Did you know that some medical practices lose up to 50% of all patient calls? Ada Andruszkiewicz, co-founder of Talkie.ai, explains how AI can solve the communication crisis in healthcare.
In this episode, you'll discover:
Ada reveals shocking statistics: the difference between 8:00 AM and 8:15 AM can mean 200 additional calls. Patients don't call when it's convenient for the practice - they all call at the same time, creating impossible-to-handle peaks.
Talkie.ai integrates with modern EHR systems within 2-4 weeks, offering a comprehensive solution that can handle complex scheduling scenarios with different insurance providers and specialties.
The company achieved SOC 2 Type II certification, implementing multi-factor authentication, encryption in transit and at rest, plus regular penetration testing and simulated phishing attacks.
Ada describes a vision where AI will proactively contact patients about scheduled check-ups, colonoscopies, or mammographies - revolutionizing preventive care.
{{lead-magnet}}
Jan Kamiński 0:02
Welcome to Keep IT Healthy podcast where we explore how technology is transforming healthcare. I'm your host, Jan Kaminski, Co-founder of Momentum. Today, I'm speaking with Ada Andruszkiewicz, Co-founder and COO of talkie.ai, a platform that streamlines patient communication in healthcare with AI. In this episode, we'll explore how automation can reduce front desk strain, improve access and bring empathy back into everyday interactions.
Jan Kamiński 0:39
What inspired you to co-found talkie.AI and obviously, for the listeners who aren't familiar, could you give us an overview of the company and what specific healthcare problems you're solving?
Ada Andruszkiewicz 0:49
In a nutshell, talkie is an AI agent for front desk teams at medical practices, and what we do is we automate conversations between patients and and practices, saving money, time and frustration actually on both sides, the patient and the practice. So this is essentially what we do. But in terms of the other part of your question, you asked me about my founding story, so I'm not a typical kind of I've always wanted to be an entrepreneur, person like I've never actually thought about becoming an entrepreneur, but when the right opportunity presented itself, I just felt like it's the right opportunity to take. And for me, there were kind of three ingredients in that rightness of this opportunity. First, it was the people. So I knew my two co-founders from previous projects, and I knew those are people that I really want to work with. We had been in lots of different situations, and I felt like if I'm to run a company, it would be with those two. And the second bit was the product itself. So a product that had real potential to change a status quo in a given market and in important markets. So the product was interesting. The people were right, and also it felt like the right time for me. And it was kind of from two angles. One, I felt like I had enough experience from my job in the UK. I ran, you know, I did various jobs in London at a company called right move, and it felt like I'm prepared to do something on my own. But also it felt like the time is ripe on the market, like the technology adoption and the readiness of the market to adopt such technology was there. So it felt like, oh my gosh, it's like the right opportunity for me.
Jan Kamiński 2:43
You didn't start with healthcare, right? You started as a more general platform, and then you pivoted into healthcare. Why transitioning?
Ada Andruszkiewicz 2:50
Yeah, so this is really interesting about healthcare. I mean, healthcare is like the perfect use case for automation with virtual assistants, because it has kind of the right characteristics. There are some characteristics that you really cannot fix with anything else. So if we look at the demand side, still, people contact their healthcare facilities, especially in the US via phone, over 70% of interactions in healthcare are still over the phone. And also, you know, the volume of the phone traffic is extremely varied. We observe that our customers experience the difference, you know, between eight o'clock and 815 can be like 200 calls. So suddenly everyone starts calling. And people don't call when it's convenient for the practice. They call when it's convenient for them, and coincidentally, it's all convenient for them at the same time. Therefore, everyone is calling at the same time. You can see the difference within a week, Monday, Tuesday being the most busy within the year. You know, the flu season for some practices. So that's the demand side. And on the on the supply side, there's never enough people to pick those calls up, because you cannot clone people at that speed like you cannot simply, you know, hire one person to work in the low traffic hours, and then, you know, have extended to 10 or 20 people when everyone is calling the job of a medical receptionist is extremely stressful. It gets a bit monotonous, which means people rotate, and nearly half of people in those jobs quit before the end of the year. So you find yourself in this perpetual cycle of having to hire, having to train, dealing with absence and dealing with huge burnout and frustration, because this is a really, really demanding job. So also what's so what we find as a result of this demand and supply situations that a lot of practices miss a lot of calls. And some of the practices tell us we miss 30% 40% 50% of our calls because we simply cannot pick them up. And we've had cases where practice said to us, we actually shut our phone phone down and we just went for voicemail because we were not able to pick those pick up those calls anyways. So what was the point? An average waiting time is over four minutes for a patient. So all these things, and just one more, which is for the practices, especially in the US, booking appointments, getting new patients, this is a revenue generating activity, because if you don't book appointments, you're going to miss your revenue, so it means that it's very difficult to deal with this traffic in any other way, and this is why we pivoted towards towards healthcare.
Jan Kamiński 5:50
Why is it so hard to solve for the practices, for the clinics? Because, as I assume, most of the other industries just solved it, by the way, right? What's so different about healthcare?
Ada Andruszkiewicz 6:03
This is a really interesting question. I don't know if anyone really knows exactly why, but people are traditionally, they have called their practices. It feels easier. So when we look at, you know, at least when we investigate why, they say it's easier for me to call. I don't have to install an app, I don't have to remember my logins. I just call because I just want to book an appointment. I just want to find something out quickly. And you know, nowadays, people are often on the run, so it's not like they will concentrate and set time set time aside. They do this when convenient. So by no means are we saying that this is the only channel you have to be multi channel nowadays. So most practices really have all their channels open to patients, because depending on the time and day, the time of the day and the ease of accessing the practice, people will choose a different channel. But you know, a lot of information is on the websites. People still call. People still call about what is you know, when are you open? Do you have parking? Is so and so available today? It's just easier.
Jan Kamiński 7:20
You mentioned answering information. The AI actually works in a way that it answers information about the organization, et cetera, et cetera, but it also deals with scheduling. So essentially, it's like a no show tool. Am I correct?
Ada Andruszkiewicz 7:34
We are deeply integrated with EHRs, so we can see both the patient chart. So when we identify the patient, we can see the whole history that we of course, need to see. So we limit our access to patient data, just to the data that we need, because this is important for data protection, but we see all the important information, and we also see the availability of various providers in this particular practice. So it means we can match the desired time for a given type of appointment with a given doctor. Because, you know, scheduling may sound really simple, but it's really complex. You have various providers dealing differently with various insurers. So as a patient, when you come in and say, you say you want to book with Dr John Smith, the practice needs to verify what is your insurance provider, because the rules may be different for this particular provider as with regards to your insurance. So we create those mappings so we don't do, like, a very simplistic approach. It's it's actually quite complex. How to match the right patient with the right doctor.
Jan Kamiński 8:55
The practices can implement token in what, like, two weeks, three weeks, what's the timeline?
Ada Andruszkiewicz 9:00
Yeah, the typical timeline is between two and four weeks. So in terms of the EHR that we are integrated with all, all of those are cloud based, extremely modern SaaS products, who are all really well prepared for working with providers like us and multiple partners. So at the moment, we are focusing on Athena health, on mod Med and Alation, and those are, you know, extremely, extremely well, set up products where they have extensive APIs, and when you integrate with them, You get a review, and you integrate with them with a purpose of realizing particular flows between the patient and the practice. It means you know what to expect in your API, and they know what to expect from you. And the beauty is that it really is just a flick. Of a switch to make this work with some traditional systems, or, you know, more, more kind of old school on premise systems. This is a lot more complex, but when these particular EHRs are, you know, they have huge marketplaces with hundreds of partners like us, they're extremely well prepared.
Jan Kamiński 10:23
Okay? And what about the patient adoption, in a sense? I mean, what are your experiences? Because when I'm, let's say, talking with an AI agent on the what I'm calling a bank, for instance, well, I mean, it's okay, but probably a run of patients after, after a few minutes, and I when I'm dealing with the doctor, it's probably way longer. I mean, not not another, not a doctor, but, for example, registration, it's probably longer. So how do patients react?
Ada Andruszkiewicz 10:51
So this frustration that you're talking about is is precisely the reason why we moved away from a horizontal approach. So we also started as a provider of a platform and service of building voice bolts or virtual assistants for different industries, and we found the same frustration, which is, if you approach this in a more kind of service or agency way there is never enough time and never enough money and never enough knowledge to create a one off custom voice assistant for a particular customer. And this really frustrated us. So we thought, actually, rather than move from one industry to another industry, we are going to focus on one industry and really perfect the way we deal with the patients, or the way our voice bots talk to the patients to avoid the frustration and all the issues that we had experienced when we were in an agency. And basically it boils down to the quality of understanding what the person says to our bot. And there are two elements to it. First is the speech to text element, and then there is the what does the person mean? Element, right? There's just two things. Because first, the way these agents work is, first, your voice has to be converted, right for the text is their input.
Jan Kamiński 12:19
How do you handle that? Because that's usually slow to convert that into voice later on.
Ada Andruszkiewicz 12:24
Well, the beauty of progress in AI is that it's getting faster and faster. So, you know, we started a couple years ago, and we were, we kind of witnessed the staggering speed of change, and especially recently, you know, you feel like you're in some kind of a hyperdrive rocket just viewing those stars. You know, move move past, so that that's the really important topic that you touched. Because, yes, with chat bots, you could afford to be slower, but in real time conversation, you have to be really, really fast. So the quality of understanding and one thing, and then the latency, the speed of understanding and processing this information, is another thing. So those two we have witnessed really improve and really shorten. So the conversation can be, of, you know, much better quality. And then there's two more elements, the way the voice feels, you know, is it natural? Does it feel pleasant? And this has, this has also, we've witnessed amazing transformation and developments here, you know, within the kind of traditional text to speech providers like Google or Amazon, but also the entrance of players, like 11 labs open, AI, you know, where the voice can breathe. You can hear it breathing. You can hear it, you know, making little noises like you can hear it laughing. Like this has not been possible until these guys came up with these, you know, amazing technologies that we can now leverage. And then, you know, to think about what doesn't frustrate the patient is when all these things are in place, and also they get the desired outcome. So the bot can actually do what they asked it to do in the first place, rather than saying, Oh, I can't help you with that. Let me transfer the patients they know that they talk with AI, right? Yeah, absolutely. Like, we never conceal this, because it just frustrates people, and because voice assistants are more and more accepted and prominent, we never, never conceal this. It's, it just frustrates people. But you know, if everything that we just discussed is in place. People can get their stuff done. They don't really mind. They don't really care if it's a human or it's a really successful voice board. They only or voice assistant. They only get frustrated if those things are. Not in place. We make mistakes, we don't understand them, or we cannot help them, and they have to go through a human anyways.
Jan Kamiński 15:07
How is it with the adoption? I mean, how do you convince practices and clinics to implement an AI solution, essentially to their workflow?
Ada Andruszkiewicz 15:15
Practices approach us at a specific point, usually something triggers the fact that they approach us, and often the trigger is that the practice wants to expand. So they're thinking, Okay, we have one or two locations, we would love to open a third or fourth location. And this means we have to do something about the phones, because by the time they have multiple locations, usually they have a team answering the phones, and the choices they face are hire more people locally, and they already struggle with that. It's already hard to maintain the team that they have. And so then they move to other choices. Should we hire offshore? And this really scares them, because it's not easy to work with a, you know, a call center team offshore. It's usually a team in India or a team in a Spanish speaking country. Perhaps the Philippines. And practices are not international businesses. They work in a particular location. So doing stuff like this really scares them. So they are thinking, either we drown in these calls or we have to do something. So this is, this is kind of what triggers their conversation with us. The other trigger they have is if they, for example, extend the scope of their specialties, if they go for multi specialty approach rather than a single specialty approach, and they're thinking, oh my gosh, now we're going to deal with twice, three times as many calls. So usually they approach us at a point where they are already feeling the pain. They're already missing 50, 30% of calls, and they know that it's only going to get worse. So this is usually the beginning of our conversation.
Jan Kamiński 17:04
When they hear AI. And I mean, your tool is predominantly based on AI. Are they cautious? What's the recurrent reaction? I know the market changed last year, especially in the last six months, but still, I know these clients are hard.
Ada Andruszkiewicz 17:17
Yeah, no, of course. I mean, they their approach really depends on their previous experiences. So if they had bad experience with AI products, they will be more cautious. If they had positive experience, they will be more outgoing. But when they approach us, they already know that we do AI based virtual assistants. So they're not like, you know, they don't they're not shocked. They usually look at multiple providers, and they're often really well educated about what these solutions can bring. They have a set of questions they ask about security certificates. So yeah, there, I think now, when we recently visited an event organized by one of our partners, modmed, we were approached by practices who said, Oh, you guys do AI. This is why we came to talk to you, because we know it can, you know, we believe it can solve some of the challenges that we have. So I think it's no longer something that people are scared of, because people started understanding that AI is a tool to help them solve their problems. It's not some science fiction concept, it's another tool that's, you know, now mature enough to use in healthcare,
Jan Kamiński 18:45
And you mentioned security and probably some compliance, did you need to go through a lot of that, I mean, of certifications, etc, in order to be able to serve these clients and these this industry?
Ada Andruszkiewicz 18:56
Yeah, definitely. We got our SOC two type two audit last December. We got the type one back last year in April, so I guess it's is exactly a year from there. But we wanted to be secure, so we we wanted a framework that would help us understand what way of dealing with patient data is a secure way so SOC really helped us to kind of organize our thoughts and our processes and all the various areas of how you run a company in a way that protects patient data sensitive data, and makes our applications and our service really secure. So the more aware practices always ask us for the SOC two audit report, and we often have to undergo security. Audit by their security teams.
Jan Kamiński 20:01
But looking at this from a, let's say, practical approach, I know that when you're talking, I mean, when there is a chatbot or a voice bot, a patient, usually they don't really share any medical information. They rather talk about an appointment, etc, some basic information, so that there is no real chance of any security breach or any compliance breach to in this regard. But what about, what if they actually start talking about it? I mean, do you How does it usually work? A bot just says, Hey, we can't talk about this. Or what's, what's the process?
Ada Andruszkiewicz 20:35
So in terms of the first part of what you said, Then they that specialty of your doctor, and the medication you take is definitely sensitive data. So if you think about it, if you want to book an appointment in the psychiatry practice, you may not want to know you. You know this may be a sensitive topic to you, right? Or perhaps I don't know reproductive health or any other endocrinology and but, but overall, the fact that you're booking an appointment with a particular doctor indicates already that you have a health issue in this area. But it this is already sensitive information. We also conduct conversations to refill prescriptions where you tell us what what drug you want to refill and dosage. You know this is already sensitive information. So even dealing with people's names, surnames, their addresses, you know, you have to be super careful. So we don't really look at the fact, I mean, we don't, for us, the fact that people don't talk about, you know, they don't go into the complexities. It doesn't mean that we can be unsecure about processing your other data, but we actually ask people to describe some of their symptoms, to help them book an appointment with the right doctor, right so to direct them whether it's a follow up or it's a first appointment. So we do ask these questions, and this is why we have a set of rules for dealing with such data, you know, we never send them via email. We always we have, like, really strict access control to the live data, multi factor authentication, you know, we encrypt all the data in transit and in rest. So all of these things are very important to us. And you know, we do various tests of our own infrastructure, penetration testing, like simulated attacks. We constantly provoke our guys, our employees, with simulated phishing attacks. So they are really, really cautious about, you know, clicking any links or sending any data in any unprotected way.
Jan Kamiński 23:06
There is this concern that this automation could, to some extent, dehumanize healthcare, right? I mean, especially that it's a lot of patients they want to talk with the doctor. They want to talk with the registration they have this. I don't know why, but it's, it seems that they, they do. So how do you see the the voice AI actually enhancing it rather than maybe diminishing the human element of care?
Ada Andruszkiewicz 23:30
Well, again, this is, this is a super interesting question, and I, you know, we spend a lot of time thinking about it, but when you think about what does it mean to dehumanize. In the history of humankind, a lot of people have done things that dehumanize humankind. But here, if you think about it, is it dehumanizing contact? If you actually facilitate smooth contact, you save people's time, save the frustration on both sides, and actually enable someone to book an appointment without waiting and see that provider. Because I don't think anyone is saying that. We want to put a barrier between the patient and the provider. It's the other way around. We want to remove that friction. But also, if we think about the other side, the people who actually provide the service, if we give them some breathing space to deal with those more complex cases that patients have, we are actually helping them to provide this service in a, you know, much more humanized and relaxed way, rather than rather than having them conduct three conversations at the same time. Because this is something we don't realize. The people who pick up the phones, they rarely have one conversation at the same time. They often have three conversations at the same time. Right? Because they pick up the phone and they tell you, please, hold on. Then they put the phone down and they carry on another conversation, which is at another point. And this is extremely frustrating for both sides. But when you think of any technological advancements, like if you think about X ray, you know, before x ray, the doctors had to touch your abdomen and touch her limp to guess what happened. And I don't think anyone would say that x ray dehumanized healthcare, because I would much rather have an x ray then you know someone guess what is happening in my thyroid or in my kidneys? So as long as we think about AI as this is a tool that's facilitating something that's solving a particular problem. It's not sending in the way between the patient and them getting helped, but rather, it's removing that friction and creating that channel. I really, like, really honestly, don't believe that it has any potential to dehumanize our healthcare. I see that it does as an enabler.
Jan Kamiński 26:12
Where do you think it's heading? In a sense, not only the AI voice bots, because I understand that where it's currently heading is where the bot can actually do things, and the more they can do, then there's an advance. There are some advancements, like they can actually make action for when you ask them to. But what's next to you? What's your take on that?
Ada Andruszkiewicz 26:36
So, you know, we used to be able to predict the future a bit more, but now again, with the speed of advancement, it's just, it's just amazing how quickly things work so but having said that, what we think will happen, and we see this happening, is that at the moment, of course, voice assistants can carry, you know, they can conduct conversations in these, let's just say, according to some predefined scenarios. And of course, we do use llms in our conversations, but we don't give them the whole control yet. They're just not ready. But what we where we see our product going is a becoming a bit more self sufficient when it comes to identifying what it knows and what it doesn't. Because at this you know, at this point, when you think about an AI agent, as you would of an agent who is learning on the job, there are things they will not know, and they would proactively ask or proactively try to fill out their knowledge right? So if you encountered three patients whom you were not able to help, you would proactively try to, you know, bridge that knowledge gap. So we are building mechanisms into our voice agents so that they can be more self sufficient in actually bridging those gaps and filling that that knowledge. Also, we see that people are more and more ready to give some space to agents so that they can be not just reactive, but proactive. So if you think about an analysis that AI could do of your patient base, who's done a checkup, you know who is due a follow up, who is due a colonoscopy or mammography, if you gave agents autonomy in actually contacting those patients and saying, Listen, I've looked through your record, you haven't booked this appointment for a while, but not it wouldn't be triggered by a person, because now you can do that with our technology. But more like you give them this data, and you say with the right prompt, figure out who to call when and who to book them with. So these are these are definitely some things that we see happening and we are working towards them so that that they will be possible.
Jan Kamiński 29:05
That's an interesting vision, I must admit. So essentially, some bots upselling the cleaning services as well, right? To some extent.
Ada Andruszkiewicz 29:14
Well, not really upselling, but more getting engaged in proactive, in more, you know, proactive outreach. So yeah, of course it ends up being an upsell. But really, at the moment, we don't have time for preventative care. You know, so many patients would have booked an appointment for a checkup if someone had reminded them, but you know, it's so easy, like I've recently visited my dentist, and he said, Do you know when you last came for a checkup? And I said, I don't think I want to know. I'm scared. Don't tell me. But if people were more proactive and now they simply don't have time to call the patients.
Jan Kamiński 29:54
No, no, I agree completely. I just I thought, because I know, similar to. Solutions from other industries that are actually getting there and doing this, but not in healthcare. As you said, you mentioned dentists. I just wonder, are you also available for other, let's say, slightly non regulated markets, like dentists like beauty as well, or not? Or at