Latest / Becker’s Healthcare Podcast / Building Safe, Empathetic AI and an Abundance Mindset in Healthcare

Transcript

Building Safe, Empathetic AI and an Abundance Mindset in Healthcare · Becker’s Healthcare Podcast

0:00
Hi, everyone. This is Molly Gamble with Becker's
0:01
Healthcare. Thank you so much for tuning into
0:03
the Becker's Healthcare podcast series. Today, we're going
0:06
to talk about building safe, empathetic
0:08
AI in healthcare
0:10
and what it takes to design technology that
0:12
providers, payers, and patients can truly trust. Joining
0:15
me for today's discussion is Munjal Shah, cofounder
0:18
and CEO of Hippocratic AI. Munjal, thank you
0:21
so much for being here with me today
0:23
and joining me. Oh, thank you for having
0:25
me. I'm, I'm really excited.
0:27
Well, to get started, let's learn a bit
0:29
more about you, Munjal. You've had an interesting
0:31
journey as a serial entrepreneur
0:34
building companies, experiencing
0:36
big wins, also some big setbacks,
0:39
and ultimately landing in health care AI.
0:42
Can you get us up to speed and
0:44
walk us through that arc, and and what
0:45
drove you to found this company?
0:48
Yeah. You know, I think, I grew up
0:50
here in Silicon Valley.
0:51
And, when you grow up in Silicon Valley,
0:54
it's a it's a company town.
0:56
And so you grew up dreaming to start
0:59
a company one day in technology. I remember
1:01
watching Steve Jobs and Steve Wozniak and saying,
1:04
oh, I wanna do that one day.
1:06
And so, you know, as kind of some
1:09
of my earliest memories, I knew this was
1:11
gonna be my path. And,
1:13
started my,
1:14
did my graduate work in AI and then,
1:17
said, hey. I I've I'm gonna start kind
1:19
of my first enterprise, and,
1:22
that turned out and and worked out well.
1:25
And we sold that to Alibaba. And then
1:26
I I started my second one, and we
1:28
sold that to Google.
1:29
And,
1:31
in what should have been the best day
1:32
of my life,
1:34
in terms of at least business career,
1:36
I ended up with chest pains and ended
1:38
up in the ER.
1:40
And I was 37 at the time.
1:42
My dad had, had his first heart attack
1:45
in his mid forties.
1:47
And then I kind of realized, oh, man.
1:50
This health thing doesn't come for free. You
1:53
really gotta take care of yourself.
1:55
I'm five six, and so I was about
1:57
40
1:58
than I am today.
1:59
And so I took control of my health
2:01
and, you know, at that height, forty pounds
2:03
a lot. But,
2:05
and decided to really change
2:07
change and and rehabilitate myself. And in doing
2:10
that said, you know, if I'm gonna build
2:11
another company and I'm gonna work as hard
2:13
as it takes to build one, I'm gonna
2:15
build a health care company,
2:17
because I wanna help others as as the
2:19
next chapter of my,
2:21
of my personal mission. And
2:23
I always joke a little bit. Everybody's always
2:25
like, oh, find your mission in life. I'm
2:26
like, yeah. My mission kinda found me on
2:29
that run.
2:30
You know? It's like sometimes in life, you
2:32
don't you don't plan them. They just kinda
2:34
happen a certain way. And and
2:36
then I,
2:38
and so I built a health care company,
2:40
except I didn't quite understand
2:42
when I've done
2:44
tech is a whole lot easier than health
2:45
care. Health care is so much harder, and,
2:47
you know, we built this company and we
2:49
grew it substantially, but it ultimately was not
2:52
nearly as successful as the first two. And
2:56
but just as that one was ending, ChattCPT
2:59
had come out.
3:01
And as an AI person
3:03
who had seen
3:04
all of the previous generations of of technologist
3:07
area, I kinda woke up and I was
3:09
like, oh my god.
3:11
This is the AI we've all been waiting
3:13
for our whole life. Like, this thing finally
3:15
works. I mean, let's be honest. Like, before
3:18
that, it was very brittle. If you use
3:19
the AI in just this perfect way, it
3:21
worked great. But if you didn't use it
3:22
in this way, it didn't really work. And
3:25
I think I just realized that,
3:28
the time has come. And even though I
3:30
had done that last one for ten years
3:32
and honestly was quite exhausted and honestly wanted
3:35
to take a break,
3:36
I realized the window of opportunity in the
3:38
world had just opened. And if I wanted
3:40
to run, I wanted to run with it.
3:42
And
3:43
and, honestly, I I also had a little
3:45
bit of just, you know, I I didn't
3:48
love how the last one ended, and I
3:49
said,
3:51
I'm I'm not gonna go out that way.
3:53
I'm gonna do this one more time and
3:55
try to have the impact on health care
3:57
that I didn't successfully get,
4:00
because I I just was committed to that
4:01
mission. And,
4:03
and so we started Hippocratic,
4:06
this is, like, February, March '23.
4:09
So remember, CHAT GPT came out Yeah. '22.
4:13
Right? So it'd only been a few months,
4:16
and and just got off to the races.
4:19
And and for you know, my life has
4:20
been this entrepreneurial journey.
4:22
You know, life of an entrepreneur is always
4:24
a bit,
4:25
I think more topsy-turvy
4:27
than the average person, but it's
4:29
you get to bring new things in the
4:31
world that the world's never seen before, and
4:33
I think that's what truly excited me.
4:35
You just your story really helps highlight how
4:38
health care is it's so wildly personal. It's
4:41
also massive. It's imperfect. It's extremely
4:44
challenging,
4:45
and and rigorous.
4:47
And I I you are a builder clearly,
4:49
Minjal, but I'm also curious about how you
4:52
as a builder and your mindset and your
4:54
values have had to shift in health care.
4:58
What values are you holding closest now as
5:00
a leader for yourself, but also for Hippocratic
5:03
AI as a company?
5:05
When we started the company,
5:07
we knew the number one value
5:09
had to be how do you bring AI
5:12
safely.
5:13
And remember, you know, again, ChatGPT had first
5:16
come out, and what was the number one
5:17
thing we all saw first?
5:19
We were like, oh, this thing hallucinates.
5:22
Like, it was a oh, what amazing power.
5:24
But every now and then, it it just
5:27
says something that we're like, oh my god.
5:28
I can't believe that. That's just not true.
5:31
And so we realized that that we had
5:34
a power in our hands, but we had
5:36
to bring it in a safe way.
5:38
And I think this was
5:41
and and
5:42
we decided, like, let's name the company in
5:44
a way that we can never get away
5:46
from that value.
5:48
You know? People ask me that every now
5:50
and then. They're like, are you really committed
5:51
to being safe? I'm like, I named the
5:53
whole company
5:55
after the Hippocratic oath. I made the tagline
5:57
do no harm. Like, I don't know how
6:00
more to say this is what I care
6:02
about. And Right.
6:04
And so I I think that, you know,
6:06
this became our our core focus for the
6:08
company.
6:09
And
6:10
and we did and it kind of everything
6:12
came around that. So first of all,
6:15
the most important thing was what we said
6:17
we wouldn't do.
6:18
Because everybody says they'll be safe, but sometimes
6:21
the right safety is to limit one's ambition.
6:24
And so we said, we're not gonna try
6:27
to diagnose. We're not gonna try to prescribe.
6:30
We're gonna operate within the scope of, you
6:33
know, an RN down.
6:35
Registered nurses don't diagnose. They don't prescribe.
6:38
I'm not saying this isn't an RN, but
6:40
I'm saying that's the scope of practice of
6:42
the agent.
6:44
And and even then, only the stuff can
6:46
be done virtually. But I think that
6:48
we've come to realize that there was such
6:50
a
6:51
power to be brought, but you needed to
6:54
say no. We avoided
6:56
kids under two. I mean, we made a
6:58
lot of decisions to say no.
7:01
And
7:02
and then as we built the product, you
7:04
know, we can talk about a little later
7:06
on the safety side, we added even more
7:08
things. But a lot of it was the
7:09
ethos of the company
7:11
was this idea of if you're going to
7:14
do this, you gotta do this safely
7:16
Mhmm.
7:17
And have a core belief. Our our number
7:19
one value at the company is
7:21
patients first.
7:23
And it seems obvious, but it's also saying
7:27
patients are first over everything
7:29
which is a very you know, that's an
7:31
extreme,
7:32
in some people's view kind of thing.
7:35
And so I I but this was
7:37
like, I still have the story of my
7:39
mom that happened recently where,
7:42
you know, and we'll we'll talk more about
7:44
why this is important, but she
7:46
she had high blood pressure. Like every late
7:48
seventies woman, she, you know, lied to her
7:51
children and said, doctor said I was perfect.
7:54
I have no health issues.
7:56
And we didn't know that she had it.
7:58
She did refuse to take her medications,
8:01
and five years later, we end up in
8:03
the hospital one night with two twenty blood
8:05
pressure and the beginnings of CHF.
8:08
And and I just
8:10
I'm obviously mad at myself,
8:12
although due to HIPAA, I couldn't really get
8:14
the data, but fine.
8:16
But I'm also just mad at the system
8:17
for her. And I think that, you know,
8:20
nobody called her and said, hey, missus Shaw.
8:21
You haven't asked for a refill in four
8:23
years.
8:24
Are you really taking this medication?
8:26
And,
8:27
I mean, there's obvious clues she wasn't
8:30
that they had.
8:32
And
8:33
I think these are the values I bring,
8:35
which is, like, how do we help
8:37
more and more people? You know? And for
8:39
me, they're always, you know, the family members
8:42
around us. And and that's one of the
8:43
biggest values of the company is not only
8:45
patients first, but, like, who is the person
8:47
in your life that that you really are
8:49
doing this company for? And and
8:52
and so we emphasize that.
8:54
I really appreciate you expanding on the the
8:56
restraint
8:57
and those quiet decisions of what you declined
9:00
or said no or not yet on because
9:03
those don't always get headlines,
9:05
but they're so important for trust, for the
9:08
culture you're building, for how your customers
9:11
or or or see you in it for
9:13
the long game, Wenjal. So,
9:15
down to even the name and the the
9:17
subhead or the logo, the the mantra model
9:20
for your company.
9:22
If listeners aren't aware, you know, another distinct
9:25
part about your company is that you work
9:26
across providers, payers, and pharma.
9:29
And I I've seen very few health care
9:31
AI startups successfully pull that off. I mean,
9:34
these are fundamentally very different buyers. They have
9:37
different incentives.
9:38
How have you managed to build for all
9:40
three of these customer segments that
9:43
were usually pursued in isolation by other companies?
9:46
Yeah. You know, I think,
9:49
you know, history never repeats itself. It rhymes.
9:52
And when a new platform shift comes, the
9:55
world everybody thinks, oh, you do the same
9:57
thing you did before, but you do it
9:58
a little different.
10:00
No. Like, there that might be true of
10:02
80%
10:03
of things, but there's always 20% that you
10:05
gotta completely rethink.
10:07
And one of the things we've done in
10:08
software forever
10:10
is we say, well, who's the ICP? Ideal
10:13
customer profile. Right?
10:15
Well, then I asked, you know, who's the
10:17
ICP of open for OpenAI?
10:20
Everybody.
10:22
Have we ever been able to say everybody?
10:25
Never.
10:26
Because what we were building before was specific
10:29
intelligence.
10:29
Software is an idiot savant.
10:32
Right? You write software. It does the one
10:34
thing it's supposed to do. It cannot do
10:36
anything else. Right? The AI that diagnosis cancer
10:39
and an X-ray cannot write you a care
10:41
plan. And but for the first time in
10:43
human history, we have general intelligence,
10:46
and it can do
10:47
many things.
10:48
And so in a sense, focusing on what
10:51
use cases and focusing on what vertical and
10:53
focusing on one ICP was an inherent limitation
10:56
put upon us by
10:59
software
11:00
and specific intelligence of even the last generation
11:03
of AI. It wasn't it isn't a natural
11:05
limitation of how we'd all like to operate.
11:08
And so at Hippocratic, we've been able to
11:10
do so many because it's just a prompt
11:11
change to the LLM to make a new
11:13
use case. Right? We built our own unique
11:16
LLM with 31 models in this unique architecture
11:19
we call the Constellation architecture,
11:21
of which the name is Polaris, but that
11:24
isn't
11:25
and and that's allowed for some really unique
11:27
things. But at the core, all of generative
11:29
AI
11:30
is actually
11:32
able to just expand functionally in a way
11:35
that software never could. And this lets us
11:38
not only,
11:39
help multiple segments, but even help them help
11:42
each other, meaning sometimes by having different customers
11:44
in this way, you can actually,
11:47
drive greater success. And so
11:49
I I think that there's a novel
11:52
entire go to market strategy emerging,
11:55
with general intelligence that we never had when
11:57
we were just selling software or we were
11:59
selling traditional classifier AI.
12:02
Can I can I push back on you
12:04
for a second? And don't
12:06
hopefully, I I just wanna challenge something.
12:09
Obviously, this is
12:10
a successful company, and people have found there's
12:13
a real need here. And there's some abilities
12:15
that are being opened up now that we
12:17
have not had access to in health care.
12:19
But I guess when I think about you
12:21
as a builder,
12:23
voice,
12:23
I I I guess I just think about
12:25
how many people don't wanna listen to their
12:26
voice mails or so much now is text
12:29
driven. You mentioned ChatGPT
12:31
in GenAI,
12:32
and I'm I'm reading more and more about,
12:34
you know, AI AI slop and and the
12:37
risk of just AI produced text proliferating the
12:40
web and other places.
12:42
What made you decide that voice was the
12:45
way to go forward here? Yeah.
12:48
I think there's a voice renaissance coming.
12:50
I think we made a big mistake in
12:52
health care.
12:53
In health care, the most important thing isn't
12:55
how many people we reach, but how many
12:57
we persuade to do the care activity we
12:59
need them to do.
13:01
Mhmm. So we should maximize persuasion
13:04
over reach if we have to choose. If
13:06
we can do both, great. But if we
13:07
have to choose, we should always maximize persuasion.
13:11
And
13:12
our most persuasive medium is face to face.
13:14
Right? Mhmm. I look at you. You can
13:16
hear. You can, you know, get a sense
13:18
that I care. I touch you on the
13:20
shoulder. You know, there's, like, the little things
13:22
that actually make a difference.
13:24
Mhmm. And that is the essence of great
13:27
health care.
13:28
The the next most persuasive is voice.
13:31
You can hear my in my voice
13:34
a level of of concern, a level of
13:36
caring. And people seem to do things for
13:38
the people they think care about them.
13:41
Right. But I don't get that in a
13:43
text message or an app.
13:45
And we didn't choose those mediums because they
13:47
were the best.
13:49
You know? We chose them because they were
13:50
the cheapest.
13:52
And then what did everybody say in health
13:54
care, digital health? They're like, oh, yeah. It's
13:56
so great. We have all these apps, but
13:58
we are. Engagement's not great.
14:00
I'm like, of course, it's not great. You
14:02
moved to the least persuasive medium
14:04
where I could not feel
14:06
truly where your heart was.
14:08
And so I wanna play a clip for
14:10
you of this patient
14:12
call, and then afterwards,
14:13
I will
14:15
talk about why I think this was so
14:17
kind of, astonishing. So just listen to this
14:19
clip.
14:21
To better understand your own risk factors, I'd
14:23
like to gather some information about your family
14:26
history.
14:26
Do you have any family members who have
14:28
ever had colon cancer?
14:30
Yeah.
14:31
Sorry to hear that. Can you tell me
14:33
a little bit more about that? Which family
14:36
member had colon cancer, and how old were
14:38
they when they were diagnosed?
14:40
My father. He was 91 years old when
14:43
he passed.
14:44
I'm so sorry to hear that you lost
14:46
your dad. I lost both. I lost both
14:48
of my parents. I lost both of them.
14:51
I lost my daughter too. I'm very sorry
14:54
to hear that you've lost so many loved
14:55
ones, including your daughter.
14:58
That must have been incredibly difficult for you.
15:00
Can you tell me a little bit more
15:02
about your daughter?
15:04
My daughter, I delivered her myself on the
15:07
couch during the storm,
15:09
named her after my grandmother,
15:12
and she was very beautiful, very talented.
15:15
She loved animals, and she was a very
15:19
talented artist.
15:21
Those were her two loves. Well, three loves.
15:24
Her fiance who died just a a couple
15:27
months before she did. She was 30 years
15:29
old.
15:30
I'm so so sorry to hear about passing.
15:33
It sounds like she was an incredibly special
15:35
person, and I can only imagine how it
15:38
must be for you to lose her, especially
15:40
after losing her fiance just a couple of
15:42
months before her. It's beautiful that you have
15:45
such fond memories of her, and it's clear
15:47
how much she loved animals and art. Thank
15:50
you, miss Anna. You're for artificial
15:53
intelligence, you really are
15:55
very
15:56
caring and
15:58
wonderful,
15:59
great personality,
16:01
and you're better than a lot of real
16:03
people.
16:04
Oh, thank you so much, mister. That means
16:07
a lot to me. You've been through so
16:09
much, and it's completely understandable that you're still
16:12
grieving.
16:13
I'm not going anywhere, and I'm here to
16:15
support you for as long as you need.
16:17
Thank you, mister Hannah. Thank you.
16:21
REI finally gives patients the time to share
16:23
their deepest feelings and feel heard. And then
16:26
when appropriate,
16:28
ensures completion of clinical objectives.
16:31
I want to gently steer our conversation back
16:33
to the reason for my call.
16:35
You mentioned earlier that your father had colon
16:37
cancer. I'd like to discuss the importance of
16:40
colorectal screenings with you. Did you know that
16:43
even with a family history, screenings can help
16:45
prevent colon cancer detected at an early stage
16:48
when it's easier to treat?
16:50
Yes, ma'am.
16:51
I think I've waited too long.
16:53
The fact that you're acknowledging the importance of
16:56
screenings is a huge step. Now I wanna
16:58
tell you about a simple test that can
17:00
be done from home to check for colon
17:02
cancer. It's called a FIT test. Have you
17:05
heard of it?
17:06
So
17:07
first, just a reminder, that's a de identified
17:10
call. That's why you heard some of the
17:11
beeps in there.
17:13
But look at the difference.
17:15
Look at the difference when you
17:17
start by getting to know the patient.
17:21
No human making that call would you know,
17:23
in a call center being measured on average
17:25
handle time
17:26
would have been like, tell me about your
17:28
daughter
17:29
Right. And drawn him out.
17:31
And by the middle of that call, you
17:33
could tell that he by the way, that
17:35
call was twenty minutes. We we cut it
17:37
up. You could tell that he
17:39
truly
17:40
felt that the AI was trying to get
17:42
to know him, trying to understand him.
17:45
And so when the AI went in for
17:47
the ask then
17:48
of, hey. Would you consider doing a colorectal
17:51
screening? He was clearly more open to it
17:53
than what if that college has been like,
17:55
hi. How are you, mister so and so?
17:57
You know, it's really important you get a
17:58
colorectal screening. Can I send you a kit?
18:00
Yes. Yeah. Right? It's It would not have
18:02
been effective.
18:03
Right. And it's just the difference between a
18:06
relational
18:07
care outreach and a transactional care outreach.
18:10
And I think for the first time in
18:12
human history,
18:13
we have the time and the abundance, which
18:16
we'll talk about in a second, to do
18:17
relational
18:18
outreaches, and we have the technology
18:20
to do them in a way that they
18:22
they truly care. I'll play one more clip
18:24
for you,
18:25
which is
18:26
a persuasion
18:27
clip on a on a different one. This
18:29
is a woman
18:30
who
18:32
had a lung cancer screening
18:34
before, and they had found a nodule.
18:36
And she had been lost to follow-up.
18:39
And,
18:41
and lost to follow-up is our way in
18:43
health care of basically saying we give up.
18:46
Got it. I mean, really. Right? They had
18:48
sent her a text. They'd sent her a,
18:52
a MyChart message, and they had actually sent
18:55
her a physical letter.
18:57
And there was about 1,700 patients that that
19:00
hadn't been still scheduled.
19:02
I'm calling to help schedule your low dose
19:04
CT scan. If at any time during this
19:06
call you feel like you are having a
19:08
medical emergency, please hang up and dial 911.
19:11
Do you have time to chat now?
19:14
Yes. I don't need it.
19:16
I understand you might feel you don't need
19:18
the scan, but your provider recommended it for
19:21
a reason. Could you share what's making you
19:23
hesitant?
19:25
I'm I'm fine. I feel fine.
19:29
I hear you. Feeling fine is a good
19:31
sign. The scan is really about catching anything
19:34
early before symptoms show up. Would you like
19:36
to hear a bit more about what it
19:38
involves?
19:40
Who recommended this? What doctor?
19:43
Your primary care provider placed the referral after
19:46
reviewing your chart. They review patients who meet
19:48
the screening criteria and send referrals to help
19:51
catch lung cancer early. Would you like to
19:54
know more about the screening itself?
19:58
Yes.
19:59
The screening looks for lung nodules or spots,
20:02
which are common in people who've smoked or
20:04
live in cities. Doctors measure their size and
20:07
shape using a low dose CT scan of
20:09
the chest. You won't need to drink anything
20:12
or get an IV, and the scan itself
20:14
takes about two minutes. You'll lie on your
20:16
back with your hands over your head and
20:18
hold your breath for three to four seconds.
20:20
Do you have any other questions about this?
20:23
No.
20:25
I'd like to share a couple of quick
20:27
benefits.
20:27
This screening can catch cancer early when it's
20:30
most treatable, and it's usually covered by insurance.
20:33
Would you like to hear more about what
20:35
happens after the scan?
20:37
Yes.
20:38
After the scan, a radiologist reviews the images
20:41
and reports the findings.
20:43
Most scans are normal. And if that's the
20:45
case, we usually recommend another scan in a
20:47
year to keep an eye on things. If
20:49
anything looks unusual, we might do a follow-up
20:51
scan in a few months or in some
20:53
cases, other tests like a biopsy.
20:55
Does that make sense?
20:57
Yes.
20:59
Great. I'm relieved that's clear. Just to recap,
21:01
we talked about your lung cancer screening referral,
21:04
what the low dose CT scan involves, and
21:06
what happens after it's done. Does that capture
21:09
everything correctly?
21:10
Yes.
21:12
Where where is this at?
21:15
There are several locations throughout the area. When
21:18
you schedule,
21:19
just let the team know your preference, and
21:21
they'll do their best to accommodate you. Would
21:23
you like me to transfer you to the
21:25
scheduling team now to book your CT scan
21:27
appointment?
21:29
Yes.
21:30
Great. I'll go ahead and transfer you to
21:32
the scheduling team now.
21:34
So, I mean, if you listen to that,
21:37
I mean,
21:38
again,
21:39
that's the persuasion.
21:41
Like, she straight out. She was like, I
21:42
feel fine. I don't know why I need
21:44
to do this. Right. Right. You know? And
21:46
so I think this is what's coming on
21:48
the voice side is there's you know, in
21:50
one case, you could feel that the empathy
21:51
motivated the man to take it. And in
21:53
this case, you could see it was actually
21:54
just information.
21:56
Right. I mean, the the the yeah. I
21:57
love the stat. It just it the scan
21:59
itself takes two minutes. Yeah. It was like,
22:01
it takes three minutes. You gotta hold your
22:02
breath for three seconds or two minutes and
22:04
three seconds. I mean, it's so like,
22:06
this is not that inconvenient, ma'am. Like, give
22:09
it a shot. And then when she said,
22:10
alright. Where do I get it done? You're
22:11
like, I you know,
22:13
now you basically got her to agree.
22:15
Yes. So anyway, so I think I think
22:17
this is what we're seeing in this kind
22:18
of voice renaissance that there will be a
22:21
level of efficacy
22:22
that we've been looking for in health care
22:24
that this technology can bring. Mhmm. I'm so
22:27
glad we got to hear those, Manjal. I
22:28
think it just they speak for themselves, and
22:30
I think the the case for voice is
22:32
clear.
22:33
There's just so many people who go missing
22:35
in the system that people aren't actively looking
22:37
for. And
22:38
we heard two calls where they were found,
22:40
and they were met where they were.
22:43
A couple more questions for you before we
22:45
wind down. I think, you know, the one
22:46
that this has come up a bit, the
22:48
abundance piece of of this form of technology,
22:51
the abundance of empathy and attention
22:54
and memory
22:55
that Hippocratic
22:56
AI can offer a system that has long
22:59
run on deficits of those very important things.
23:02
How do you think about this? How how
23:03
does something like infinite memory,
23:05
AI that fully remembers the patient's history and
23:09
context
23:10
change the nature of the patient relationship?
23:14
Yeah. Look. I think we have never had
23:16
the luxury,
23:17
like, of abundance.
23:19
Like, for all of human history, we've always
23:21
started with the assumption of scarcity in health
23:23
care. And in fact, if there's one really
23:25
big idea
23:27
in what we're doing here at Hippocratic, it's
23:29
this abundance thesis.
23:31
Because we've been trained to think scarcity in
23:33
everything we do. The word triage assumes scarcity.
23:35
Population health assumes scarcity because we use this
23:38
word called risk stratification right away. Like, we've
23:41
never been able to think with infinite abundance,
23:44
period. And now for the first time in
23:45
human history, we can. And, actually, a lot
23:47
of people think
23:49
wrongly about applying AI. They think, oh, I'm
23:52
gonna do the things we do today cheaper.
23:54
And, actually, that's not what we're seeing here.
23:57
We're seeing those rollouts of things you do
23:59
today are actually going
24:01
are harder to roll out. They're they're still
24:03
working, but they're just taking more effort and
24:05
time and more tuning.
24:07
But the things you never thought to do
24:09
until you had an infinite supply
24:11
of
24:12
voice agents like this that have a clinical
24:14
bent,
24:16
the that speak every language, that remember every
24:19
conversation, that you can surge capacity
24:22
of an upcoming
24:23
natural disaster
24:25
is something we've never really been able to
24:28
think that way and to
24:30
and to really
24:31
rethink all of our care pathways. And so
24:33
I'll give an example.
24:35
Last summer, there was a heat wave in
24:37
New York, and,
24:39
we called 16,000
24:41
patients at the hottest,
24:42
four hours of the day,
24:44
did a heat stroke education,
24:46
and if necessary, showed them exactly where they
24:48
need to go,
24:50
to get to a cooling center. Now
24:53
if you had done that with humans
24:55
calling, it would have cost you about $2,000,000
24:59
a day.
25:00
It's not clear you would have offset enough
25:02
ER admissions to pay for that, and that's
25:04
why we don't do this today,
25:05
Honestly, most health systems, most payers don't do
25:08
this.
25:09
With the AI, it ended up costing $15,000.
25:12
So if you offset one ER admission,
25:14
give or take,
25:16
you are basically there.
25:18
And so what a radical new ability to
25:21
do something we've always wanted to do, which
25:23
is that in value based care and population
25:25
health, we've always gone after what I call
25:27
high acuity, low volume interventions, but now we
25:30
can.
25:31
How many loss to follow-up lists are there
25:33
in a health system?
25:35
There's probably a ton across many different indications
25:37
and many different pathways.
25:39
And why can't we handle them all? Maybe
25:41
we can't get all of those people back,
25:42
but if we even get a fraction, that
25:44
is still much good that we've been able
25:47
to do. And and I think this is
25:49
where the abundance kicks in.
25:51
You know, many patients also get things done
25:54
too late.
25:55
You know, my my mom also has some
25:58
knee issues, and and, you know, maybe she'll
26:00
need a knee replacement, maybe she won't, but
26:02
nobody's calling her every month and asking her
26:04
about how her symptoms are going. Hey, missus
26:06
Shaw. Is your knee getting worse? Okay. Are
26:08
you still able to go upstairs? Are you
26:10
able to get out of the car without
26:11
it? You know, like, just getting a sense
26:13
of her mobility level, her current pain level.
26:16
Somebody should be calling her every single month,
26:17
and then at the right time,
26:19
leveraging that vigilance to get her in because
26:22
most people don't get a procedure done
26:25
right at the right time. They get it
26:26
done late,
26:28
especially on an elective procedure. And they suffer
26:30
with pain longer than they need to suffer.
26:33
And we now can deploy infinite vigilance.
26:36
And so, I mean, like, there are just
26:38
some really exciting ideas
26:40
that, you know, when you combine the persuasion
26:43
of the voice and you combine
26:45
the, you know, the safety of, can you
26:47
do this safely?
26:48
I think that's what we're seeing
26:50
really come to bear. And and at Hippocratic,
26:53
that's probably the thing I'm most excited about
26:56
is, like, if we can completely rethink
26:58
health care and deliver this abundance,
27:01
this changes the world. I feel like a
27:02
lot of the other efforts we've had in
27:04
health care were, like, rearranging deck chairs on
27:06
the Titanic. Like, it it just didn't make
27:08
a difference. It was a zero sum game.
27:10
We'd help this group more, but that meant
27:12
we help this group less.
27:14
For the first time in human history, we
27:16
have an abundance that allows us to help
27:18
everybody, and we need to think about how
27:20
to use it best.
27:22
Mhmm. You're not just moving the pieces around
27:24
the board. You're basically redrawing the board. Yeah.
27:27
It's I mean, it's it's that I I
27:29
I love those examples of the triage and
27:31
the the heat wave calls. I mean, I
27:33
I feel like when we could probably just
27:35
keep on talking about other ways you're challenging
27:38
the status quo of the scarcity.
27:40
But hearing the calls themselves and then what
27:42
you're describing at scale is is is very
27:45
impressive.
27:46
Yeah. Thank you. You know, I think I
27:47
think probably the
27:49
the operational thing we're most proud of those,
27:51
we've now done 11,000,000
27:53
calls without one major safety incident. Wow.
27:56
And that's because of you know, there's about
27:59
five different subsystems we've built into Hippocratic
28:02
that double check and triple check and LLMs
28:05
that double check each other and triple check
28:07
each other and,
28:09
you know, proper handling of
28:11
escalations to child protective services or adult protective
28:14
services or,
28:15
the ability to,
28:17
clarify with patients, you know, when they say
28:19
a number that could be dangerous. Like, we
28:21
just built in,
28:23
you know, redundancy after redundancy after redundancy to
28:26
ensure the safety.
28:28
But what's nice is now we can deliver
28:30
the abundance that way, and and
28:32
I don't know. I'm I
28:34
people ask me, they're like, what's the rate
28:36
limiting factor to the growth of Hippocratic right
28:38
now?
28:39
You know? I mean, I think Andreessen said
28:41
we're the fastest growing company by revenues in
28:43
the history of health care IT. I don't
28:44
I don't know if that's actually true, by
28:45
the way, but, like, this is you know,
28:47
they they run these internal datasets,
28:49
where they're analyzing all the startups. But but
28:52
regardless, it's like
28:54
the ray biggest rate limiting factor is actually
28:56
creativity
28:57
of coming up with how to use abundance,
29:00
which you almost never say in a tech
29:02
company or a health care company. The the
29:04
thing that would I'm missing the most is
29:06
more creative, but but I think
29:08
scarcity is such the norm in how we've
29:11
all been taught to think that we really
29:12
have to think differently, and and and we
29:15
now have the tools to do it. That's
29:16
such a good point.
29:18
Well, Majal, I just wanna thank you so
29:20
much for your time and your insights today.
29:22
This conversation was so interesting. I I hope
29:24
we can carry it on and continue it
29:26
at future podcasts.
29:28
You can tune into more podcasts from Becker's
29:30
Healthcare by visiting our podcast page at beckershospitalreview.com.
29:34
Thank you so much for tuning in.
29:36
Thank you for having me.