ThinkAndor: AI that helps doctors (not replaces them)
39:35 Share

Welcome to today's episode of Lexicon. I'm Christopher McFadden, Contributing Writer for Interesting Engineering. In this episode, we sit down with Raj Toletti, Chairman and CEO of Andor Health, to explore how AI is transforming virtual care without replacing the human touch.

From reducing clinical burnout to accelerating time to treatment, Raj shares how ThinkAndOr is building a more intelligent, more connected healthcare system powered by AI agents that collaborate across every stage of care. If you're curious about the future of healthcare tech, this conversation is a must listen.

Raj, thanks for joining us. How are you today? Good, good. How are you? Very well, thank you. Thanks for joining us. For our audience's benefit, can you tell us a little bit about yourself first, please? Sure.

I'm a healthcare technology entrepreneur. This is my fourth startup. My mission in life was to actually help patients access care better electronically. I would also help care team members, clinicians, and nurses with their daily tasks, automate them, truly create efficiencies in the way they deliver care.

And so that's been my mission for the last 25 years. And that's what I'm still passionate about. Excellent. That's on the professional side. Yeah. Living the dream.

And what was it they say? You won't work a day in your life if you do something you love. Right. I think that's the saying, something like that. Probably just murdered that. Yeah. Yeah. Anyway, so on that subject then. So what specific gaps in health care delivery were you aiming to solve when you founded Andor Health? And how does Think Andor, I think it's called, address them today? Yeah, yeah.

So the vision was to actually have a set of AI agents. This was in 2019. So today we are on a video call and my vision was to actually have an AI agent actually start this conversation between you and I. And if I'm the physician and you're the patient, I would actually be able to bring up your clinical content right onto the screen.

While I'm talking to you, the AI agents are transcribing our dialogue. And then through the camera, I'll be able to observe your vitals and other

types of observations so that I can bring all of that together to provide better care for you as a clinician, but also bring other care team members and families. So my vision was to actually build a care collaboration platform that is based on a set of AI agents. And so we set out to actually build the first

AI software infrastructure for healthcare that is cross-cloud compatible. So some of our customers are on Azure, Microsoft, or some customers are on Google. Our AI agents and our AI software infrastructure for healthcare can be published on any cloud and leverage the best of breed capabilities from that AI cloud. So that was the vision and mission

for hand or health late 2018 and early 2019. Fantastic. How much, is there a limit to that though? You say I could potentially diagnose your vitals through the camera. Presumably you can only go so far. It's a bit like triage, I guess. At some point you're going to have to physically see. Sure, yeah.

I think AI in general is the AI agents, the agentification that we've done, they're adjunct to a clinician. In some tasks that we are able to automate. So if you think about healthcare, I mean, you go see a doctor in your office, like their clinic,

And then if you're sick for whatever reason and you have to be admitted into the hospital, that's the acute care or the inpatient setting. And then once you get discharged, you might actually, if you're an older patient, you might go to a nursing home or you might go home. So

where we think AI starts and stops is basically, while at home and while you're trying to access care, I mean, our AI agents can help you, triage you, bring you to the right speciality, right? And then as much information I can automate from you, and then we can connect virtually. But then once you transition into a care setting like a hospital, I can have a camera in the room that's able to see that, hey, the room is empty, but

It's not marked as empty in the electronic medical record. We might as well tell someone to actually mark that as empty and increase capacity or somebody is falling or somebody is actually wanting a blanket because they're feeling cold. Today, all of those tasks are being delegated to a nurse practitioner or a nurse, a clinician.

So if you can automate a lot of that through vision, observe, listen, and then send the signal to the right care team, or it could be the housekeeping for that blanket because somebody is cold. So we can automate a lot of these tasks so that these clinicians can perform their duties at their license level.

It's an adjunct to the clinician, but in some cases it's also helping the clinicians make better decisions. And then once you transition into your home or nursing home, these agents still follow you. They will make sure that they prompt you to adhere to the care plan that you're supposed to so that you don't come back to the hospital.

and you get better. So since the agents actually understand your behavior, how we can actually engage you with the right rules of engagement at these different care settings, even post discharge, these agents are able to follow you and help you have a better outcome, whether it's from a surgery or from any kind of hospitalization, or you're just home and being treated at home. So

I know it's a long-winded answer, but I think the platform itself is very relevant in any given care setting and these agents are helping you as a patient, you know, for better, you know, engage and adhere to a care plan better. It's also helping the clinician automate a lot of their tasks and bring the right clinical content to the clinician with the right clinical context.

So they're much more productive. So we've been able to take out three and a half hours of nursing time per shift. We've been able to take out 180 keystrokes for nurses. We've been able to increase the capacity 1.5 times or reduce the number of days you stay in a hospital. We've been able to increase or decrease 65%

unnecessary visits to an emergency room because of these AI agents that are helping patients and the clinicians. So we see this being a way of life and instead of me as a patient or you as a patient trying to understand a lot of content or the clinician trying to go into an electronic medical record and spending a lot of time,

If we're able to surface this clinical content in context for both parties, the ROI is significant.

Yeah, so it enables the clinicians to actually do their real job, which is looking after your health rather than all the nonsense around administration, paperwork, whatever. I got you. Excellent. Brings us nicely into the next question then. So AI, like generative AI, is still a new concept, especially in clinical settings, I presume. So how do you ensure it's deployed safely and meaningfully without overwhelming conditions or compromising care quality?

Yes, it's fundamentally the question to ask. What is responsible AI? Basically for us, and we are very passionate about delivering very responsible AI. We work with very large research institutes as well across the globe. For us, first, the AI should actually be demonstrating true value. Like I said, is it three and a half hours of time savings?

Are there no false positives? Is the AI giving the right materials to the clinician and to the patient? So we've gone through years of basically putting guardrails on the capabilities that we provide. And we watermark that and said, this is from Andor Health and this will actually give you these results based on the deployment, the best practices to deploy this type of an AI.

And so it is very, very imperative for healthcare organizations to look at research. So we at AndorHealth have made a consensus decision that we don't publish our own return on investment studies on AI. We actually work with our clients and our clients actually publish peer reviewed general publications based on the value that they've seen from our AI technology.

So it is imperative again for other healthcare systems to look at and say, what is my AI roadmap? What are the studies out there that have very, very definitive concrete return on investment? And is that AI capability very, very responsible from an outcomes perspective and then deploy those. So that's the tag that we've taken. And we are probably...

overzealous in terms of making sure that we deliver responsible AI to our client base and to healthcare in general. Okay. What kind of feedback are you getting? Is it support or resistance from specific areas like the public sector? I'm not sure. Yeah, I think there is a lot of, I would say, I mean, a lot of health systems want to leverage AI or just healthcare in general.

but they're also kind of very guarded because you know they want to make sure that you know they don't they do the right things to their patients to their caregivers so i don't think so it's resistance it's basically us being able to partner shoulder to shoulder with our customers and showing how this has actually been proven out in the marketplace and once that that is established

I think the health systems are very, very accepting of this technology. And in fact, I mean, there's so much need right now because there's burnout. There's nursing burnout, physician burnout, there's staff shortages. So these are all the right, I would say, patterns for something like an AI technology to come and provide significant relief to

for this industry so we feel that the industry is hungry to actually kind of adapt and adopt this type of ai technology but on this you know on the other hand they're really making sure that this is very responsible and and deploying it in that way yeah i'm sure like the frontline staff would sort of bite your hand off to get this kind of technology it's more convincing the management and the senior executives i presume they're pretty more of a problem

Yeah, I think, you know, first of all, I don't think so this is like, you know, at least our customers, they don't see it as

an immediate replacement of 10 people. It's more of an adjunct and it's making their jobs easier. It's increasing their capacity by 30 percent. Those are some low hanging fruits, but it's not just about cost reduction. It's about the AI technology. You as a patient, you come in and search for a vaccine. Then if I'm able to navigate you through AI to the right health system, now I'm increasing the patient

lower and so I'm actually increasing revenue. So there are, it's just not about, you know, operational savings, but it's also about significant revenue lift and being able to capture those visits that were never possible before due to capacity or due to the ability for call centers to actually engage the patient. So, so I think, I think that's where I think, you know, because healthcare, you know, the industry itself,

It's somewhat kind of like, you know, it's hurting from top line revenues for some of the health systems. And if you can actually go and increase the revenue, better the outcomes, and at the same time, improve patient satisfaction, because you as a patient now, you feel that there's always somebody there. You might not even know it's an AI, you know, persona that's actually engaging you. I'm not sure the statistics, but I think men are less likely to

go to a doctor or a nurse if they've got any kind of ailments. They'd rather just get on with it. Presumably a tool like this would encourage them more to kind of play more of an active role in looking after their health or if they've got a problem, a concern, go to an AI doctor to try and help diagnose a problem. I don't know. I mean, I think I agree, but I think men are probably, I mean, in many households,

the you know women are kind of the caregivers and and they are more responsible about their health than you know statistically you know this industry you know data out there that men are less compliant with their care plans into their care than women right now

Also, this type of a patient engagement where it is asynchronous, right? So if I have an agent that basically is able to understand that Raj is usually more responsive between 4 and 5 p.m. and is actually helping me with my care plan at that particular time, more susceptible to actually listen to that, you know, call to action and actually, you know, follow that care plan better.

And also, I think, you know, on the other side of the coin, the same AI agents also can actually prompt, like if it's my wife or my daughter that's actually taking care of my health, it also can actually connect that care team, which includes the family member, and create that collaborative, you know, fabric that can actually bring the care team members, including the family members together, to help me in being more compliant.

So I think we see a lot of that, I think. So the short answer is yes, I think it is. We've seen really good, good results. Actually, one result, one machine learning project that we've done is for people that have prostate cancer, men. And we've been able to actually use our large language models, actually took the largest data set of patients with prostate cancer for robotic surgery.

we've been able to predict days to continence days to potency and we were able to engage the patient pre-surgery and post-surgery and reduce and this is a journal published uh publication and reduce the days to potency and days to confidence significantly and so yes i mean you know we are seeing actual results on some of this and which is very very meaningful for men's health

Very cool. And presumably patients can use it if they've got any questions, queries or doubts about dosage or how much the doctor said I've got to take this medicine as well. Yeah. So I think when you get discharged from a hospital, that's the last time you want to learn about your whole education around your discharge. So you go home and then you have all these questions.

So we've been able to actually have these agents kind of learn, for lack of better words, the discharge plan.

and kind of engage with the patients to help them navigate that care plan so much more better. Yes, they can ask questions. And basically always, you know, we actually try to bring the clinician in the loop for the right reasons, right? But if there are certain things that are actually tacit and that can be explained to the patient, yes, the agents will actually do that.

Oh, fantastic. The next question then. So one of your stated goals is to shorten the time from diagnosis to treatment, which you've kind of touched on. Can you walk us through how Think Andor makes that possible in real world clinical environments, please, if possible? From diagnosis to treatment? Yes. Okay. Yeah. I mean, so we have many examples. I mean, we have a journal publication in Canada for sick kids, the largest pediatric institute in the world,

So if a child gets sick or there's some symptoms,

they can just go in and basically use our agent and intelligent adaptive triage. Basically, the agent is basically asking the patient and they're actually bringing the patient to the right point of care within that health system. So we've been able to cut significantly hours and days for that patient to get treatment. Another example I'll give you is

Medical University of South Carolina, they set up a virtual endocrinology and rheumatology offering. And before this virtual connection existed, it was months to get to an appointment. And now we've been able to reduce that to 10 days or less. So

So there's many an example of actually how this automation and this virtual collaboration are reducing that time to clinician from door to treatment in an emergency room setting. So it's really compressing that time and more importantly, it's scaling the backend, right? So it's actually, it's helping the throughput and optimization of the patient flow

within any care setting, basically. - Well, yeah, that kind of time saving would, yeah, potentially save lives. I mean, heart attacks are severe case. Yeah, the early you pick up on things like cancer or something, yeah, the better. - I mean, there's so many patient stories I can tell you. I mean, our agents, after the patient has been discharged, we kind of follow up and say, "How are you doing?" You know, so on and so forth.

We have stories where there's some mental health problems, the agents were able to engage and save lives for suicidality. There's just very, very neat stories and outcomes and 26 to 38% patient satisfaction because of this type of engagement, additional.

So, yeah, it's very gratifying to see because I come from a family full of clinicians, my parents, my grandparents, and I have 33 doctors in my family. So I think having that extra burst capacity, if you will, through these AI capabilities, it is such a pleasure to see and gratifying to see that there's immediate

So we don't have to wait six months on a technology to show return on investment. Month one, we're able to show here's what we've done for you, health system or pharma or payer. Fantastic. It's in the blood, as they say.

and no pun intended so virtual nursing and remote patient monitoring have become critical care delivery models as we've discussed what do you see as the next evolution of these models and how is andor health preparing for it yeah i mean remote patient monitoring you know historically has been first of all you know we're talking about all these different capabilities and and historically over the last 15 20 years there's been like point solutions there's a

a camera in the room by one technology company. There's a set of devices that are sent to your home for remote patient monitoring, your blood pressure cuff, your pulse ox and so on and so forth. That's another technology company that was providing that service or another technology company that was providing a digital front door in the primary clinic setting.

We believe there has to be one pane of glass. This is a platform play for virtual care collaboration. Why I'm going there before I answer the question about remote patient monitoring is, if you are engaging with a piece of technology in the clinic setting or in the hospital setting, where you're waving to the camera, feeling cold, you're about to fall, whatever that technology that you've now consistently experienced, and when you go home,

If you get a completely different experience and also a disjointed experience, we've seen that remote patient monitoring has not scaled as a capability in a lot of health systems. The number of patients that are being monitored through devices is also a very small cohort, statistically not that significant.

If you broaden that cohort of patients to people that have devices and then any patient that's actually getting discharged and that can put them on a care plan, we are now at scale, many health systems with tens of thousands of patients that are being monitored on a daily basis. So where we see this going is patient monitoring doesn't mean it's the set of devices I ship to the patient's home.

Patient monitoring could be you trying to access care and I monitor you and try to get you the right care. Patient monitoring could be I'm monitoring you and observing you inside the hospital. And then once you leave the hospital, I'm still able to monitor you with or without devices, just like we talked about before a care plan and say like, you know, make sure that you call the patient between 4 and 5 p.m. They're more engaged. And then all the devices that you need to measure your vitals.

if you compress them into like, you know, one or two apertures, then the usage of them we've seen significantly increase. So a lot of health systems that we are working with today, they look at us and say, okay, this is a platform player. So if I start with remote patient monitoring with Andor, I can also do the virtual nursing or virtual sitting or

telestroke, tele-ICU, or I can also put it on the clinic side so that I can attract more patients and bring them to my physicians. Or I can stop the patients going outside of my four walls. So again, you asked me a question about remote patient monitoring, but I think patient monitoring itself is a more broader capability in our minds. - No, this is fair enough, fair answer.

But so does it would it be reliant on wearable technology then? I think as you know today it's a way of life. I mean we all wear something you know that that's monitoring us. I think yes I mean the ability for the AI layer to be able to interpret these device data

And it's just not the data, right? So if I get an abnormal signal, what do I do with it? Yeah. Right? So that's where the ambient orchestration comes into play. I need to be able to document. I need to be able to orchestrate. At the same time, I have to observe you as well, right? So wearable devices, we feel that more and more will be highly effective from post-care perspective, right?

We also feel there'll be consolidation of these devices. But then what's really important is that software layer that actually is able to read or actually receive these signals and orchestrate the right care team into the right care setting, whether it's inside the hospital or at home or in pre-care. So would the ultimate sort of end of the road be, I don't know,

like an army, a team of robots or something in the home that would, not only train you well, but could sort of actually, if there's a problem, they can interact immediately or deliver the medication you need or, I don't know, basically do stuff for you that's required. Is that kind of the ultimate end goal of this? Yeah, I think we believe that this agentification is real and we've built our platform

on a set of agents today that are working very well in various different care settings. We see these agents grow that can help augment just your care setting, whether it's at home or in the hospital. So instead of you having to search or understand

like what this data really means. If there's a co-pilot, an agent, you know, shoulder to shoulder with you as a clinician, as a nurse, as a patient, we have seen significant results already. So we see that agentification actually expand towards the outer years. And also like the EMRs, the electronic medical records that have all of this data,

can now signal a lot more data because we can take that signal from the EMR or from a device and action it through these agents in the right context. Okay, great. Let's get on that subject then. As AI takes on more clinical and operational decision-making, what ethical or regulatory frameworks do you think are most important to guide its use responsibly?

Yeah, there are some regulatory frameworks for AI already that CMS and other agencies have already embarked upon that journey. There's guidance that they've already started providing in terms of the thought process around these frameworks. We also see kind of AI committees within each one of the healthcare systems that are being formed.

In some health systems, they also have a chief AI officer. So I think the industry as a whole is getting more prepared to kind of see how best to adapt and adopt this technology into their four walls. There's been a lot of false positives, people with a camera that has some AI

or some devices that might have some kind of observational capabilities. But since there was so much noise around them, there's so much false positives that are being generated. Now we see it's more methodical, but more importantly, they want to have a single platform. It could be Android, maybe tomorrow it could be somebody else, but basically a platform play

based on regulatory compliance or governance that are being created by these health systems is going to be key. Yeah, I think potential users of this would probably be concerned about privacy because it sounds quite intrusive into your life. How is that kind of maintained or controlled? Yeah, I mean, it's a very fair question. I think

You as the patient have always have to be empowered because it's your data, right? At the end of the day, you and the physician are the only two people that know about your data, right? So again, when I said the guardrails, so when I basically, when we actually develop a large language model to say, I'm going to do ambient documentation or ambient observation,

We're just using that data for you and your physician. So it's not going outside. So our AI doesn't go outside to the web and try to have identifiable data be used for any kind of learning. So what we're doing is very specifically making sure that our data separation, segregation is all siloed out.

And the AI technology can be deployed on a very focused manner. It doesn't have to be deployed so that like, you know, I take your medical data and bounce it on the internet so that we can learn, you know, something that would help you. I think that those approaches obviously are not even being allowed in healthcare. So from a privacy standpoint of view, one,

At least in our platform, you as the patient have complete authority to say what you want the physician to see. I mean, physician, obviously, you want them to see everything. But, you know, even your loved ones or including, you know, other family members, you have complete authority to basically help curate and provide inclusion-exclusion criteria to the system so that you know you're fully empowered.

Okay, great. Excellent. So looking ahead then, five to ten years, what do you think would be the most transformative use case of AI in healthcare that we're only just beginning to explore now? Excuse me, that's a mouthful. Yeah, I mean it's a very interesting question because I think it won't be just one use case, okay.

I mean, you have AI being used in drug discovery, right? And from a life sciences perspective, you're looking at AI being used in precision medicine, like the example I gave you around prostate cancer, right? And then you have AI being used for operational efficiencies, like I talked about, right? Just pure because the administrative costs associated with care delivery are so high today.

So, and the real burnout, you know, in clinicians and nurses. I personally, you know, I'm very, very hopeful that the operational efficiencies and having these clinicians practice at their license level and reduce the burden on them, right, for nursing and for other collaborative tasks will have significant impact from a cost

resource and outcomes perspective. Because still at the end of the day, I mean, you still need physicians, you still need nurses. How do you make their jobs easier and how do you increase their capacity? I think that to me is the most interesting use case. And I'm very, very hopeful that we will change the way care is delivered by creating these efficiencies.

Absolutely. I mean, I can foresee a future where you'd have an AI doctor as good as a human doctor diagnosing problems and whatnot, but

At the end of the day, you still want that human connection. I don't think you can ever replicate that, replace that even with robots. So I think you're bang on the money there. Completely different subject now. So you founded several successful healthcare IT companies, as you mentioned. What common thread or principle has guided you through each one, especially in balancing innovation with real world impact? Yeah.

So throughout my journey over the last from 1998, you know, that's when I started my first company through 2025. You know, you can see it's like almost 27 years of innovation. My mission has always been to make access to care more easy for patients, not standing in lines, you know, just there's so much angst in health care.

you go to a doctor you fill out the same forms over and over again right so right from my first company you know where during the dot-com days we were helping thousands of hospitals bring an online presence my second company we were we actually introduced the first online bill pay in healthcare where you don't have to mail bills and like you know you can actually go pay online

So a lot of consumer and patient self-service that we brought to market. We extended that in our third company and the second company became part of National Cash Register, which does airline check-in to ATMs and so on and so forth. So my technology helped patients access care better, faster, better, cheaper, but also with better outcomes.

The third company, we were the first patient engagement platform that ran on a population health management engine. Again, the theme was, while you're checking in, if I say, Raj, you need to have your physical done this year. Oh, by the way, you have to pay $20, but here are the care gaps for you.

And so we made access to care even more easier, but also by telling you what your care gaps are, we're able to reduce those care gaps and have a better outcome for you as a patient. So the theme that you see in these companies is like, how do you make access to care better? And then by definition, if you make access to care better, then you gain a lot of throughput and efficiencies on the back end, whether it's a clinic or a hospital.

And then on the back end, the other side of the coin, we focused on that more at Andor. We said, like, listen, we really need to look at the problem of access to care. There are a lot of tools right now. We've pioneered some of those tools for patient access.

And this other side of the coin where we looked at the clinicians, the care teams in the hospitals and clinics, they just don't have tools to collaborate virtually more efficiently. And if you can actually inject some AI agents into that mix and bring clinical content and context into that collaboration, whether they're collaborating with the patient or among themselves,

we make it more efficient. So again, so there's two general themes in all the companies that we formed and we obviously strive to do better and better is the way we help patients access care and the way we help care teams collaborate virtually. Fantastic.

It's like an exponential growth from your first company to Andor Health. I wonder if this is the end of the road for you or what's the next step that's going to be even bigger again? I'm an engineer, so I keep building. So, I mean, this company is just one of the most, I mean, it's the best, like the

the most gratifying, I would say. And I think it's such so much value that this business is creating to patients and to caregivers. We really want to build something really, really global and big so that we can actually, you know, solve these problems of access to care and care collaboration. We're right now in three or four countries in the UK and Canada and going in the Middle East. And so we, so yeah,

We were all very focused, checked in, no pun intended, to actually make this something incredibly valuable to the industry. And after that, you know, I'm sure I have one more in me. Well, best of luck to you. That's all my questions, Raj. Is there anything else you'd like to mention you think is important? No, I really appreciate you taking the time to talk with me and speak with me. And so...

I'm looking forward to seeing the podcast. Fantastic. With that then, Raj, thank you for your time. That was very interesting. Also, don't forget to subscribe to IE Plus for premium insights and exclusive content.