Is the healthcare system drowning in paperwork and phone calls? The U.S. healthcare system spends over $80 billion annually on administrative tasks, a hidden tax that slows down patient care and burns out clinicians. Much of this cost comes from endless, inefficient phone calls for things like prior authorizations and benefits verification. This episode explores a transformative solution using AI for healthcare administration to automate these frustrating conversations and create a more efficient, "weightless" healthcare system. We're joined by Ankit Jain, the founder and CEO of Infinitus Systems, who is pioneering this change.
In this deep dive, we uncover the true scale of the administrative burden in medicine, where physicians can spend up to 15 hours a week on the phone for tasks that 98% of the time result in an approval anyway. This friction not only creates delays but can even have a chilling effect on care, as providers may avoid treatments they anticipate will be a bureaucratic nightmare. We explore how Infinitus Systems is leveraging sophisticated voice AI in healthcare to tackle this problem head-on. Ankit Jain explains how their technology, born from his experience at Google, goes beyond simply making calls. It navigates complex Interactive Voice Response (IVR) menus, engages in real-time conversations with payer representatives, and uses a proprietary knowledge graph to push back against incorrect information, correcting errors in real-time. This focus on prior authorization automation has already saved over 100 million minutes of conversation time for their partners.
We also discuss how to reduce healthcare administrative costs and the critical importance of building trustworthy AI in healthcare. Ankit details the safety guardrails Infinitus has built to avoid "hallucinations" and ensure every action complies with pre-approved protocols and regulations like HIPAA and SOC 2. We explore the difference between a fully autonomous Voice AI Agent and a Voice AI Co-pilot, which works alongside human staff to eliminate tedious work like waiting on hold. Finally, Ankit Jain of Infinitus shares his personal journey into health tech, offering powerful advice for anyone looking to build a career at the intersection of technology and healthcare.
Introduction
Dr. Tim Showalter: Hey, Paul.
Dr. Paul Gerrard: Hey, good morning. How are you doing?
Dr. Tim Showalter: Not too bad. I'm a little embarrassed that we've had our first mispronunciation snafu.
Dr. Paul Gerrard: Oh, I'm proud that it took us this long to have it.
Dr. Tim Showalter: Infinitus, not Infinitus. We should convince them to just change their name to Infinitus, just to make it easier.
Dr. Paul Gerrard: It'd be easier for us. I don't know about for them, but I agree.
Dr. Tim Showalter: I guess we should warn our listeners that we don't discover the error in pronunciation until our interview. So you're going to have to get through a lot of infinituses before discovering that it's actually pronounced infinitus. So as a PSA for the company, we should say it's infinitus.
Dr. Paul Gerrard: Yeah, so our apologies to the listeners and the company for the mispronunciation.
Dr. Tim Showalter: We should always apologize to the listeners just for making them suffer through it.
Dr. Paul Gerrard: That's a fair point.
Dr. Tim Showalter: Speaking of that, I guess let's hop on and torture them some more.
Dr. Paul Gerrard: Sounds like a plan. So this is a podcast on how to pronounce things correctly in the English language?
Dr. Tim Showalter: That's right.
Dr. Paul Gerrard: Welcome to Health Tech Remedy. I'm Paul Gerrard. I began my career as a physical medicine and rehabilitation physician and later focused on reimbursement policy and AI-enabled diagnostics. Missing in action today is Trevor Royce, a radiation oncologist and researcher specializing in real-world evidence, informatics, and AI diagnostics.
Dr. Tim Showalter: Well, Paul, you've still got me here today, so we'll be doing this together. For our listeners, I'm Tim Showalter, a radiation oncologist, former med device entrepreneur, and now devoted to bringing AI into healthcare. And today, Paul and I are taking a close look at Infinitus Systems, a company transforming the way healthcare organizations handle one of the most frustrating and inefficient aspects of their operations, phone-based administrative workflows.
The $80 Billion Problem: Healthcare's Administrative Burden
Dr. Paul Gerrard: I think prepping for this, I read somewhere that the U.S. Healthcare system spends over $80 billion annually on administrative costs. I don't know how true that number is, but I think everybody can agree it's a big number that we spend on administrative costs, both in terms of money and just time that probably could be spent on better things like caring for patients.
A major contributor to this is phone-based workflows. And to listeners not familiar with healthcare, yes, there's still a lot of things done by phone. We can't just text each other for everything in medicine, although we still use Pager sometimes too. These are calls between providers, payers, pharmacies, and it's to handle tasks like prior authorizations, benefit verification, prescription management, things that it would seem could be automated and computerized, but we just aren't there yet in healthcare.
They're not flashy problems. They're expensive and they're time consuming. And so anybody who can do something to solve that, even if it's not a flashy solution, it's just practical and there's a lot of value to it.
Dr. Tim Showalter: I think this is a good call out, really in healthcare. The things that are most efficient, I think happened within the EMR. And when you're busy and in your midst of your clinic day, a lot of your documentation, your billing, you're consuming information and digesting it and making recommendations for patients, it's really all within the EMR.
It seems to me like whenever as a clinician you're hopping on the phone, it's because there's a friction point that's slowing you down and it's going to drag on the other tasks you have to do and you're really not scheduled for it. I think a really good example for this that I know Infinitus Systems is involved in is prior authorizations. I think that's a really big example. As a radiation oncologist, it's something that a lot of my colleagues spend time complaining about, I would say, and identifying as an area for optimization.
There are some reports that physicians spend up to 15 hours a week dealing with prior authorization. I haven't ever seen such a burden in my own practice as a radiation oncologist, but I think that certainly may vary from region to region and specialty to specialty. And it does generally cause some delays, and that can lead to frustration on the side of providers as well as for patients.
A single prior authorization at the staff level can take more than 15 minutes of staff time. And if you multiply that across all the encounters and think about what staffing looks like, it's clearly a contributor to inefficiencies. And ever since the COVID era restrictions have lifted, I've read that the volume has surged even more in terms of the burden for this.
Dr. Paul Gerrard: You call the prior authorization line, you get the robot, which tells you our menu options have changed when they really haven't. You go through various menus. Ultimately, it cuts you off and you get dumped after going through menus for 15 minutes. And that's really a hidden tax on the system. It's not just the direct time of the provider. As you mentioned, Tim, it's not scheduled time. You don't necessarily have it baked into your day.
Sometimes it doesn't happen that day. It happens a couple of days later. The other thing is it also has a chilling effect on care. This is certainly not ideal, but at the end of the day, if providers know, if they anticipate, hey, there's going to be all this friction to getting prior authorization, then there may be decreased tendencies to pursue some diagnostics or treatments. If those diagnostics or treatments are not ideal for the patient, that's okay. But in a perfect world, really what's perfect for the patient should be driving medical decision making rather than anticipated friction in the system.
Introducing Infinitus Systems: A Vision for a "Weightless" Healthcare System
Dr. Tim Showalter: I think that's spot on. Our systems and our jobs are not really structured in a way that we have the resources to spend that much time on the phone and deal with this. I guess that's a good transition point for talking about what Infinitus' vision is and what they're solving for.
This company was founded by Ankit Jain, who we're going to speak with later, who's previously at Google, and Shyam Rajagopalan, who is an expert in cloud systems and machine learning. And Ankit is the CEO right now, and he previously had experience with a voice AI tool developed at Google. And my understanding is that that's what led to the spark for the idea.
So I think the story that's been told about the founding of Infinitus is that he was having a conversation with his wife, who's a healthcare professional, and really helped him realize that the technology that had been developed at Google in terms of a voice AI tool really could find its clinical impact within healthcare. We're automating conversations to push through patient care and prior authorization could really deliver benefits for providers and patients.
Dr. Paul Gerrard: And so it sounds like their mission here is this idea of a weightless healthcare system, which I think is really a great vision, a great dream. What they mean by that is replacing repetitive phone calls with intelligent and compliant AI agents that can streamline workflows across payers, providers, and pharmacies. Imagine if the AI can do this instead of you as the doctor having to sit on the phone and work through things, or even having clinical staff go through this.
Dr. Tim Showalter: Why not have AI to AI?
Dr. Paul Gerrard: Well, I guess that becomes the question. Where is this going? Are we using AI to AI to solve something that could be solved with simple databases and connectivity? But I guess that's a bigger question.
Dr. Tim Showalter: That's fair. I think that's a good question. And maybe we'll get there. It does seem like, look, we're still using some degree of faxing technology in health care. And maybe pointing a voice based technology within the existing infrastructure is the way to go. It certainly seems to be working in terms of their company growth and looking at where they've gotten support from investors.
Looking at the published reports of their investment rounds, looks like in total they've raised over $100 million from top-tier investors. They've got Andreessen Horowitz, Google Ventures, Kleiner Perkins, Coatue, and even the Memorial Hermann Health System, with some pretty convincing rationale underlying their investment thesis. Particularly the last one for Memorial Hermann is a direct signal from the front line that validates that this technology matters and is working and really makes sense within the healthcare ecosystem and among its most important constituents.
How Infinitus Works: Voice AI with Trustworthy Guardrails
Dr. Paul Gerrard: My understanding is their primary product, their core product is FastTrack, which is this AI-powered voice agent that automates phone calls. And the idea is to bypass the interactive voice response systems, those menu trees that we have to go through, and enable speaking directly with payer reps to retrieve real-time data about coverage, authorizations, and claims.
And they have automated 5 million calls and over 100 million minutes of conversation. Let me just reiterate that final point. 100 million minutes of conversation. So these guys are not ubiquitously out there yet. And even still, there was 100 million minutes of conversation on administrative tasks to automate.
Dr. Tim Showalter: That's amazing. And I can see it being a huge time saver. And to put it in context for people, I think that providers have so few minutes to spare for these phone calls. And they're trying to accomplish a lot with a relatively small staff to get these things done. But ultimately, it's not just about speed. I think the more that we can have AI agents that are connected with the source data handle some of these conversations, the more that the data accuracy could be improved.
Certainly, there's a lot of information to digest. And so I can see that there's a real opportunity there to leverage AI. The data that I've read in terms of their published reports from Infinitus suggests that they are seeing a data accuracy improvement, and they've demonstrated that their workflows can be broadly adopted. I think I saw some number, like nearly half of Fortune 50 companies are implementing some sort of AI systems to improve workflows. And so I think it's a really promising trend overall.
Dr. Paul Gerrard: And it sounds like the technology stack is impressive. They have over 100 custom models, which are these fine-tuned large language models like GPT and Gemini, and a proprietary knowledge graph that maps insurance rules across commercial and government plans. So the idea is that they can have these tools that can power real-time context-aware conversations.
Dr. Tim Showalter: Obviously, they're working within highly regulated spaces. I did see some press releases and sections on their website about working within really a safety first or AI for good type system of guardrails. And I think that's really important for this space. Of course, you want to leverage the technology and the large language models, but you have to make sure that the information has been pre-evaluated and is high quality, and that you're avoiding hallucinations, and that every output matches with pre-approved protocols.
Obviously, that's what's important in healthcare. And although these are not in primary treatment recommendation tasks, they're still important to the overall healthcare outcomes continuum. They're really leading the way in showing how to leverage that AI infrastructure and push things forward.
Dr. Paul Gerrard: That's a really good point. You don't want to have hallucinations. I guess at some point, if you're getting prior authorization based on information that isn't real, that could have serious consequences down the line. So having guardrails against that is important. And so it sounds like maybe one of their strategic advantages in this world where lots of AI and LLMs are coming out is this ability to impose regulatory constraints and build trust into the product that way.
Dr. Tim Showalter: They do have listed on their website that they've gotten through their HIPAA compliance, Sock 2, high trust. And I think having that experience and track record of deploying and being trusted partners within the healthcare ecosystem is very important. And I could see that that really gives them a bit of a barrier to entry compared to other companies that come in. Paul, I think you looked at some of the other integrations they have within the ecosystem. Do you want to comment on that?
Dr. Paul Gerrard: Integrations are important because they help you get out into the market and make the product more readily usable. And so they have integrations into Salesforce, Health Cloud, MuleSoft, and Google's Vertex AI. So they're becoming not just this AI piece that exists on the side, but more integrated into the healthcare infrastructure. And it looks like their strategy goes beyond replacing phone calls. That's obviously a big value add right now, but potentially down the line, we are talking about full end-to-end automation of a lot of the tedious and time-consuming administrative tasks in healthcare.
The Future of Healthcare Communication: Opportunities and Risks
Dr. Tim Showalter: What's interesting to me is that I think this is definitely an area where healthcare providers and hospital administrators, and I imagine payers as well, would love to streamline the process. And these are not phone calls that people have time for, nor do they enjoy. For example, I love talking to other doctors that are in the prostate cancer multidisciplinary group with me in my practice. And I love fielding calls from colleagues at other places, but I don't love getting on to prior authorization calls where I know that it's basically just have to clarify some information.
And what I would say is that most in most instances, when I get on that phone call, it just turns out that the right information has not been provided. I can imagine that an AI agent would do a really good job of connecting that payer staff member who needs to make a decision with the right information. So I think it's a great opportunity to really use technology to streamline the system. And I can imagine it being really impactful.
Dr. Paul Gerrard: I agree. And I think one of the things that we talked about earlier that I think is exciting here is they're not just talking guardrails. It sounds like they're really trying to, they think they can get seriously and implementing it to make this trustworthy AI. And I think that's one of the kinds of things that takes AI from being a toy to a real usable tool.
Having said that, on the flip side, for this to be broadly usable, you have to scale across lots of payers. Even one doctor is probably dealing with 10 to 20 payers in a given month. If you go to doctors all over the country, we're talking hundreds, maybe thousands of payers. I imagine that's going to be non-trivial to scale and make your AI work across so many. Especially if the menu options really do change.
Dr. Tim Showalter: I think that's fair. And I also think it's hard to predict what direction healthcare is going. And we're always looking for paradigm shifts and they take a while to come in healthcare. But to your point that you started off with before is, what if this is throwing voice-based technology at something that ultimately needs to be fundamentally changed and there need to be tighter integrations across systems?
I could imagine that the risk for Infinitus systems is that their voice-based technology could be replaced by something. On the other hand, though, to mitigate that risk, I kind of think that if they're providing a voice-based solution, it seems like that's probably harder to do than a simple technology integration. And it also occurs to me that maybe they're being really smart here to position themselves as the key technology-based intermediary, and that if the future changed and healthcare were more streamlined, they'd be best positioned to deliver on that.
Dr. Paul Gerrard: I agree with that.
Dr. Tim Showalter: Well, I guess, Paul, thanks for chatting to me about this. I guess let's hop over to our conversation with the CEO, Ankit Jain, who's joining with us next. And we'll dive deeper into how he built the company and what are the hardest challenges that he's had along the way.
A Conversation with Ankit Jain, CEO of Infinitus
Dr. Paul Gerrard: We're joined today by Ankit Jain, founder and CEO of Infinitus Systems. Ankit, welcome to Health Tech Remedy.
Ankit Jain: Thanks for having me. And it's Infinitus Systems.
Dr. Tim Showalter: Oh thanks for cracking us infinitus.
Ankit Jain: Infinitis would be a terrible disease to have, and I really don't want a disease.
Dr. Tim Showalter: It's just an inflammation of infinity.
Dr. Paul Gerrard: I don't want any disease with the word infinity in it. Okay, with that, I'm Paul Gerrard, and luckily, my expertise is in healthcare policy and reimbursement rather than how to pronounce words in the English language. And I'm here with my co-host, Tim Showalter, and we're excited to learn more about your company and your journey through technology and healthcare.
Dr. Tim Showalter: Ankit, thank you for correcting us on the pronunciation. It's only going to get better from here, I guarantee you. So stick with us for this interview.
Ankit Jain: I always think about it and I go, I wonder how people pronounce Google in their early days. I'm sure Sergey and Larry had to go and say, it's not Google, it's Google. Hopefully at 20 years from now, Infinitus will be something that everybody in the world goes, of course, that made healthcare so much better. And it's to Infinitus and beyond.
Dr. Tim Showalter: Well, so I think that's a fair, fair way to start. So I'm a radiation oncologist, and I thought a good way to frame this discussion would be, I can tell you about Monday afternoons as I'm wrapping up my clinic, usually it's around 3:30 or so, and the machines are still treating patients, I'm still getting paid, blah, blah, blah, and I start to go through my email. And one thing I really dread is getting those emails saying PET scan was denied or this particular radiation therapy code was denied prior authorization reviews required.
And they just give me the patient's name, member number and the phone number. Then you dial and it's like choose option four and you start off there and I have no time blocked in my schedule as a clinician to make these phone calls and I dread them. They're always nice when you get on the phone with them but I wait on hold. I'm usually talking to someone who's not the medical director where it needs to be reviewed for. And I will tell you that more times than not what I found out is that the great staff at my hospital has sent a ton of information.
Some of which includes the information that the payer needs, but it was so much information it was hard for me to get through it. And some of it was contradictory. And then ultimately, after 15 to 30 minutes of my time, that decision is approval for the patient. And that's just an example of one of the phone calls that we just don't have time to do and don't enjoy doing as physicians. And I'm wondering if you've ever heard of a story like that before.
Why Prior Authorization is Broken: The "Fraud, Waste, and Abuse" Tax
Ankit Jain: It's fascinating that you ask that, Tim, or start with that, because the reality is that eventually, over 98, maybe 99% of prior auths eventually get approved. So the whole system is working incredibly hard to eventually have a less than 1% or 2% denial rate. So then the question is, why do you need that? And it's really more than a prior auth or anything. I like to describe it as more fraud, waste, and abuse. You really want to prevent the 1% to 2% of fraud, waste, and abuse from happening.
And that's why 98% to 99% of folks in the medical system are paying the tax for that 1% or 2% of fraud, waste, and abuse. So I hear those stories every single day. Every clinician, every administrative worker has waited on hold for hours to provide the information and say, hey, on page 72, I gave you what you were asking for. When the person says, oh, sorry, I was going through this 200-page fax that you sent me and I didn't look at page 72.
So there's so many pieces of this that need to be improved. It's not just the phone call. The phone call is one piece of it and we're excited and proud of the work we're doing here. When you think about prior authorization, it's about figuring out what is the real proof of medical necessity, who determines what those medical necessities should be, and what is the mechanism by which you can instantaneously share the data from the provider to the payer in a way that the payer can almost instantaneously understand it.
Adjudicate it, and 95% of the time or 98% of the time approve it, and then really escalate 1% to 2% or 3% to 4%, not 60% of the time or 40% of the time that we see depending on payer today. So the opportunities there on the provider side, the opportunities there on the payer side, and the opportunities there for all the startups and technology partners and vendors who are trying to do this in the middle to make this data transfer a little more manageable.
From Google's AI Fund to Solving Healthcare's Biggest Headaches
Dr. Tim Showalter: That's fantastic. And I'm sitting here listening to you as a practicing physician, and it sounds like you've really developed a deep knowledge of the friction points and the real need to get information shared more quickly and to remove a lot of these bottlenecks. We all want to hear a ton more about what you're doing with Infinitus, but I would love to go back actually and discover a little bit more about how you came to learn so much about this problem. My understanding is you have a technology background and I'd love to hear more about how did you end up in healthcare? What led you to found the company and leverage all your technology background into this particular problem?
Ankit Jain: I have three backgrounds that I think bring all of this together. The first is the technology background and we'll talk about that. The second is my wife has been working in healthcare her entire career. And the third one, like both of you, I'm a human that has health issues and I have a family that has health issues. So healthcare issues are something that all of us face. And so the intersection of those three is what brings me to where I am today.
So growing up, I studied computer science and business. Specifically the math that goes into much of machine and deep learning and the current generation of LLMs, all the linear algebra that makes those systems work the way they do, the algorithms run the way they do, and played a role in the startups I've done in the past, the roles I had at Google, both as an engineering leader for Google Play, but also the person who helped start Google's AI venture fund, Gradient Ventures.
In all of those, I was increasingly getting exposed to how technology and how large amounts of data could be used to drive smoothness in industries and to solve problems where lack of data transparency was at the root of all problems. And when we think about what we're doing at Infinitus, while we're building cutting edge technology to have really complicated clinical and administrative phone calls, some of which are 35 minutes long, or actually we had one call that was three and a half hours long a couple of months ago. It's crazy to imagine a machine talking to a human for that long.
But the fundamental issue is data transparency, data access, and data exchange. That's the thing that we're excited to solve at Infinitus, making it so that the system is transparent, making it so that the healthcare system is proactive rather than reactive. We've all been a patient or a caregiver where we have to pound the table for somebody to move the process forward. We have to ask, hey, why isn't that MRI scheduled? Or is there any way for me to have that procedure done a week ahead? And then we're stopped by prior authorizations, or we're stopped by the fact that the specialist isn't available in a way that makes sense with schedules. And I think there's an opportunity for technology and AI to really play a big role there.
Defining the Solution: Voice AI Agents vs. Voice AI Co-Pilots
Dr. Paul Gerrard: That really takes us to what your company does. And could you tell us a little bit about what you do and how you are tackling some of these problems that you've identified? And I just want to comment, I can't even imagine a three-hour long prior authorization call. That is stories I would tell my children to make them behave.
Dr. Tim Showalter: Thank God it was an AI agent.
Ankit Jain: Luckily, my kids hear stories about made up characters and not about call center agents who have to answer these or make these calls all day long. But it's the reality of almost a million workers in this country every single day, which is scary. But at the heart of it, what we have built at Infinitus is a platform to build any kind of voice AI agent and a whole set of voice AI co-pilots. So when I say a voice AI agent, it is a piece of software that can automate a phone call from start to finish.
This could be an outbound phone call to a patient on behalf of a Medicare Advantage plan to do a health risk assessment. It could be on behalf of a provider to do a pre-surgery assessment. Prep call, it could be post-discharge follow-up call, or it could be a back office phone call between a pharmacy and a payer to do a benefits verification for a specialty drug or a prior authorization status check.
Those are on the outbound side. On the inbound side, it could be that first line of defense when you call a patient assistance program for a pharmaceutical manufacturer's drug, or it could be the inbound phone call when you call your local health system and you want to ask about the status of the procedure that needs to be scheduled. So inbound and outbound phone calls, we've built a whole suite of voice AI agents, administrative and clinical.
In addition to that, we have voice AI co-pilots. To me, co-pilot is a piece of technology or software that is sitting side by side with the healthcare worker to drive them to be more efficient. So in the revenue cycle world, we've got AI co-pilots that go through the payer's IVRs and wait on hold for an hour, two hours, three hours. And when that payer agent picks up, drops in that revenue cycle worker. So that if instead of doing 10 cases a day, they can do 20 cases a day. Because while they're talking to one agent, the next call is already waiting on hold so that they can go from agent to agent and be extremely efficient.
Another co-pilot is in the world of interpretive services where you've got a nurse that speaks English, a patient that speaks Tagalog. Well, AI has come a long way. There's no reason why the nurse needs to click a button, dial someone in to translate every line back and forth. AI can do that perfectly well. So just some examples of AI agents and co-pilots that we're excited about bringing to market.
Building Trustworthy AI: The Importance of Context and Safety Guardrails
Dr. Tim Showalter: Have you ever had your own two AI agents have a conversation?
Ankit Jain: Our hypothesis is in the back office, eventually you want to get rid of the AI agent. You want it to be an API connection. So again, I didn't come from the world of healthcare prior to starting Infinitus. I was showing my wife a demo of this thing called Google Duplex, where you could say, hey, Google, make me a reservation at a spa, salon, restaurant. And if OpenTable or something like that was there, it would do it through an API. If not, the Google Assistant would make the phone call to the business and talk and make a reservation.
And I was talking about the technology and she said, man, I wish someone would do this for healthcare because doctor's offices talk to payers all day long to get to move the process forward. And I said, why isn't this an API? And she said, because that's just not where healthcare is today. So our vision is over time, we replace the back office phone calls with API calls. So if we ever find ourselves in two places, the caller and the recipient, which we have in some cases, we replace that with an API call rather than have two AI agents talk to each other in English.
On the other hand, in the front office, when you're talking to a patient, you're actually giving them the gift of time. We've all had a call with a well-meaning nurse or physician. In their voice, you hear the empathy, you hear the answers, but you also hear that they're thinking about what they need to do next because they have so many things to do. So one of the opportunities with AI is to just allow that patient to have that time.
That patient is going through probably not a pleasure generating journey. A chief medical officer I talked to a few years ago told me every patient goes through three journeys. The first journey is the diagnosis journey when you're told what you have. And in that moment, you're with your physician, but you're not emotionally ready to ask the questions. You're just taking it in. What does this mean for my life? What does this mean for my dreams? At 10 p.m., you finally have your questions, but you don't have your physician. So what can AI do in that context?
The second journey is, okay, do I have the right resources going through the industrial complex of healthcare? Do I need to get a referral? Do I need to get a second opinion? Who do I go to? And can AI give you that kind of a pathway, a navigation? And we've got all kinds of roles in healthcare. My wife is a patient navigator. Her job is to help patients and their families through these journeys. But what can AI do at 10 p.m. our best people do in the daytime.
And then finally, the financial journey. Can I afford this? How am I going to make my life work if I need a specialty medication or a cellular gene therapy, or am I going to be bankrupt for the rest of my life? So there's many parts of this that the patient is going through and their family is going through a lot of angst. And I think AI being side by side with our healthcare workforce can really enrich that experience and make it as good as you can imagine it to be in that hard time.
Dr. Tim Showalter: I can see in some of these AI agent and AI co-pilot use cases that it's really critical to have AI review a whole ton of information. And I think that's one of the bottlenecks for humans to go through the, I think you said page 72 of the documentation to find the key shred of information. How have you thought about embedding access to patient-specific information or information directly from the provider in informing the AI agents and what solutions have you guys come up with?
Ankit Jain: A couple of thoughts on this. The first thought is that in the last few years, it's gotten increasingly easier. And I think this trend will continue to build a demo of anything AI-related. Any of us can slap together a prompt into GPT, Cloud, Gemini, you name it. And in the case of a voice agent, you've got speech recognition, your NLP system, and then text-to-speech, your speech synthesis, and now you're starting to see voice models, voice-to-voice models, where you just give it a prompt and it does everything else.
So prompt engineering is where you're seeing a lot of sexy demos, as I like to call them. The jump from there to something that's production-grade with the right guardrails, with the right ability to be trusted and hallucination-free, is getting harder and harder because these machines are getting so good that it's hard to tell when it's right versus making something up. And so we've invested a lot in the safety guardrails before the call, during the call, and after the call.
And Tim, to your question, having the right context is almost more important than having a decent prompt or a decent NLP stack. Because if you don't have the context, you're not going to be able to have a conversation or an experience that is anywhere near what you would expect from the healthcare system. And so we spend a lot of time getting the right knowledge bases, the right access to the right pieces of data, and understanding how to balance different styles of each customer into the in-call experience. Because every brand, every healthcare entity wants to show up as themselves. Every health system, every pharma company, every payer, every pharmacy has a brand that you have to enrich with your AI offerings.
You can't just say, well, you're getting the generic offering because that loses their ability. So we do the pre-call prep with the context. During the call, we are making sure we're collecting the right information, know when to push back. Just as an example statistic, when we call a payer to collect benefit information for specialty drugs, 25% of the time on first ask, we get incorrect information because the person is looking at the wrong benefit plan is looking at the wrong page. And so as we've done millions of these calls, we're learning what to expect. And that knowledge graph is used by our AI agents to push back and say, are you sure about that? Could you check that please?
75 to 80% of the time we push back, we get a different answer. That's crazy. You expect your best agent to do that, but the new hire or the person who doesn't have that detailed and understanding is going to get incorrect information, which by the way, if you get the incorrect information and benefits, you're going to get a denial in the claim that you'll try to fight. So all of this fraud, waste, and abuse on one side is actually mistakes being made upstream. So by fixing stuff upstream, you can solve downstream problems as well.
Dr. Tim Showalter: That's a much more straightforward problem solving solution that's not dependent on massive amounts of data. I can see how that would really have an impact into the workflow at a large scale. And this is where you've learned so much about that space. It's interesting to me as a clinician to hear about that. It makes obvious sense. But I'm picturing, in my mind, AI digesting 500 pages of a patient's history in order to interface with insurance for prior authorization. But what I'm hearing from you is that there are these even more common benefits investigations that are just at a more basic level and occupy millions of minutes every year. Sounds like a great solution for that.
Ankit Jain: So here's another philosophical debate that we're having internally. When most people design voice AI agents, they say, this is the goal of the call. These are the 10 pieces of information you either need to give or get in a call. And when you talk to these systems, you get a feeling that the system is driving the conversation. And the physicians on our team, the folks that are saying, that's not how a physician has a conversation with a person. They let the patient drive the conversation.
And so one of the areas of active research for us is how do you understand the goals, but then let the patient drive the conversation so it feels more natural that when we talk about bedside manner or being able to connect with the patient, it's the flow of the conversation that plays as much as the role. And so all that goes into the kind of design of the system and how it achieves its end goal.
Career Advice: Why Now is the Most Exciting Time for Health Tech
Dr. Paul Gerrard: We have a lot of listeners who are interested in the intersection of healthcare and technology and maybe are trying to pivot their careers. What career advice might you give to somebody who is interested in making a difference in that intersection of healthcare and technology?
Ankit Jain: A few things. One, I think it's the most exciting time to come into healthcare and technology. I have friends or mentors who were trying to do these kind of things 10 years ago from a philosophical perspective, and they said it was hard to deploy technology in healthcare. It was hard to get funded as an entrepreneur in healthcare technology. And it was hard to hire people in the technology world to work on healthcare problems because they didn't understand. They thought everything was behind regulation.
And from a technology deployment perspective, it's all gotten easier. There's a lot more money, investments, there's a lot more ability to deploy technology. And there's a thirst for to improve the system. The one thing I've learned in the last six and a half years being part of the healthcare ecosystem is that everyone wants to do right by the patient, because at the end of the day, we're all going to be patients. So when we improve the system, we're improving it for ourselves. It's not just for some other party out there. It's literally for ourselves. And so that fundamental alignment allows for forward motion. Sometimes it seems slow, but it's coming from the right place. And that's a really fulfilling and exciting place to be.
Dr. Paul Gerrard: Ankit Jain, thank you for joining us. It's been a real pleasure. And that's it for this episode of Health Tech Remedy. Be sure to like and subscribe on your favorite local podcast platform and share with friends and colleagues who may be interested. Look for us on LinkedIn and reach out with questions, suggestions, or just a chat.
Credits
HealthTech Remedy is produced and marketed by Podcast Studio X.
