Revolutionizing Medical Tech with AI | Hamed Shahbazi

[00:00:01] This is in your average business podcast and he's not your average host.

[00:00:06] This is the James Altucher Show.

[00:00:18] Okay, so we're talking about the aura ring.

[00:00:20] Spell that as an OUR A.

[00:00:23] Okay.

[00:00:23] And you're wearing one?

[00:00:25] Yeah, so basically based on its collecting things like heart rate variability, temperature,

[00:00:31] resting heart rate.

[00:00:32] And even my new son out of his perspiration around that it can total detect all these elements

[00:00:37] and so based on all these things it starts to infer.

[00:00:40] It has a bit of an algorithm where it determines whether or not you're more stressed or less stressed.

[00:00:45] And I guess when we are stressed, things happen to our heart rate, things happen

[00:00:48] to the temperature even of our skin, you know, it could sort of go up and down and then they've

[00:00:53] been able to basically push that into some kind of algorithm that then punches out some kind

[00:00:58] of stress response.

[00:01:00] And have you noticed that since you got the aura ring?

[00:01:03] Because now you measure your stress levels every day, have you become less stressed?

[00:01:07] Yeah, it's really interesting.

[00:01:09] I do think it's helped because it does.

[00:01:12] I go back and I look at my day and I'm like, okay, that's interesting.

[00:01:15] I didn't really notice that if I was stressed there but was I, you know, what was I doing?

[00:01:20] You know, and it does sort of train you a little bit, particularly with sleep.

[00:01:26] So when you're asleep, you know, you're not aware of all the things that are happening to you.

[00:01:30] You don't aware of all the things that you do before sleep that impact your sleep, right?

[00:01:33] And, and you know, if you had drinks, if you had a late meal, if you slept late,

[00:01:38] you know, all these things impact the call out of your sleep.

[00:01:40] And so it's really helped train me because I want better quality sleep because it's very important for health.

[00:01:46] Ultimately, you know, having that rest in restoration.

[00:01:48] So I would say, yeah, it's just sort of trains you.

[00:01:51] And having an active guide as to what happened during the day, the things that you did.

[00:01:57] Your behaviors during the day that contributed to your stress response

[00:02:00] and all these different vitals, you know, you know, it has this kind of cumulative effect that it kind of helps you act better over time.

[00:02:09] I've been afraid to get one of these devices that track everything because I'm afraid I'll just get obsessive

[00:02:15] about, you know, how many steps that I walk, how much hour, how much hour is the quality sleep that I get.

[00:02:21] And my, is my heart okay, am I too stressed?

[00:02:24] I'm afraid I'll just become like overly, the way people check their email all the time, I'll pray it.

[00:02:28] I'll just check this all the time and get obsessed with the data.

[00:02:32] But maybe it's useful.

[00:02:35] You know what, I think you're right, though.

[00:02:37] I do think about this sometimes.

[00:02:40] I used to check my email when I woke up.

[00:02:42] They now I check my sleep score like I am pretty obsessed about it.

[00:02:46] But what I've talked to people who have had it for a long time and they say, look, it's good to have

[00:02:51] for at least, you know, a couple of years.

[00:02:54] And then you've kind of gone through that training.

[00:02:56] I just spoke about earlier.

[00:02:58] And then once you've done that, you don't necessarily need it as much.

[00:03:00] So then you can sort of put it away if you think you've gotten a bit too obsessed over it

[00:03:04] and you're tracking things a bit too closely because really the benefit is in that training.

[00:03:08] Once you've had the training, you know, you know what happens when you eat late, when you drink,

[00:03:13] you know how your body works.

[00:03:14] So now you don't necessarily need all the, all the detailed analytics every day.

[00:03:19] All right, I'm going to get one.

[00:03:20] And the interesting thing is, and this is going to drive us to one of the topics I really want

[00:03:25] to talk to you about is the data or the company must be accumulating

[00:03:32] is probably amazing for in terms of training AI.

[00:03:37] So it's probably really figuring out the, the roots of stress and heart attacks and strokes

[00:03:42] and, and, you know, maybe other things like, oh,

[00:03:48] if among the million people who use our ring, the people who went down an income this year,

[00:03:54] only got six hours of sleep as opposed to people who went up and income,

[00:03:58] got nine hours of sleep on average.

[00:04:00] So there's going to be all sorts of conclusions and data for that the AI is going to determine from this.

[00:04:06] It's, it's unreal. You're absolutely right. Actually, we're just talking about this earlier,

[00:04:10] but, or has got so much data, I, I heard the CEO talk about this at a conference years ago.

[00:04:15] He was saying that they may be one of the best predictors of whether or not someone has COVID,

[00:04:20] just based on the signature of all these things that I just mentioned, heart rate,

[00:04:25] availability, temperature, you know, resting heart rate and your sleep patterns.

[00:04:30] And yeah, it, it, it tells a story, right? And, and so now what they do,

[00:04:36] is they also are able to tell you your, um, your cardiology age like that,

[00:04:40] your heart health age, how, how old is your heart compared to how old are you?

[00:04:46] That's interesting. The other thing that it does is it, it, it's now starting to tell you if

[00:04:51] it thinks you're going to get sick based on some of the AI response, um, in terms of,

[00:04:57] of being able to learn based on your data and other people's data, what, you know,

[00:05:02] and what happens to your vitals right before you get sick. So it has, because it has a million people

[00:05:08] who have gotten sick at different points. And if you start showing characteristics like,

[00:05:13] oh, you just went down in steps per day. Your sleep went down. Um, your heart shows you probably

[00:05:20] are on a plain traveling because maybe that, your heart changes there. So, so now you're going to

[00:05:24] get more likely to be exposed to some bad air. Who knows? Well, let me ask you, can you, can you

[00:05:30] buy the data? Can you license the data? I don't know. It's a really good, it's a really good

[00:05:36] question. I think you're right though. That's probably their biggest, uh, you know, kind of

[00:05:43] accumulated all this, you know, sort of first-party data, right? It's their device. They've collected

[00:05:49] all of those data and, and they're likely looking for patterns and, and no, anomaly detection

[00:05:54] amongst all of it. Um, so it's, it's, it's fascinating. I don't know if they'll ever open sourced it,

[00:06:00] but I know, I know this is an area, you know, this kind of activity tracking and biometric

[00:06:05] tracking this really, really on the rise. You know, that the whole idea of our lives medically

[00:06:10] is that, you know, we used to be checked out once in a while. Well, once in a while we'd get

[00:06:15] it checked up once in a while, we'd get our vitals checked and now we're getting into a world where

[00:06:19] we're going to be continuously monitored and, and that continuous monitoring, that data becomes

[00:06:25] really valuable and because, you know, people have said, hey, you know, my Apple Watch save my life

[00:06:30] because it was able to predict that I have a roof, yeah. And, and, and, and, and I was, you know,

[00:06:36] we're getting into that, into that stage of life now where these things are starting to get, you know,

[00:06:42] quite quite smart and quite continuous and sort of their their their evolution in terms of

[00:06:49] focusing on our data. Yeah, and, and, and using the AI or even very sophisticated statistics

[00:06:55] to kind of make, you know, and it's not like we're looking at this as well. Okay, it's collecting

[00:07:01] healthcare data so we can make healthcare predictions, but maybe we can predict other things like,

[00:07:06] people with these healthcare statistics both Democrat, people with these characteristics,

[00:07:11] yeah, probably. Or people with these characteristics are more likely to make

[00:07:16] a buy when they when they see an ad because they didn't get much sleep last night

[00:07:21] and here's an ad for clothes and somehow people would get to leave who have these other

[00:07:25] characteristics need clothes. Totally, I, I think those data really should, there's

[00:07:31] be really interesting. I, I saw an X the other day someone had built an ETF

[00:07:37] of stocks based on CEOs that lift weights like their their their avid, you know,

[00:07:43] avid lean to wellness, they lift weights, they they don't and they're really dedicated to it.

[00:07:48] And it was like striking how how much better than the S&P they'd performed as a group.

[00:07:54] Like that's interesting, right? I believe it, you know, about a year ago, I was,

[00:07:59] so I was competing in one of the various US senior championships for chess. And somebody pointed

[00:08:08] out to me that the winners of this are like three or four tournaments that kind of are the championship

[00:08:13] for for older people in the US for for chess. So one point out to me, all the winners that year

[00:08:19] you could tell lifted weights like they had, even though it's chess the game where you're just

[00:08:24] sitting down and you don't, you wouldn't think it's related lifting weights. It does give you more

[00:08:27] stamina. I give you more focus and there's something to building muscles that is that is healthy.

[00:08:34] Absolutely, you know, we are our minds and our bodies are connected. And if we have a stronger

[00:08:40] body, I think we I think we would have stronger minds. And you know, also what happens is when you

[00:08:47] lift weights, you have better glucose control. We now know that glucose has a lot of insidious

[00:08:54] effects not only in terms of our body, but also, you know, all sort of brain. You know,

[00:09:01] we're starting to think that, you know, dementia and neurodegenerative diseases are a form

[00:09:07] of Type 3 diabetes, right? You know, we started to hear that a little bit. So lifting weights is a great

[00:09:12] way of glucose control because that, you know, our muscles so cut more of that glucose. So yeah,

[00:09:19] you know, the more all this story is you got to lift especially as a man as you get older.

[00:09:24] You know, if you don't use it, you lose it. You start to get sarcopina then you've got other

[00:09:28] revetisks, you know, you fall down and, you know, like early death, you know, all this stuff, right?

[00:09:33] I got to start lifting weights. I have the weights are sitting right there. I'm looking at weights,

[00:09:37] but I just never touched that. So, Ahmed Chibazi, I really, I want to talk about

[00:09:46] the most important topic of our lifetime, which is the role of AI and healthcare and how healthcare

[00:09:51] is going to change not 15 years from now. But like in the next few months or year or two years,

[00:09:58] three years, like the whole world is going to be unrecognizable. And like people talk about AI

[00:10:03] with robots and self-driving. I don't care about that at all, but healthcare is going to 100

[00:10:07] percent change. Your the founder of well health, co-founder of heal well, which has been out

[00:10:15] of well health. These are both public companies, but heal well is there's a fascinating, fascinating

[00:10:20] thing with AI and healthcare, which we'll talk about. But first off, you start out as an entrepreneur,

[00:10:27] classic story from nothing, soldier first company. I think it was your first going for $340 million

[00:10:32] to paypal. What's the story? We're going to talk about healthcare and AI, but I just really

[00:10:35] know how to become an entrepreneur. Thanks for asking it and it's just delighted to be with you.

[00:10:42] Love your show and your broad podcast. It's great to be here. Look, I love solving problems as an

[00:10:49] entrepreneur. I was always curious about solving problems. I consider myself to be a tech journalist

[00:10:56] so the first place where I was applying technology to help improve processes and different businesses

[00:11:03] was in Fintech where we were trying to speed up payments, lower the latency of having the post

[00:11:09] to the back offices of different billers like Utility and Wireless Billers. This was important

[00:11:15] because a lot of people were paying a lot of money for last-minute bill payments. These emergency

[00:11:21] bill payments. We spread that up and we started to save consumers hundreds of millions of dollars

[00:11:27] because we were able to do that. I don't understand. What was it doing? I'm a consumer.

[00:11:34] You need to pay your utility bill or your car bill. If you don't pay it today, you're going

[00:11:39] to get shut off. You've blown through your due date and whatnot. You would typically go to

[00:11:48] a Western Union or money grant and they would do a money transfer for your bill payment. They

[00:11:53] would charge you like 8-10-12 bucks for to do that. We came along and we said, that's a courageous.

[00:12:00] That's crazy. We should be able to confidently post a payment to someone's back office

[00:12:07] for a lot less than that. We established links into the back office of lots of different

[00:12:15] billing companies. How would I make my, let's say, I'm past my grace period. You totally

[00:12:20] company has set me five notices. I'm about to get my electricity shut off. How do I hear about your

[00:12:26] company as opposed to just going to Western Union? I always do. And be what do I do?

[00:12:32] We actually became an option for a lot of the billers. You could go to the billers site directly.

[00:12:37] You could go to the convenience store. We actually had self-service automated kiosks spread around

[00:12:42] where you could actually put cash into those kiosks and you could then if you were under bank

[00:12:47] to let's say your cash preferred, there was a bunch of these different ways that we could acquire

[00:12:52] the transaction mainly outside of the bank channel. So that was the thing that we were focused on

[00:12:57] is the loader moderner didn't come demographic that was paying outside of the bank channel.

[00:13:36] It in retrospect, the name maybe is not so good. But I called it the poverty index.

[00:13:41] And I recommended because there was a recession. It was 2002. There was a bear market.

[00:13:46] It was recession was happening. And I recommended like dollar stores, rent-affurniture stores,

[00:13:54] pawn shops and payday lender stocks. And those stocks actually did do very well. And they were

[00:14:01] good businesses. Because like you say, the under bank, when you're banked, you don't understand

[00:14:06] that there are actually a large percentage of the population that don't even have bank accounts

[00:14:11] or bank accounts they can regularly use because they're expensive. There's all these fees.

[00:14:16] So services for the underbank do very well. And because they're local and community-based,

[00:14:24] it's surprising how few people default on like let's say rent-affurniture situations or

[00:14:32] payday loans and so on. Yeah, absolutely. And you know, it's expensive to be poor,

[00:14:38] unfortunately. And so it's sort of counter-intuitive. I should have access to cheaper services

[00:14:44] because I don't have as much money. But no, I actually have to pay more just to

[00:14:51] expedite a payment. And so this issue of accessibility and technology is a really big thing.

[00:14:58] And this is what we've tried to do at well health is to try to give that higher end experience

[00:15:06] to just normal primary care visitors, you know, allow them not to have to come in and submit

[00:15:14] for, you know, a couple hours for their, you know, doctor to show up. You know, to be able

[00:15:19] to respect their time through digital technologies, allow them the book online, allow them to,

[00:15:25] you know, feed information into the clinic to save them time, allow them to check in

[00:15:29] the underbobled phone. So we've developed all this digital patient engagement to help

[00:15:35] patients and clinics interact with each other. And then we've applied that to an enormous

[00:15:42] owned and operated network. So we're the largest owner operator of outpatient medical clinics in

[00:15:49] Canada, and we own a significant number in the US as well. So overall, wells of about $1 billion

[00:15:54] in revenue this year, I'm very profitable 130 million in the, but does likely what we'll get to

[00:16:01] generating, you know, pretty significant cash flow. And, you know, we're doing millions of

[00:16:05] patients and visits. And a lot, and our, our worldview is, is tech enabled healthcare, you know,

[00:16:11] it's incredibly important to us. We're not just out there, just trying to be an owner operator.

[00:16:16] We really want to be as tech enabled as possible. So we respect people's time and deliver a better

[00:16:21] experience. So, and I want to, I want to talk about well health, your clinics, and I am

[00:16:27] but also heal well, which is, you have, you also, as we were talking about with the ordering,

[00:16:31] you have millions and millions of patient records that you've been able to create an AI model on,

[00:16:38] and I want to talk about how this is going to, going to literally change the future. But I still

[00:16:44] am fascinated, how did you start your first company? Like, what was the transformation that took

[00:16:51] you from, you know, whatever, to, to entrepreneur, to successful entrepreneur. Thank you. Well,

[00:16:59] thank you. You know, I, I think it was just, I think as a entrepreneur, you have to be

[00:17:07] have a bit of audacity to think you can make a difference. And but what was happening in your life?

[00:17:12] Like, what were you doing? You know, it's interesting. I, it was really right out of school for me.

[00:17:18] I had, I had just finished my engineering degree and I did engineering

[00:17:23] because, you know, growing up in a, in a, in a, in a, in a, in a Persian family, I, I was sort of

[00:17:28] given the choices of being a doctor or being an engineer. And my dad was an engineer. So,

[00:17:36] once I finished engineering, I loved it was very interesting, but I just really realized that I don't

[00:17:40] want to be an engineer. I, I realized that I wanted to be a businessman. And so, it was around the time

[00:17:46] when, you know, we started to see really the rise of, you know, the graphical internet. And I just

[00:17:54] knew that I really wanted to be the part of that, of that growth. I wanted it to be involved in,

[00:18:01] and, you know, this is before iPhones, you know, I'm not, I'm not in my 20s or 30s anymore. Right?

[00:18:07] So, and, and I could just see that the internet was going to disrupt everything. And so,

[00:18:12] you know, it became a really, a really, a really big priority for me to get involved in the

[00:18:18] technology revolution and apply, you know, digitization and modernization theme to different businesses.

[00:18:25] And, and that, and that's really what I've done on my life now. I've applied it to Fintek. I'm now

[00:18:29] planning it to healthcare. It was remarkable that I even had the chance to do that with healthcare.

[00:18:34] It just goes to show you how, how much lag there is in adoption of digital technology.

[00:18:40] It was healthcare. Like healthcare, I feel it was 30 years behind the times.

[00:18:44] 100% and so, you know, in 2017, 2018, you know, around the time that that I'd sold my previous

[00:18:52] business. I remember having these experiences in doctors offices and I was thinking,

[00:18:56] this is unreal. Like I could have had the same experience 10 to 20, 30 years ago. The doctor

[00:19:01] came in, it was handed his, you know, paper folder and he wrote paper notes. And, and I got a

[00:19:10] kind of referral that I had to then use to call someone else with the phone. And it's just like

[00:19:17] everything about it was was not digital. It took forever for me to get in. There were people running

[00:19:21] around everywhere in the clinic. It didn't look very organized and I was like, you know, I wonder

[00:19:26] if this is like an anomalous situation or is this the state of play in healthcare? And so,

[00:19:32] I started to do some research and I was like, I was just floored. And at that time, in Canada,

[00:19:37] we had 0.25% penetration and telehealth throughout the country, which is unreal. It likes

[00:19:42] really, really low at a time when how do you define telehealth? Telehealth is a patient

[00:19:49] physician consultation that occurs in any other format rather than in person consultation.

[00:19:57] So it could be over the phone, it could be an audio visual kind of link like this. And,

[00:20:05] 1617 we got iPhones everywhere all over the world. And so, and so the idea that there wouldn't be

[00:20:10] you know, video calls or commonplace, but they had not really made them way into healthcare.

[00:20:17] Some of some of the some of the different HMOs that we're doing this like Kaiser,

[00:20:22] permanent in the US had had already quite a bit of growth in telehealth and whatnot. But as

[00:20:27] a country here in Canada, we were quite you know, low and just just for contrast purposes,

[00:20:33] two years later during the pandemic because of physical distancing, we got to a high of 80

[00:20:39] percent penetration of telehealth. So 80% of all visits were attributable to telehealth and 20

[00:20:45] percent will not. And now it's now post-pandemic, we're back down, but we're not back down to

[00:20:50] the same levels we were before. Pretty much an all industrialized countries were in that 40 to 50

[00:20:55] percent range. So, forever this industry's now changed as a result of the pandemic. You can't

[00:21:00] say that about too many industries where forever there changed. Now there's an entrenched kind

[00:21:06] of kind of behavior and preference to see physicians online for certain types of appointments.

[00:21:12] That could be mental health. It could be chronic disease checkings. It could be for prescription

[00:21:17] renewals. All that kind of stuff. Yeah, it seems like well first off, it seems like the general

[00:21:23] trend for humanity is we don't ever want to leave our house. Like, you know, right now I can get a

[00:21:30] chef to cook me food and somebody else will then pick up that food, deliver it to me and then I

[00:21:37] could watch a newly released movie streaming right into my house like I just never have to go anywhere.

[00:21:41] And I can have a telehealth position to, you know, give me a prescription for something and I can go out

[00:21:50] and get anti-depressants and testosterone and whatever just deliver to me. All deliver perfectly.

[00:21:57] Yep. And if you want to actually give you the shot, there are actually people now. There's all

[00:22:03] kinds of businesses across the country that will come over and they'll give you that B12 shot

[00:22:06] or they'll give you a little minister of the shot for you if you don't want to do it yourself.

[00:22:10] Yeah, I just applied for life insurance with someone came here and took my blood and

[00:22:16] did all the blood test and then and then also I used to use the, I haven't used this lately but

[00:22:22] the IV doctor where they come over and give you like vitamin C and whatever, you know,

[00:22:27] I would do that before traveling. But also with telehealth now you have more opportunities

[00:22:33] to collect digital records, right? So like what sorts of, what's more to data? Did you,

[00:22:38] once you were doing having millions of patients do telehealth with your clinics,

[00:22:41] what sort of data did you start to collect? Yeah, so I'm really glad you brought that up

[00:22:48] because you know, thematically speaking electronic medical records have been around for a while

[00:22:54] and while there has been less usage in the past, that's growing tremendously. And for a long time,

[00:23:01] doctors have been entering in all this data, entering in their observations of the patients,

[00:23:08] you know, what they're dealing with, what the assessment of the visit is, what the care plan is,

[00:23:12] you know, what drugs they're taking, whether or not they're working, all this kind of stuff. All

[00:23:16] this information is going into the patient record and you know, if you're seeing a specialist

[00:23:21] or you're getting labs done on your blood, all that stuff is coming into the primary care,

[00:23:26] you know, record and the problem is all this information has been going in and very little value

[00:23:32] has been coming out. And this is where our partnership with Healwell

[00:23:36] our spin out of our data into Healwell has been as being tremendous because we now have, you know,

[00:23:43] physician copylates like AI copylates. What is AI really good at? Taking a vast amount of

[00:23:49] information, very complex, very complicated and being able to actually create very simple insights

[00:23:55] and be able to deliver those to people at the right time in generate those insights based on

[00:24:01] that kind of anomaly detection and learning of from data. And once with example,

[00:24:05] there may be a certain pattern, a certain signature to a certain disease. Like let's say,

[00:24:11] how your blood tests come back, you know, the types of vitals that that you exhibit,

[00:24:16] you know, in terms of your enzyme levels or or you know, whether or not you're prone to diabetes

[00:24:21] or your BMI, you know, all this kind of stuff could inform the view that maybe you have, you know,

[00:24:27] some kind of, you know, diabetes or chronic kidney disease or anything like that. And maybe

[00:24:33] the doctor hasn't picked that up because they have not been able to review all of the information.

[00:24:38] There's so much information in the patient record, it's really hard for the doctor to go in

[00:24:44] and connect the dots and integrate the information on the fly. Especially when you consider that

[00:24:48] here in Canada, doctors are getting paid 30, 40 bucks for primary care visit in the US. It's not

[00:24:53] that much more, right? So it's a lot of work to do for a little bit of money. And so, so basically

[00:25:01] this software will, you know, what we call risk stratify the patient for particular diseases.

[00:25:07] It'll say, hey, these are the patients that are, you know, red, you know, yellow green,

[00:25:14] they're, they're, they're risk for, you know, heart disease, they're risk for diabetes,

[00:25:19] they're risk for chronic kidney disease or, you know, like their, they're, they're red, meaning

[00:25:24] like, I'm sorry I keep it drop thing. I know I'm going to get complaints later from

[00:25:28] listeners that, oh, you dropped too much. But I just want to, I get curious about, is the problem

[00:25:33] that like, I feel like when I get a blood test and I, I barely in my life, I haven't been to

[00:25:39] a doctor's visit since I was like 18 years old. So, and I get a lot of criticism for this

[00:25:44] from my wife in particular. But it's the problem because when you get a blood test, okay, here's

[00:25:49] my liver counts, here's my enzyme counts, here's this thing, that thing. And the doctor just

[00:25:55] looked at one thing at a time and say, oh, these high liver counts mean liver disease is the

[00:26:00] problem that they're just looking at one, you know, one thing means this, one, another thing

[00:26:05] means this. And they're not looking at all of the, how the different data piece of data

[00:26:10] interact with each other. And maybe it's slightly elevated count in this area is related to a

[00:26:17] slightly, I kind of, this here, could mean some cancer or some other horrific disease that you

[00:26:22] have. But most doctors don't know all these correlated relationships, but AI does.

[00:26:28] You're absolutely right on exactly. So, like what's an example of that though? And I'm sorry,

[00:26:33] what I just trying to understand. Yeah, I mean, I think it's a lot of those, you know,

[00:26:39] I think part of it is observational things that the doctor will write over time and so there's

[00:26:46] a longitudinal element here, there's a timeline element here. So, so the AI will review

[00:26:51] all the different, you know, clinical notes. And this is what AI's really good at. It's actually

[00:26:57] taking unstructured information and making it structured. So, imagine all these various notes

[00:27:02] that are just sitting in the record over time. And it's, it, it, it'd be hard to go in and read

[00:27:09] all of them. They could be, you know, thousands of pages of PDF, but the machine can do that fairly

[00:27:14] quickly, right? And so, it can, it can, it can look at, you know, that those doctors observations

[00:27:19] over time, you know, what your BMI was, what these results were like. And, and it can piece together,

[00:27:25] you know, stories, like, oh, look, it's not just what happened at this blood test. It's what

[00:27:29] happened to all those blood tests that occurred over the last, you know, several years and all

[00:27:35] the different, you know, observations the physician had, you know, based on this,

[00:27:40] it could mean that you have a rare disease. And rare diseases are interesting, because there are

[00:27:44] literally thousands of rare diseases and ultra rare diseases. This is a $200 billion market for pharmaceutical

[00:27:53] companies. A lot of those rare diseases have pharmaceutical interventions. And some, like in

[00:27:59] Canada here, there's some rare diseases where they'll have three or four customers, you know,

[00:28:03] you know, patients that that that become customers for particular pharmaceutical interventions. But it's

[00:28:08] really hard to actually identify what whether or not someone is, you know, it is subject or

[00:28:14] has exposure to some kind of rare disease. So what's a story slash disease where prior to AI,

[00:28:21] nobody would have or would have been very difficult to figure this out. After AI,

[00:28:26] this is a specific example. Like you have a story of a patient where they had symptoms over years,

[00:28:33] but weren't getting diagnosed correctly and then suddenly AI and boom, they're their diagnosed.

[00:28:38] Yeah, I mean, we have tons of those. I mean, I, I, I probably couldn't, couldn't

[00:28:43] riddle off, you know, specific ones. But, but you know, I've heard tons of these types of stories.

[00:28:48] And you know, we all, we all know someone who hasn't been feeling well for a while and no one can

[00:28:54] quite, you know, put the mark on what's going on with them. And you know, they're like, I've had so many

[00:29:00] tests and, and, and, and, and I've talked to so many people and no one's been able to figure it

[00:29:05] out and have gone from person to person to person. And it's because a lot of the times,

[00:29:09] it is some form of rare or ultra rare disease. And, and you know, we now, you know, support hundreds

[00:29:17] of them through the Healwell technology. There are hundreds of these diseases that, that are copilot,

[00:29:23] you know, supports and can identify and, and, and, and, and that's extremely valuable for

[00:29:30] a physician. But the most common diseases are chronic care diseases. So, so we've, we've, we've

[00:29:36] seen this a lot. We've seen, we've seen people that are that are degrading and, you know,

[00:29:42] good things are not happening. Uh, even if they're identified for a particular disease,

[00:29:47] like let's call it chronic kidney disease. If they're, if they're under, uh,

[00:29:52] dosed in terms of the, the, the pharmaceutical interventions that they're taking,

[00:29:57] we've had a particular case I was looking at the other day. This particular patient was

[00:30:02] not doing well. And, and the copilot identified that they're not taking the right dosage.

[00:30:08] So it's just something simple as that changed this, this patient's life forever because now

[00:30:13] they're taking the right dosage and, and they started to improve right away. What, what did the

[00:30:18] doctor say? Oh, if you're not feeling well, but we think we got the diagnosis right, it's got to be

[00:30:23] the dosage. What, why was, why did the AI have an advantage over the doctor? Because it was able

[00:30:29] to kind of look at the data and, and recommend a particular dosage and, and for that,

[00:30:34] for the particular progression that that disease had with that person, the dosage must

[00:30:39] should have been much much higher. And you know, these pharmaceutical interventions are sometimes

[00:30:42] just useless if you don't have the right dosage. Like you might as well not even be taking it.

[00:30:47] Right? And so dosage and adherence are incredibly important in terms of, you know, beating back

[00:30:52] whatever, whatever ailment that you have. So okay, how did you train the AI? So you have millions

[00:31:16] and millions of patient records starting from, let's hit the initial phone call. I've got a stomach ache.

[00:31:20] I've got a headache. Then you have a doctor's visit and a doctor takes all the notes. Then you have

[00:31:25] the next visit, the next visit, the next visit. You got all these millions of records. Just tell me

[00:31:30] technically, how did that, how did you train the AI? Like what did you do? It's, it's not easy.

[00:31:36] So because of the complexity of the human body, you have to go disease by disease and you have to go

[00:31:42] you have to do a really deep dive. You have to kind of bring on specialists and experts in that

[00:31:47] particular field. You need to look at tons and tons of data. You need to identify essentially

[00:31:53] what the signature of that disease is. Like it, you know, like it, it is characterized by

[00:31:59] you know, certain types of labs, certain types of progression in terms of the degradation of

[00:32:04] certain, you know, vitals that could be showing up. You know, so some of that could be rules-based,

[00:32:10] like it doesn't have to be AI. But what's interesting about AI is just the ability to take

[00:32:16] unstructured information and structure it. That's really hard to do because, because most of

[00:32:21] the information in the patient record isn't just perfectly laid out in tables. Now, this is what we

[00:32:26] call structured information. You can run a search easily, comes back. But the vast majority

[00:32:31] information isn't there's unstructured. So someone in AI has to read that whatever was written

[00:32:37] there and try to interpret what it actually means. What was intended by the physician,

[00:32:43] the wrote is in context with everything else. That's where it gets really exciting because

[00:32:48] it then structures it and then it can do a anomaly detection. It can say, oh, based on all this

[00:32:52] information now we know that we are actually pretty close to this particular disease. And this

[00:32:59] is what rules-based doesn't do because rules-based requires you to perfectly hit on those criteria.

[00:33:05] If there's 10 variables, there's like 10 to the 10th number of possible variations of those

[00:33:12] variables and conclusions. So obviously this AI was trained on data. So you have millions of records

[00:33:20] where you have years of patient record, millions of patients where you have years of records. And

[00:33:25] at the end of the day, there is a diagnosis that doctors agree on. You have this training data.

[00:33:34] How did you then feed it into an AI? How did you build the AI?

[00:33:38] It's really- so what the AI is doing here is basically processing all of this data and learning.

[00:33:48] So a lot of times you're basically writing algorithms that allow the AI to just review the data

[00:34:00] process, it learn from it and identify the anomalies. And as you find the anomalies,

[00:34:06] you start to build software around these anomalies mean certain things. And a lot of times,

[00:34:14] machine learning is essentially trying to find out something based on this data that you don't

[00:34:19] know about. Rules-based filter, you know, you're looking for something you say if this and this

[00:34:25] are true, tell me what the output is. And machine learning, you don't know it, you say,

[00:34:29] tell me what the needle in the haystack is. Find me the needle in the haystack. That's what

[00:34:34] machine learning is good at. And when you do that enough times, and this is why you need so much

[00:34:40] if you do that enough times, you start to correlate relationships. You start to say, oh, we found

[00:34:45] the needle in the haystack this many times in this data. It means that we have now identified a real

[00:34:52] pattern and it's that pattern detection that AI is so good at. Does that make any sense?

[00:34:58] Yeah, and so- so that's where the- the kind of supervised learning first takes place, right? So you

[00:35:04] had all these patients with all these diagnoses, they're in a big system. And then this system

[00:35:10] sort of figures out all the possible contexts that are related to each disease. And now given

[00:35:16] a new patient record, it then basically spits back, oh, and you don't even know which patients

[00:35:23] it's looked at, which patients is correlated to it just, but it figures it out. And it spits back,

[00:35:29] okay, this person might be at risk for this rare disease. Right. Yeah. In the process that I was talking

[00:35:34] about and you touched on to is about training the data. So once you train the AI, so the AI

[00:35:40] starts to understand certain things then it becomes a lot more useful and being able to respond

[00:35:46] to queries and make that data a lot more useful. And then, you know, there's basically two forms of

[00:35:52] AI, there's symbolic AI and then there's generative AI. You know what's interesting about generative AI

[00:35:58] is that it's, it's literally predicting what the answer is going to be based on this enormous

[00:36:06] amount of data. And so one of the things that we've done with one of our co-pilets is actually

[00:36:11] having the co-pilot, the service actually listened to a patient and provider conversation. And

[00:36:19] taking all of that conversation and generating a medically relevant note, this is being a huge

[00:36:25] change of your physician's because typically physicians will have to take a transcript of the

[00:36:30] entire meeting, entire consultation and they'll have to actually make that note. That note has

[00:36:36] regulatory relevance. Typically that's how you get paid as a physician. You need to make that

[00:36:41] note into the record and that note has a particular structure. It can't be a super long note. It has

[00:36:45] to be fairly short. And it sort of talks about this situation, what the assessment is, what the

[00:36:55] you know, a good part of their day just generating this note. Right? And this is why they're taking

[00:36:59] all these notes and transcripts while you're speaking. But we found that that's returning

[00:37:05] 20 to 30 percent of physicians' day back to them just by just by generating this note. So

[00:37:15] that's pretty incredible. So we're already starting to see back to your comment earlier

[00:37:19] that physicians are starting to change. Not two, five, 10 years from now. Now, I'm starting to impact

[00:37:26] them. They're starting to get real time at the AI make a diagnosis while the patient and the doctor are

[00:37:30] talking. It can. It can start to take that information that's just taken into the gender

[00:37:37] of AI as it's generating. And it can start to identify things about it. And especially it's

[00:37:44] powerful if it's connected to the rest of the record. So then it could say,

[00:37:49] you know, based on this visit and based on the patient record, we recommend a referral or we

[00:37:55] believe that there's, you know, the assessment of the chart now is that this person should go

[00:38:02] see a specialist or this person should try this pharmaceutical intervention. So yeah, I mean,

[00:38:08] all of it's connected. And that's where we're going. Basically we're going in a direction where

[00:38:13] physicians are going to get a lot more help. I think they've been very expensive data entry

[00:38:18] clerks for many years entering data, you know, into these patient records and to the EHRs

[00:38:25] and not getting much back. And now we're starting to see AI driven co-pilets actually be able to,

[00:38:32] you know, you know, pull all this patterns together and identify these these, these,

[00:38:42] move, move things forward. It's like think about a plane and and and a pilot, you know,

[00:38:47] you get into a plane and you fly and you don't even think anything about it but the,

[00:38:52] the onboard computers doing everything, but you still probably would not fly less that

[00:38:57] that pilot got into the plane, right? We feel we feel much better that there's a pilot there.

[00:39:04] I think this is where healthcare is going. You're going to have these copilets all this machinery

[00:39:10] and intelligence that helps the physician make a good decision. But for the most part,

[00:39:18] a lot of the work, the hard work's going to be done by machines in the future. And that sounds

[00:39:24] a little, um, a little much but when you consider the fact that a human body has far more complexity

[00:39:30] than at Boeing 747, there's no way a pilot could run those those planes properly with it out on

[00:39:37] computers. So how are they developing patient visits, right? Like let's say let's say I'm a patient

[00:39:43] and let's say some days I have stomach aches, some days I have migraines, some days I'm a little

[00:39:50] bit more for teagued than others and I go to a doctor, they might say something like well when you

[00:39:55] have a stomach ache take this pepto visible and when you have a headache take et cetera and

[00:40:00] when you're fatigue take melatonin and but but you're you're saying I could be I could go in

[00:40:08] and given my prior records and given the conversation I'm having with a doctor, the air I can say

[00:40:13] no you have this kidney problem, this rare kidney problem and you need to see a specialist right now.

[00:40:20] So that's sort of a leap that that could happen where the doctor is just not going to make the

[00:40:30] rare kidney disease, the doctor might not make that connection because he's never encountered

[00:40:34] before but the AI among its millions of records has encountered it many times. Correct, I mean a lot

[00:40:41] of what you see when you go to a walk in clinic or some kind of urgent care center is basically

[00:40:47] a physician treating your symptoms like okay I'm sorry you're feeling that way here but you're

[00:40:51] things that you could do but there may be something going on and from a preventative health

[00:40:56] solve our preventative health you know like if we can if we can identify the issue earlier time

[00:41:01] in medicine is everything right so if you're not feeling well if you've got to pay and somewhere

[00:41:05] listen to it don't ever so depressed that because pain is a signal you know these are all these

[00:41:12] things that happen in your body are signals and you need to pay attention you need to be aware right

[00:41:16] and so you know and what we're doing with the eyes we're just we're just getting time

[00:41:22] and that's allowing us that's giving us really precious opportunity to intervene before things get

[00:41:28] works. Do you have a sense of what percentage of the time the AI makes a recommendation

[00:41:33] that's correct versus incorrect? Oh it's a good question the vast majority of the time it's

[00:41:40] got to it's got to point now sometimes the doctor could know so it could sort of you know

[00:41:44] dismiss the notification or would have you sometimes the AI depend due to a lack of information

[00:41:50] it may not make the right recommendations but that's also in the quality of the AI

[00:41:55] so if it doesn't have enough information it shouldn't make a call so we've sort of seen a

[00:41:59] little bit of that is that you know it doesn't always have to give it an output it should

[00:42:03] give an output when it feels you know quite certain based on you know all the things that it

[00:42:08] needs to see in terms of pattern detection so that does get to call it a bit of it but but

[00:42:13] but I would say that with the type of copylates that we're building and that we see you know

[00:42:19] there's there's a lot of you know rigor and discipline and quality that goes into the stuff

[00:42:26] I would say you know it's sort of rarely wrong you know it's probably going to be right you know

[00:42:32] 95 plus percent of the time or it have some kind of point to make so you'd have to review those

[00:42:38] notifications as a doctor and and try to understand why it's making those inferences.

[00:42:43] Are we getting to a point where we can even do or you could implement health care as a

[00:42:50] service so I could log on to heal well and say oh I have some of the eggs today and in tomorrow

[00:42:56] I could log in and say oh I had a migraine today and I took it's said right and then the next day

[00:43:02] oh I slept 14 hours for some reason and then and then it comes back and says look you need to see

[00:43:08] this doctor he's available right now I'm going to make the call for you and boom like so the whole

[00:43:14] so health so has HAA health care as a service. Absolutely yeah I know I think that in the future

[00:43:22] you're going to have a lot of intelligent questionnaires A.I. that asks you a bunch of questions and

[00:43:27] basically going through a chat and then it'll get to a point where it says you know I'm going to

[00:43:38] do about the situation and the beautiful thing about that is that as soon as you speak to that

[00:43:42] person it's already delivered kind of this assessment and all the notes required for that person

[00:43:46] to understand exactly you know the situation and where they can start to make a valuable contribution.

[00:43:53] I because I feel like now there are simple versions of that but it's very rule based like

[00:43:59] let's say it's specifically about sleep like how many hours on average do you sleep a night

[00:44:03] okay you probably need this drug boom so it's like we're very well based.

[00:44:10] If it's based on like the AI of you know understanding the data of millions of patients

[00:44:15] paired with their diagnoses and then comparing my specific language and context to the AI database

[00:44:23] seems like that's very powerful. Yeah I mean look I think I think the more the more history you have

[00:44:30] the better the AI is going to be and what's interesting it's not just about your data see

[00:44:36] a capable AI is doing this deep learning where it's comparing your data with all the other data that

[00:44:41] it seems so art of this is about unlocking the value of your data but in context with with

[00:44:48] these greater data sets. Well how many patients records do you have in the AI? Oh I mean we have

[00:44:55] you know we're over 20 million patient records you know just just in our Canadian you know

[00:45:01] in our Canadian business so it's it's it's a pretty vast amount of patient records. It's

[00:45:06] terabytes of data. So it's like a doctor like a really smart doctor with perfect memory on 20

[00:45:11] million patients. Yeah exactly I mean it's hard to beat right? Jay maybe we this could be

[00:45:17] Jay do you think we have like a good part one and then we will do a part two where I would ask

[00:45:22] more about the specifics of healthcare AI and your companies and also more about entrepreneurship

[00:45:29] but I think this idea of healthcare as a service will get people thinking and uh and I'm very

[00:45:37] excited about the future and what it brings so thanks so much Ahmed and we'll talk maybe not next

[00:45:42] week but the week after we'll end Jay will help schedule something.