Latest / The Advancing Healthcare Innovation Show / AI Can’t Just Be Another Healthcare Project | Justin Norden, MD | E199

Transcript

AI Can’t Just Be Another Healthcare Project | Justin Norden, MD | E199 · The Advancing Healthcare Innovation Show

0:01
[music]
0:11
Hello and welcome back to the Advancing
0:13
Healthcare Innovation Show. I am your
0:15
host, Michael Samatinos, and if you're
0:17
new to us, awesome to have you here.
0:20
We're having some incredible
0:20
conversations with real leaders that are
0:23
painstakingly reshaping the future of
0:26
healthcare. And if you haven't had an
0:28
opportunity to check out any of our
0:30
historical interviews, wherever you are
0:32
now, if you're watching this or if
0:33
you're listening to this, please hit the
0:35
subscribe button. Now, we're really
0:37
passionate about a few different areas.
0:38
We're passionate about access,
0:40
innovation, transformation, and
0:42
leadership. And the objective here, I'm
0:44
just kind of taking a step back, is is
0:47
we're trying to get the real essence of
0:49
people's stories, you know, not just the
0:51
wins, the learnings as well and sharing
0:54
them with our community and ultimately
0:55
with the world. So, I am uh super
0:58
excited about today's guest. But before
1:00
I kind of dig in, you know what's
1:03
actually really wild to me right now is
1:05
that every health system in the country
1:08
is buying AI. Every single one of them.
1:12
And yet I think when you sit back and
1:15
you ask many of them like how is it
1:17
going like how is it really really going
1:19
the answer is usually
1:22
like some version of the same thing.
1:24
It's like yeah we've got tools and we're
1:26
just not sure we have control anymore.
1:29
Like that's a real problem. And what I
1:32
find fascinating about today's guest is
1:36
that he has already solved a version of
1:39
this problem before. it just wasn't in
1:42
healthcare. So, Dr. Justin Nordon built
1:45
a company called Trustworthy AI and the
1:48
whole thing was about making AI systems
1:51
safe enough so you can actually trust
1:52
them. And back in, I think it was him in
1:56
2023, sat around a table full of
1:59
executives and tech leaders and he
2:01
listened to them. Uh they were
2:03
describing exactly what I just said. So
2:06
they bought the tools, they just kind of
2:08
lost the wheel. And maybe he can verify
2:12
this, but it seems as though like him
2:14
starting this company, he just couldn't
2:16
stop himself from starting this company.
2:18
And this company is called Qualified
2:20
Health.
2:21
They initially launched out of the gate.
2:23
They had raised somewhere around like 30
2:24
million bucks. Uh they just closed
2:26
another round a little while ago. I
2:28
think over
2:30
140 million or something like that, just
2:31
to keep me honest. Um, but they're
2:34
working with over a dozen health
2:35
systems, really big names. You would
2:38
know all those names and he can mention
2:40
them. But I think one of them had
2:43
already documented some savings around
2:45
like $15 million in under six months.
2:48
That's amazing. Justin's a physician.
2:51
He's a computer scientist. His CV is
2:54
absolutely ridiculous. Stanford has been
2:56
at Apple. Um, is also a venture
2:59
capitalist at GSR Ventures. He was
3:01
backing AI companies. So now he's he got
3:04
an itch again. He's in the seat. He's
3:06
seen healthc care from almost every
3:07
single apple uh angle and um we're going
3:10
to have him on the show. So Justin,
3:12
great to see you. Welcome to the show.
3:16
>> Well, Michael, thank you for having me.
3:18
And let's be clear, I can live up to
3:19
none of those wonderful, very kind words
3:21
that you just said. So let's reset
3:22
expectations. Super looking forward to
3:25
the conversation. Before we get into
3:27
anything like super serious, um, I got
3:29
to ask you about something that I found
3:31
out through one of your customers, and I
3:34
think it's actually the most interesting
3:35
thing that's on your bio. You're a
3:37
threetime world champion in ultimate
3:40
frisbee, team USA captain, alltime
3:43
assists, and completions record for your
3:46
team. Like I I need to know like how
3:48
does someone who is doing all of that uh
3:52
at the same time as medical school
3:55
building companies
3:57
actually like how the hell do you
3:59
function like what like what's going on
4:00
in your head?
4:02
>> Uh what so yes I guess those things are
4:06
all technically true. Uh you pick a
4:09
really niche sport
4:11
key item one you know I I grew up
4:14
playing everything. uh have an older
4:16
brother, every sport imaginable, you
4:19
know, baseball, basketball, soccer, you
4:20
know, early on had a dream like many,
4:23
you know, young boys do, be a
4:24
professional, be a professional athlete
4:26
someday. Uh I did not at the time uh
4:30
think that would be ultimate frisbee,
4:32
but ended up choosing college based on
4:34
ultimate frisbee based on some amazing
4:36
coaches and kind of falling into it on
4:38
accident at the end of middle school and
4:40
early high school. And then, you know,
4:43
because it's not as serious, let's be
4:46
frank, as the time come as some other
4:48
sports, you know, was able to compete
4:49
through med school and after um for for
4:52
quite some time. And so, uh no, an
4:54
absolutely amazing uh sport people um
4:59
and was wonderful to you know, keep
5:01
something in your identity, especially
5:03
through the hard years of medical
5:04
training and things like that. Uh but
5:06
you know sadly uh with uh new company
5:10
starting qualified and with kind of now
5:13
two uh babies uh two two kids uh the the
5:17
cleats have been hung up for some time
5:19
now. So maybe maybe someday we'll go
5:21
back.
5:22
>> I I I understand. Um the other thing I
5:25
wanted to know is look you're a
5:26
physician you're a computer scientist. I
5:29
think those are two two different
5:31
callings. Um, was there like a specific
5:33
moment, maybe a a patient, uh, a a
5:37
professor or even a problem that you ran
5:39
into where those two worlds sort of
5:42
collided and you thought like this this
5:45
is like where I need to be. Like take
5:46
like take me back to that.
5:48
>> Yeah. Uh, I guess it really starts first
5:52
with my family. So I grew up in Seattle
5:56
and came from a family of physicians. My
5:58
father was a physician. My mom wasn't,
6:00
but three of her siblings were. And so,
6:03
you know, like many, you know, people
6:04
who go into medicine, there's medicine
6:06
was around. It was never pushed by
6:08
family. It was never, [clears throat]
6:10
hey, you have to do this. But,
6:11
>> you know, as I was excited and wanted to
6:13
learn more about science and things like
6:15
that, there was support. Um, but the
6:17
other relevant piece was I, you know,
6:18
grew up in Seattle and went to call kind
6:21
of the Bill Gates Microsoft grade school
6:22
high school
6:24
>> and was always interested in computers.
6:26
you know, built my own computers, early
6:28
bits of programming, and you know, at
6:30
the time we were, you know, handed
6:32
laptops in middle school, which at the
6:33
time was weird,
6:34
>> you know, and so when I look back, these
6:38
are kind of the two forces. They I
6:40
didn't know I was going to combine them
6:41
at the time. I didn't know that was
6:43
going to be my career, but these were
6:44
just two separate interests. and
6:47
ultimately, you know, was going to do
6:49
pre-med in college, but had an
6:51
incredible professor who kind of
6:53
convinced me to actually switch to do
6:55
study computer science um as my major,
6:58
which fulfills zero requirements, right,
7:00
of the premed, you know, curriculum
7:02
where if you do something like biology
7:04
or chemistry, but just had a truly
7:06
incredible professor who went out of his
7:08
way um because I was interested and it's
7:11
no, I promise you this will come back
7:13
someday to be useful. And that same
7:15
professor convinced me to go out to
7:16
England to masters in machine learning
7:18
and genomics. And then you know by this
7:22
time now coming to Stanford for medical
7:24
school which was
7:25
>> kind of the obvious choice now trying to
7:28
put these two things together.
7:30
>> Um you know that was kind of the the
7:32
start you know early I thought it would
7:34
more be on the research side. That's
7:36
what you know as a medical school I did
7:38
>> cancer research data science you know
7:40
computational biology work. Um, but I
7:43
think the thing that really got to me in
7:45
medical school was, you know, incredible
7:48
classmates who were going to do new
7:49
basic science, develop new drugs. I was
7:51
fascinated by that, but that wasn't me.
7:53
>> I think what drew me most was
7:56
>> as you're there starting to learn about
7:58
patient care, you know, the thing that
8:00
kind of would break my heart is
8:02
>> all of these patients who came in for
8:04
things we know how to solve in medicine.
8:07
We just didn't do it for whatever
8:09
reason. we know the guidelines, we have
8:11
the drugs, it's not even about costs,
8:13
which you know we could talk about. It
8:14
was just people fell through the cracks
8:17
and this was something that coming from
8:18
a technology background and seeing these
8:20
guidelines was something I was super
8:21
drawn to. And that really kind of kept
8:23
turning the cranks to Apple or center
8:25
for digital health building AI tools in
8:27
their practice. And candidly, some of
8:29
the work I'm most proud of we're doing
8:31
today with our systems and partners is
8:33
just closing this gap between what we
8:35
know how to do and medicine has already
8:37
figured out without a doubt and just
8:38
getting it done across our
8:40
>> Yeah. you you've had such an interesting
8:42
tour of duty because you've worked at
8:44
Apple and some of these other places
8:45
like and I know Apple is famously very
8:48
secretive and I'm not going to ask you
8:49
to like you know break any NDAs or
8:51
anything like that but what I want to
8:52
know is is through all of these tours of
8:54
duty what what did they in essence sort
8:56
of teach you about trust
9:00
um because your your company has figured
9:02
something out about trust that I think
9:04
many companies don't ever like fully
9:08
track.
9:11
>> So,
9:13
I mean, it's funny. Trust is such a
9:15
complicated word and I'll just describe
9:17
a few dimensions for us. There's, you
9:19
know, we've known each other and had a
9:20
handful of conversations over the years
9:22
and got to build a relationship and have
9:24
many mutual friends. And so, that's like
9:26
one way to build trust.
9:27
>> Another way to build trust is through
9:29
data and performance. there's like
9:31
algorithmic trust you can show with data
9:34
past performance future performance
9:36
running things in shadow evaluations
9:39
and so and both are important right and
9:43
I think often it gets conflated as one
9:46
thing but there's so many different
9:47
dimensions to it and so yeah
9:49
>> you know our
9:51
>> and for healthcare when the default is
9:54
to not change the default is to go slow
9:57
on these new tools to do no harm to kind
10:00
of overcome that when you know we're
10:03
both in this industry because it is the
10:05
most important industry you know when
10:07
you don't have your health nothing else
10:08
matters and so
10:11
>> it just you have to treat it so
10:12
carefully it takes years and years to
10:14
build up in an instant to lose
10:17
>> and so I think coming in kind of with
10:20
this recognition of how important this
10:22
is it came from building the team super
10:26
intentionally you know one of my
10:27
co-founders Kar Mate is our chief
10:29
medical officer officer. Again, he's
10:31
he's been in healthcare his his whole
10:32
career from CMS to IHI to, you know,
10:36
running the IHI for the last 5 years is
10:38
one of the most trusted institutions to
10:40
work on the hardest problems, quality of
10:42
care, improvement and change at health
10:44
systems. And it took you know balancing
10:46
the technology you know one foot in the
10:48
door on what is happening you know at
10:50
open AI and propic and the frontier labs
10:52
and what is happening on the ground in
10:54
healthcare and being able to kind of
10:56
merge these two again being able to
10:58
speak the language and coming from the
11:00
context of years of working with
11:01
healthcare as well as kind of speak the
11:03
language and be on point and trusted to
11:06
stay up on kind of what's happening in
11:08
the frontier labs. And so it it really
11:11
is the core of kind of how we've built
11:13
the company um and how we've built our
11:15
team and then how we build kind of the
11:17
tooling to kind of measure and show that
11:18
we're doing what we said.
11:20
>> Yeah, this intentional curation,
11:22
bringing the right people at the right
11:23
time is paramount. I I want to make sure
11:26
that the audience actually really gets
11:30
like what qualified health is. Not not
11:32
necessarily like the press release
11:33
version, but like the real version. So
11:35
if I'm like uh if I'm a CIO and I'm
11:37
calling you tomorrow like what actually
11:40
happens like what what does that that
11:43
conversation look like the weeks
11:45
unfolding look like?
11:48
>> Yes. So a few things I think you know we
11:50
consider ourselves you know a really
11:52
platform partner for health system AI
11:56
deployments.
11:57
>> So what does that mean? Um, we're having
11:58
a conversation with the CIO. One of the
12:00
first things we can deploy is access to
12:04
frontier models from OpenAI, Anthropic,
12:06
Google in a HIPPA compliant way. Uh,
12:10
PII, PHI preserving as well as training
12:13
to now your entire workforce. This could
12:15
roll out to 20 50,000 employees where
12:18
you [clears throat] know you can prevent
12:19
people from you know using their phones
12:22
leaking patient information as well as
12:24
have kind of a single place people can
12:26
get access to all the new models in a
12:29
safe and secure way. So that's kind of
12:30
one bottom.
12:33
>> The second thing you know especially for
12:35
the CIO lens is we are getting the data
12:38
infrastructure set up for scale. One of
12:40
our core thesis from starting the
12:42
company is instead of the CIO in the
12:44
system having to do 50 different data
12:48
integration, security reviews, audit
12:50
processes, contracting processes
12:53
um to get access to some of the same
12:55
sets of things if we set up that data
12:57
infrastructure once. we can now power
13:00
tens, 50, hundreds of use cases all
13:03
based on that same modern infrastructure
13:05
getting ready for AI with audit trails,
13:07
monitoring and governance attached to
13:09
it. And so that's why, you know, we call
13:11
ourselves an infrastructure
13:12
>> company where we get that foundation set
13:14
up right. And maybe my simple analogy
13:17
for this is it doesn't matter if your
13:20
use case is a revenue cycle use case, a
13:22
care quality management use case, an
13:24
internal hospital operations use case,
13:26
you need these three components. access
13:28
to structured and unstructured
13:29
information, ability to follow a set of
13:32
rules, payer guidelines, clinical
13:34
guidelines, hospital operation
13:35
guidelines, and then source for a user
13:38
the exact data while a rule was followed
13:40
and then taking an action or allowing
13:41
that user to take an action.
13:43
>> It's the same thing technically over and
13:45
over and over again. And so we've built
13:48
the tooling um to really kind of deploy
13:50
use case after use case in the same way
13:52
that can be managed. So there's the data
13:54
infrastructure work that we get set up,
13:56
evaluation monitoring, there's the
13:58
general purpose kind of employee
14:00
workforce at scale, access and training.
14:03
Um, and then there's kind of end toend
14:05
workflows that we're building off of
14:06
those same components to kind of really
14:08
drive [clears throat]
14:09
value scorecards for the chief medical
14:11
officer or really the CFO to kind of
14:14
prove those outcomes that we're
14:15
searching for.
14:16
>> Yeah. I mean, I I mentioned that one of
14:18
your health system partners had
14:20
documented $15 million, you know, in
14:23
impact in in under six months. I'd love
14:26
to hear that story like to to the degree
14:27
that you're able to share like what was
14:30
broken before you got there, you know,
14:32
what did the team inside that system
14:34
actually experience and like and then
14:36
what changed?
14:37
>> Yeah. So I think you know this was one
14:40
of our early partners um down in down in
14:43
Texas and I think you know how did the
14:46
partnership start? I think they were
14:48
really forward thinking in their
14:50
leadership team. You know they were
14:51
starting to use these tools experiment
14:53
with these tools and kind of
14:55
>> wanted to make an early bet about where
14:57
the where the world would go and really
14:58
wanted to partner
15:00
>> on you know not a one-off but really
15:02
doing this at scale. And so that was
15:04
really why there was a fit early to run
15:06
at this. Uh I think the other key
15:08
component was we had a leadership
15:11
behind. It wasn't just the CIO, it was
15:13
the CIO, the chief officer, the CEO, the
15:15
COO, the CFO
15:17
>> and we're able to pick problems that
15:19
could move the needle that they were
15:22
measuring and that we could show a delta
15:25
for for what that change looked like.
15:26
And um so a few things you know I think
15:29
the roll out started like many of our
15:31
others do with kind of access to the
15:34
frontier models for for them around
15:36
15,000 uh plus employees where people
15:38
were able to again go to a safe place to
15:41
get information start to learn start to
15:43
kind of build some of their own tools
15:45
idea generate
15:46
>> different things as first component. um
15:49
one way, you know, they talked about
15:51
kind of what they might have purchased
15:52
for something else. Um which was one way
15:55
u to think about kind of the the cost.
15:57
Uh but then the more interesting ones,
15:59
we worked on a handful of revenue cycle
16:01
problems. There was a few things the
16:02
team had tracked where they knew they
16:04
had gaps compared to kind of national
16:06
benchmarks and guidelines and we were
16:08
able to kind of again use the AI system
16:10
and those components to put something
16:12
together to flag again to their users
16:15
and to the team areas where um they were
16:19
kind of based on guidelines you know
16:21
missing opportunities uh to improve kind
16:24
of what they were doing from a capture
16:25
perspective um and submit that and again
16:28
because we can track the very clear
16:30
baselines we're able to show kind of the
16:32
performance that changes over time. Uh
16:34
and then the the final piece which I'll
16:35
mention um which I briefly mentioned a
16:39
little bit before and actually we've
16:40
published on was work we did with uh
16:43
Enthropic and and the UT folks to go
16:46
through their patient populations in
16:48
cardiology specifically in heart failure
16:50
to start and show where there were gaps
16:52
in changes on medications the patients
16:55
really should have been on. um which
16:57
happens across the country and again
16:59
just because things are busy, chronic
17:00
conditions, you know, things fall
17:03
through the cracks as well as bring the
17:04
patients in who really need the biggest
17:06
changes um which [clears throat] again
17:08
are the right things for the patients to
17:10
kind of get the care and follow the
17:11
guidelines that they need and kind of
17:12
bring the most important patients and
17:14
then prioritize those um who really
17:16
drive the economics for a health system
17:18
as well. And so it's really one of these
17:20
win-winwin things. we're able to bring
17:22
in the right patients, prioritize those
17:24
who really need, you know, your
17:26
specialist access to care and those
17:27
limited resources and prioritize that.
17:29
And so, uh, and then the fun part is
17:31
this is just the start because it's the
17:33
same infrastructure and tools that you
17:35
just apply again and again and again,
17:38
>> um, on the next use case. Uh, because it
17:41
is more work to set up the data
17:42
infrastructure. It is more work to get
17:45
to the strategic level with the
17:47
leadership team to talk about their
17:48
priorities and where it's a good fit.
17:50
Um, but then instead of bolting on one
17:52
more solution, which adds even more
17:54
overhead and even more cost and even
17:56
more management issues, we're putting it
17:58
through that same foundation.
18:00
>> Yeah. It's like, listen, if you're going
18:02
to build a skyscraper, you need to have
18:03
a really sound, solid foundation. You're
18:05
like, all right, let's do this. Let's do
18:07
it once and let's build and go as
18:11
quickly as we want, but we just know
18:12
that we have the the foundation
18:15
to be able to support what it is that
18:18
we're trying to do. Um
18:21
maybe this hasn't been asked you often
18:22
but there's so many companies that are
18:25
out there and a lot of them are you know
18:30
state they do infrastructure and AI and
18:32
this and that how does how does someone
18:35
sort of
18:37
pick through the noise feel like CIOS
18:40
are so uh in general hospitals they're
18:43
inundated bombarded with messages where
18:46
do they distill what's real and what's
18:49
not what what have you found to be sort
18:51
of allows you to kind of cut through the
18:53
noise because I would imagine sometimes
18:54
you even you uh despite you know a
18:57
stellar reputation team it's
18:59
unbelievable kind of get caught up in
19:00
that sometimes
19:03
it you know this was one of the problems
19:05
as you mentioned back to 23 of hundreds
19:09
and hundreds of companies popping up how
19:11
do we know where to start how to
19:13
evaluate and run through and I think
19:15
>> um and this this isn't a new AI problem
19:18
this has been true for decades in health
19:21
IT you know as we've gone through health
19:22
IT digital health you know our different
19:24
names that our industry has evolved over
19:25
time and um you know at the end of the
19:29
day you know what actually happens
19:32
people want to talk to your customers
19:33
and partners and hear results you can
19:35
have fancy marketing language you know
19:38
uh we haven't done a lot we don't have a
19:39
marketing department we don't have a
19:40
marketing team we barely we don't really
19:43
have a
19:45
we barely have a have a has a salesforce
19:48
Um because it's and companies
19:52
go at this in different ways because the
19:53
most important thing we do is deliver
19:55
for our partners period.
19:57
>> Yeah. Your customers and your business
19:58
level that we're doing we it and it's I
20:01
feel super fortunate in even how we do
20:03
summer contracting like most of our fees
20:05
are on the impacts that we generate
20:06
together
20:07
>> where we win we win when they win. Um
20:10
and it's an incredible group of partners
20:11
who want to share talk about their
20:14
success and really have a hand in what
20:16
this journey looks like. And so um you
20:20
know I
20:22
wish I had a perfect formula and you
20:24
know a team where yes you create you
20:27
know your perfect sandbox environment
20:29
and have your whole team to do evals and
20:31
hundreds of different things coming in
20:32
and you know while this is kind of one
20:35
of the things you know we can do
20:36
internally like health systems aren't
20:38
set up for that at scale and by the time
20:40
you set it up right you know the models
20:42
are changing and we want access to more
20:43
information and by the time you've
20:45
tested properly and integrated with
20:47
someone's new system. We spent, you
20:49
know,
20:51
hundreds of thousands of dollars just in
20:52
the process, you know, and and people
20:54
power to get to that end point. And so,
20:56
you know, really it this is why, you
20:58
know, we focus so much on being able to
21:00
deliver for for our partners because
21:02
those those customer references are
21:03
everything.
21:05
>> Yeah. Look, at the end of the day, um
21:08
what you say is never going to be as
21:10
valuable as what someone else says about
21:12
you. And if it's a customer who's
21:15
delighted, it is worth its weight and
21:17
gold. That's that's for sure. Um, before
21:21
we kind of wrap, I just So, where do
21:23
where can people follow your work? You
21:25
mentioned you don't have a robust
21:26
marketing department, but like where do
21:28
where do people kind of follow the the
21:29
qualified health story? Give a shout
21:31
out, give a little commercial for that
21:33
before we uh end up wrapping. Yeah, I
21:36
think uh you know we have a website uh
21:40
qualified.com
21:43
you know we have our LinkedIn page uh as
21:46
well as you know it's the only place I'm
21:48
really active I do you know posting on
21:50
LinkedIn from time to time uh as you
21:53
know again it's not just things about us
21:55
it's really about where the field is
21:56
moving you know I still do a a podcast
21:58
for Stanford on healthcare AI and it's
22:01
really a conversation on what's
22:04
happening
22:05
new papers, new developments coming out
22:07
that isn't related to qualified health
22:09
actually. It's just about how do we keep
22:11
up on the field and and the
22:13
conversations that are happening. And so
22:15
those are the places uh please feel free
22:17
to reach out if if you're a health
22:19
system listener and want to have this
22:20
conversation in more depth. Um but it it
22:23
really is just an exciting moment when
22:26
healthc care can change. the tools are
22:28
available now to do so and now is about
22:31
how do you partner and put the pieces
22:32
together and work through the changes to
22:34
our institutions and systems that will
22:36
come of it. you know, how far out do you
22:39
look, you know, as a CEO? Because
22:41
obviously you're a part of a fast
22:42
growing company. There's a lot going on.
22:44
And I asked that question just because
22:47
if you try to peer out 5 years, 10
22:49
years, and if things are going according
22:52
to plan in terms of what how you can
22:55
impact American healthcare,
22:58
how different does it look? How does it
23:00
impact the person in the waiting room?
23:01
Like talk to me about that vision.
23:05
So 10 years is too long. I guess I'll
23:08
start there. It
23:09
>> five years. I mean 10 years.
23:11
>> It's it's too far. Uh which in
23:14
healthcare people often will argue it's
23:15
not too far but at least I'm on one side
23:18
10 years is too far.
23:19
>> Uh
23:21
>> in
23:23
there's a few different ways kind of I
23:25
think about it. You know take this work
23:27
on you know care gaps that we do and
23:29
improving quality of care and
23:31
guidelines. We focused on cardiology
23:33
first, but we're going to do this across
23:34
everything.
23:36
What happens when you have proactive
23:40
guideline driven care on every single
23:42
patient? How does that change? Should
23:44
you see your triage, your throughput?
23:47
These are questions we should be
23:48
answering now. And it'll take a couple
23:51
years to get that through the systems
23:53
and for those to evolve. But it
23:55
fundamentally looks different,
23:57
>> right? Even as you start thinking as we
24:00
have these tools or someone else builds
24:02
these tools, you know, how you think
24:04
about a capital investment on a building
24:06
starts to change, right? From from a
24:08
health system, how you think about a a
24:10
workforce that's using these tools on a
24:13
daily basis and OpenAI published a paper
24:15
last week at internally at OpenAI, it's
24:17
not just their engineers anymore. It's
24:20
legal, finance, customer support that
24:23
has switched all their workflows from
24:24
chat based interactions really to agent
24:26
based interactions to offload tasks.
24:28
>> These things are happening now. And I
24:30
think the difference between
24:33
>> systems who choose to treat AI as how
24:36
they run the business versus as a
24:38
project within two years are going to be
24:42
pretty different. And you know I walk
24:44
you know system partners through you
24:46
know what is the margin level impact how
24:47
much can we change quality of care and
24:50
um
24:52
as I kind of talk through these numbers
24:55
you know forget AI there's a 3 to 5%
24:58
margin change if leaders just did all
25:00
the things that they wanted to do and
25:01
get done what they would typically hire
25:03
Mackenzie for.
25:05
Then you start to layer on kind of the
25:07
impact of AI and revenue cycle,
25:09
caregaps, all these other places where
25:10
it truly can kind of change how a task
25:12
is done and that's at least another 3 to
25:15
5%. McKenzie has an 11 to 19% paper that
25:18
they published last week.
25:20
>> But this is possible now. Forget new
25:22
models, forget all the, you know,
25:24
>> hype coming out from Silicon Valley on
25:26
what might come. That's just today. And
25:29
so it really that is possible today. And
25:32
we're starting to prove that with
25:33
different partners for kind of when we
25:35
really think systematically about
25:37
knocking these things out. And so within
25:40
this year, we're going to start to see
25:42
more results um where systems are truly
25:44
finding our way changing. By the end of
25:47
next year, I think systems are going to
25:49
start to pull away. Um but in other
25:52
places, healthcare will be exactly
25:54
identical. And so I think this is the
25:56
interesting conundrum of as we watch US
25:57
healthcare, but I think you'll start to
25:59
break away into two packs starting this
26:02
year and then really by the end of next
26:03
year systems are going to start to look
26:05
different and it's going to show and
26:07
financials.
26:08
>> They're kind of case shaped. You know,
26:09
basically the ones that are going to
26:10
take are going to be dramatically better
26:12
and the ones are just going to that are
26:14
not on board with this because the way
26:16
that you mentioned it's like hey this is
26:17
this is the new operating system and
26:19
you're integrating it. You're not just
26:21
you said you're you said the it's not
26:24
just a project. You're like no no no
26:25
this is this is a completely kind of
26:27
embedded
26:29
operating system of doing things.
26:32
Fascinating.
26:34
>> That's correct. Again super biased one
26:36
one end of the spectrum but um getting
26:39
to work with leaders who see what's
26:42
coming and get to do that. I just feel
26:43
so fortunate to get to do that every day
26:45
and and the improvements are are
26:47
starting to build up and it's super
26:48
exciting.
26:50
>> Justin, it was awesome having you come
26:51
on. I just love how bold you are in
26:53
terms of like the work that you're
26:54
doing. You're getting out there making
26:56
it happen. I see you on the road all the
26:57
time, so we're both on the road. So,
26:59
it's good to run into each other seeing
27:00
it happen. It's uh I don't say that
27:03
lightly. So, for those that have been
27:04
listening, if you found this
27:06
conversation interesting or if there's
27:07
another leader that you would like for
27:10
us to have come on to the show or a
27:12
topic that you want us to explore, leave
27:15
a note, leave a comment. And if you
27:16
found this particular conversation to be
27:19
enlightening, please share it, like it.
27:21
If you haven't subscribed, make sure you
27:22
subscribe now because it is important.
27:24
This is an interactive community. So,
27:27
thank you for listening. Keep building
27:28
with purpose and we'll see you next time
27:31
on the Advancing Healthcare Innovation
27:33
Show.