters based off of that appointment.

You can click a couple of buttons and there's a letter put together for you to print out to send home with them. Reduces time of your support staff, as well as the patients calling you back later to ask for it again. There are several uses for that. We put on here the Meaningful Use documentation. There are so many things that you can use it for, based off of just the transcript that comes from that appointment.

So just a little bit of data here. We had about 20 providers utilizing it this time, so this was before we were fully implemented. Now we have all our providers utilizing it every day, but we've saved about 54,740 minutes using this for our documentation. If you transcribe that into monetary means, that's $91,000 in saved time, so it's a pretty good ROI on that.

And then the next is just a quick visual of it that I'm actually going to let Dr. Patel walk through with you guys, because it's his workflow in the day-to-day.

Neal Patel: Yeah, go ahead. So this is basically starting after you've done an encounter. So I essentially will pre-put in some context before I walk in the room. I do the encounter, I leave the room. I give a verbal plan to my staff while it's also recording, and it generates an initial SOAP document.

Now the great thing about AI and what's really changed with generative AI over the last year is that you can take that transcription or that generated note and do anything you want to it, however you like to write your notes. For example, this is a template for a urology patient.

This is not our template per se, but this is a good example template where you can actually tell it and give it instructions like you would a ChatGPT on how you like your notes to be done. And there's a lot of ways to really 10X what you're doing here, because the great thing about these notes is the AI's listening the whole time while you're in there when you may sometimes zone out if the patient's going off on a tangent.

And it picks up everything that they're talking about, things that you may not even consider relating. They're talking about another disease or disorder, and they may ask you, "Is that related to my urologic condition?" The AI picks all that up, and it just makes your note way more comprehensive to anyone that's reading it. So you can see here, it's applying a template.

This is all done automatically, by the way. But just to show you how that works, it automatically puts in diagnosis codes. We've actually taught ours to automatically do AMA 2021 MDM coding based off the note, and suggest a code with the appropriate documentation underneath. So it passes all the audit musters and compliance musters that we need.

You can do fun things where you can actually take your transcript and create Meaningful Use documents. So our patients always leave with some sort of Meaningful Use document that's very meaningless. And this allows us to actually say, "This is what was discussed." And you can even put in whatever you want it to print out about your visit.

It can be emailed, it could be printed, it could be on your letterhead, whatever you want to make it basically almost an instantaneous document. You can also create separate billing documents, and these can all automatically be generated that your staff can then pull, or it can be automatically pushed to wherever you need it to be pushed.

But it applies everything from the transcription, so that your coders have what they need, and it can make it a very robust document. What we've found is that a lot of docs were just putting in the minimal amount of documentation because they just wanted to get through the day. They just wanted to write the note and move on. And what we found is with AI, it's capturing so much that their actual level of coding went up.

And we first thought it was a mistake; we went through, we audited everything, and it turns out that it actually helped us code better. And it was better than even some of the teams that we were using to do third-party coding teams, so we found that we could do a lot of this in-house. It's done things for a tumor board presentation; I said, "Can you help me write this tumor board from this note?"

And it'll do it, so it can do whatever you want. But the great news is that for me personally, I used to document a couple of hours a day, now it's five minutes and it's just editing and making sure that the note is accurate. So this has been a huge game changer for our docs. I think the second it starts going slow, we get about 15 complaints instantaneously that they just get a little frozen, so they rely on it quite a bit.

And it's really changed the paradigm for them from not having to dictate as much or type as much. And people—not just us—but everyone around us that sees our notes actually truly sees the amount of work that we do as physicians for our patients. And it's great to just see that reflected in our documentation. So with that—

David: Do you have a metric on whether your doctors are happier? Can you put some numbers to that, some time saved or anything of that order?

Neal Patel: Yeah, it's about two hours a day for every doc. I used to spend my entire lunch trying to do my morning notes. Now I go out to lunch.

I don't have to worry about it. My notes are there. I just review them real quick, make sure they're accurate, and I just sign them, and it's been a life changer.

David: Nice.

Speaker 11: Neal, I'll just address David's question. So I was always very efficient. It didn't change the time that I spent in a day documenting, but now I walk into a patient's room with a laptop, I press record, I sit. I look the patient in the eye, we have a great conversation. I walk out of the room, I say, "Stop record," and the note's done.

I don't have to go to another computer or sit there and dictate something. I do review the note, each and every single note, to make sure it's accurate. But let's say I'm using ORGOVYX or JELMYTO and it misspells it, I tell it ORGOVYX is pronounced this way, and it knows for the next time how to correct it.

David: In the note, so when someone else looks at the chart, they have a pretty good idea of what happened that visit.

And if you send a note to the referring doctor, he has a good idea?

Speaker 11: Right. It also tells you, "Oh, how are you?" "Oh, I had a knee replacement." "Well, what side did you have?"

"Right knee." And it'll catch that and it'll tell you, which normally I didn't pick up.

David: Right.

Speaker 11: So to Neal's point, the EMR code is a higher code because you're incorporating a lot more data, while I'm sitting there and having a conversation.

Which is the valuable time that I used to not have with the patients.

David: Thank you. The only one awkward thing about that is you got to tell them at the physical exam. You can't see what the heck you're doing, correct?

Neal Patel: Yeah. So one thing about it is it does change your visit a little bit, because you have to change how you talk to the patient. Because you're obviously talking to them, but you're also verbalizing other things like, "I'm going to do a physical exam now. Your testicle feels like this." Or you have to tell them about the name of the test.

You can't be like, "Oh, we're going to run a special PSA test to make sure that maybe a biopsy is right or not for you." It is really you have to enunciate what you're going to do because then otherwise, you got to edit it later. Or maybe it'll say in the note you're running a special PSA test.

So there is some changes in how you verbalize, but it's a very quick transition into doing that because of the time savings that you get out of it, and your patients are educated better because you're actually telling them names of tests. And you were maybe reviewing labs or documents that you would just say, "Your labs were all normal," but you're reviewing them and then you get credit for all of that.

Speaker 12: Was this built into an EMR program that you had? Did you buy it off the shelf? Did you code it? How can you build this?

Neal Patel: Yes. So there are EMR companies that will sell you or have a feature that you can add in. In our experience, they were charging a dollar per chart, and at our volumes, that was just way too expensive per year. There are companies, there are many companies out there, companies that are here that are doing it as a third party, so they can integrate with any EMR, but they sit on top.

So it's ambient listening. You have it on your phone, so I hit record, it's all HIPAA-compliant. When I'm done, you just copy and paste it in from the website. But the company's actually developing technology where you could be in your EMR, if it's a web-based EMR.

And it can read and see the patient's name, the chart, and be like, "Hey, you just saw this patient in our system. Is this the same patient in your chart? Do you want to pull in your note immediately?" So it can integrate with your EMR without actually being truly integrated with your EMR.

Speaker 4: So you contracted with a third party for this? For this scribe technology, you contracted with a third party?

Neal Patel: And it's very economical, like very economical. You'd be surprised.

David: Which company?

Neal Patel: So there's many companies; the company we use is Heidi, but there's a ton of companies out there. We actually found the bigger companies like Dragon or Microsoft, those companies were way more expensive and were not as nimble. It wasn't as good to customize the way that we wanted to customize.

They were very restrictive, but there's a lot of companies out there. You just search AI scribe on Google, you'll see a bunch. We don't necessarily have to—they're all very economical, and I think you'd be happy with anyone. We don't want to push one company or the other. Yeah.

Speaker 12: For this technology, when they call the patient to talk to them, is anybody auditing that or making sure they're not saying something crazy?

Neal Patel: Yeah, Alicia?

Alicia Davis: Yeah. So to start, the company that we were using is doing 100% nurse live calls, so they have a nurse listening in to every call.

And then we're going behind them and having our staff review the transcripts and listening to the audio, just to make sure that there's nothing that deviates from either our protocols or pathways, and there's no care concerns.

Speaker 12: And then eventually you would stop doing that?

Alicia Davis: Yeah, exactly.

Speaker 12: That was an outside company too?

Alicia Davis: It was, yeah.

Speaker 11: There was just another session about the use of the G2211 complexity code.

I'm just wondering if you've incorporated either of these tools to create some rules about when to bill a G2211?

Neal Patel: Yeah. We have, we have. So as part of our coding, MDM coding, we did add a special prompt in there where we loaded basically all the criteria for G2211 coding, and said, "You have to have a greater than 95% confidence that this note meets the G2211 criteria." And if so, at the bottom, it writes a little sentence that says G2211.

And then the reasoning why it thinks G2211 should be applied. Now sometimes we'll get it and I'll be like, "Ah, I don't think that this qualifies." But then we go to auditor compliance, and so actually yeah, it does. But obviously, we have auditing teams on the back end making sure that what it's saying is obviously correct.

And then giving us that feedback back so that we can modify the prompt language in our template creation to make sure that it's more accurate.

Speaker 13: Sounds like the process requires a champion to get going. You're obviously it for yours, so maybe you guys all became champions at the same time.

What would it take the practice that's uninitiated, how many months to push out the first product?

Neal Patel: So for AI scribe technology?

Speaker 13: Sure.

Neal Patel: You could start it on Monday. It's actually free for most of these things that you can actually just get it up and running pretty quickly.

Now, minus the administrative red tape that your practice may have for all of that, to get it up and running, it took us a day.

Candice Sonko: I will say this was our third solution, so we had tried two others before that, so we had spent months trying to perfect those.

And then we had this pushed out to all physicians within four weeks.

Speaker 13: Got it. Thank you.

Speaker 14: So these systems are great for new patients. How do you guys handle established patients?

Neal Patel: It's a great question. So there's a couple things that we do. So the first is for established patients, we have a few workflows. The first workflow is you're looking at that patient and you're quickly summarizing to the AI before you walk in the room while they're there. So the patient's coming here for follow-up, they have these problems.

Most of these AI systems also have the ability to add context. So you can put in your previous note and say, "This note was dated October 23rd, 2024, please use this as context." And you can paste your entire last note in, and it uses that data for your follow-up because then you can see why they were here. You can have that conversation and follow up on each assessment or problem that they had.

And then it's pulling from that data and helping you write your note while not leaving out certain things. So it'll pull in prior pathology data, data of their past surgery, so we do that. We also have built an RPA bot that's actually going into each chart and actually pulling the last note, the last procedure they had, the most recent diagnostic test, and automatically typing that into our patient's chart as context.

So all we have to do is actually look at that system and not even have to go into our EMR beforehand just to be able to review the patient before we walk in. So you're basically like I have an iPad up or my phone up, and I'm walking to the room, and I'm doing a chart review without ever having to log into my EMR.

Alicia Davis: Yes, it can. It comes with a telephone number that you can put on it. So if you were to put something up on your website, it can dial directly into the AI.

And then you can program the AI to do what you want it to do with that information, whether it's just an educational conversation, like I said earlier with OAB or BPH. And then what do you want it to do at the end of the phone call?

Speaker 15: Talk about, "Hey, I had this surgery," and they're trying to get through.

Alicia Davis: Yeah. You would just build it for what use case you would like to use it for.

So if you're like, "I want it to be available for post-op concerns," then you would just tell it that's what the phone number's for.

Neal Patel: But you can see the potential of introducing an AI-based IVR system customized to your practice. When someone calls, they immediately get someone on the line that can answer questions. So you're not relying on them waiting through a whole bunch of options. Someone picks up immediately and they say, "I just need directions here. What's the address here?"

Or, "Hey, when's my appointment? I forgot." Or, "I have billing questions." It can automatically start routing them where they need to go, and it's conversational, instead of like press one, press two, and then they get lost. And then you can also have it so that it can answer basic triage questions that you have put strict guardrails in, so that it's not going to give medical advice.

Obviously, if you told it to, yes. But we have made sure that it is very strict in what information it can give the patient and what advice it can give the patient. So if we're okay with it counseling the patient on post-operative stuff, we find that that is safe stuff to counsel on, and there's also escalation.

So if they're doing a post-op and the patient has said something that triggers an escalation, it doesn't just say, "Oh, that's good to know. We'll let the team know." It will actually start sending notifications or transferring it to a nurse or do some escalation management. So it's not just like, "This patient's having a massive post-op MI and they're saying they're having chest pain," and it just says, "Okay."

So all those guardrails are built in and are constantly being updated. But the other great thing about it is, for us, it's cardiac clearances and blood thinners, like pre-op counseling for these patients. One of our biggest challenges is staying up to speed on getting patients properly pre-opped beforehand and not canceling them at the last minute because they didn't get a clearance.

Or they're still taking their blood-thinning agents, or they forgot to stop their Ozempic or whatever it is. Now we can have it call them ahead of time, but anyway, thank you. I hope this was informative.