AI-Driven Healthcare: Moving Beyond Analytics to Actionable Insights

In the inaugural episode of "Hard Problems, Smart Solutions," Newfire's VP of Data and AI, Gordon Wong, hosts Emily Lindemer, Executive Director of Data and Healthcare Innovation at JPMorgan Chase & Co.

They discuss AI's transformative potential in healthcare, emphasizing the importance of streamlining data and addressing administrative inefficiencies to improve patient outcomes. The conversation explores Emily's career journey, the current state of AI in healthcare, and practical advice for organizations looking to implement AI solutions with a focus on technical, regulatory, and ROI perspectives. Particular emphasis is placed on addressing healthcare disparities through robust data infrastructure and scalable platforms, with strategic insights for organizations looking to leverage AI effectively.

Key takeaways include focusing on feasible, high-impact applications and the necessity for robust data infrastructure.

Creators & Guests

Host

Gordon Wong

Head of Data and AI at Newfire Global Partners

Guest

Emily Lindemer

Executive Director of Data at Morgan Health, JPMorgan Chase & Co.

What is Hard Problems, Smart Solutions - The Newfire Podcast?

Join Newfire’s engineering, product, data, and people experts as they tackle today’s most pressing technology questions alongside industry leaders from some of the world’s most notable companies. This is where the hard problems you’re facing finally get the smart solutions you were looking for.

Welcome to Hard Problems, Smart
Solutions - the Newfire Podcast, where

we dive into complex challenges and ways
to solve them with top industry leaders.

I'm Gordon Wong, VP of Data and
AI at Newfire Global Partners,

and your host for this episode.

In each episode, we bring you
conversations with top innovators

and decision-makers, tackling the
biggest issues across industries.

Whether you're looking for insights to
drive your own strategies or to learn

from the best, you're in the right place.

Let's get started.

Hi, everybody.

Welcome to the first episode
of Hard Problems, Smart

Solutions, the Newfire podcast.

Today, I'm thrilled to introduce
our guest, Emily Lindemir, Executive

Director of Data and Healthcare
Innovation at JPMorgan Chase Co.

Emily has led data science
initiatives at Cityblock Health and

Wellframe, focusing on innovative
AI solutions to improve healthcare.

She holds a PhD in medical
engineering and medical physics

from the Harvard MIT Division of
Health Sciences and Technology.

Emily has been instrumental in
advancing how we use data to

improve healthcare outcomes.

This conversation is especially
timely, as many organizations Including

us and our clients at Newfire.

We'll look at how AI can move
the needle from analytics to

actionable insights in healthcare.

Welcome, Emily.

Thanks, Gordon.

Great to be here.

Excited for the conversation.

Yeah, me too.

I've been looking forward
to this for weeks.

Um, so a little bit of background
for those of you who are listening.

Emily and I actually have crossed
paths multiple times within the

Boston healthcare ecosystem.

Uh, we've actually worked at some
of the same firms, we've consulted

with each other as colleagues.

And so I have a lot of respect for Emily.

So I'm, again, I feel really
privileged to have this opportunity

to have this conversation with her.

And so today's topic is, you know, when
we were looking for guests to explore

this topic, Emily is really one of the
first people that came to mind, because

of her focus on data science and AI and
really, impactful, actionable insights.

Today, we're going to dig into AI's
role in healthcare transformation.

Right, so Emily, you've been at
the forefront in this industry for a

while, and arguably, your entire career
and education has led to this point.

So, how do you see AI evolving from
really just analyzing data and producing

descriptive analytics to driving
actionable business focused solutions?

It's a great question, Gordon.

This is one of my favorite topics
and probably one of the things

I get asked about the most.

Before I answer, let me give a little
bit of an overview of kind of like

where my journey in AI healthcare
has been just for listeners to

see kind of like what I've seen.

Um, I, I started my career in the
imaging world, which is where AI,

I think, really started showing
its first promise in healthcare.

Um, it is the perfect application of AI

images are, um, you know, we
think back to, like, Identifying

cats versus dogs, right?

Like the early, early applications
of AI that were across the world.

Like you can apply them
to medical images as well.

Um, and so I started in neuroimaging
and one of my first jobs out of grad

school was working at IBM Watson Health,

where I was part of their AI imaging
team trying to make new algorithms

that we were bringing to market,
um, in mammography, chest CT.

My career slowly transitioned out of
imaging more to broader health systems,

more broad healthcare applications,
kind of working with digital health

companies, working with, uh, providers
like Cityblock, where Gordon and I

crossed paths, um, and really kind
of seeing more at a global scale

rather than singular applications, how
AI, and just data science in general

was evolving across the industry.

And so just to say, I did all of my
training like way before LLMs came out.

Um, there is, we are in a new generation
of AI when we're thinking about

LLMs and generative AI and, um, and
so I have a lot of thoughts there.

I think that where we are today is that
there's a lot of excitement about AI.

People are making discoveries, training
algorithms to be incrementally better

than the last and sometimes more than
incrementally at like light speed.

But if you know healthcare, you also
know that healthcare kind of like

exists on somewhat of a delayed timeline
when it comes to data and technology.

A lot of our data and infrastructure
in the industry isn't there yet

to support these super advanced
technological applications.

So, I am kind of coming at this
conversation from a place of, um, I've

kind of, I've seen under the hood of what
our data infrastructure often looks like,

and, and I'm coming at it with this view
of a little bit of skepticism, honestly.

So I'll, I'll give you that as sort
of a backdrop for my perspective.

And it really is just my
perspective that I'm sharing here.

So let me, after I say that, let me give
the question back to Gordon and kind

of where do you want to go with that?

Yeah, absolutely.

So that actually resonates
with me quite a bit.

Um, because as a, uh, someone who's
been building data platforms for

more years than I care to count,
um, we've been asked do many

cutting-edge things, but frequently
the organizations are not ready for it.

So I use the term pragmatic
cutting-edge a lot, and I suspect

that might resonate with you a bit.

I love that.

So let's, let's really
drill into that then, right?

So we, we are really looking
for, you know, solving for

actionable business problems.

Right?

So, you know, what, what, where are
you seeing progress in terms of you

using AI to actually move the needle?

So I think about AI and healthcare
in like these two broad camps.

There's administrative
and there's clinical.

Um, and people are interested in
both and there's really, really

important applications in both.

If we start with administrative,
administrative are the things like, rev

cycle management, billing, coding, um,
even just scheduling patients, like, all

of, all of these things in healthcare that
aren't directly caring for the patient,

but cause a lot of burden on the industry.

Clinical applications are the things
that I think we think about as like

these really sexy, like, going to move
the needle and change patient care.

Those are things like automatically
diagnosing a patient or, you

know, prescribing like the
absolute best treatment regimen

based on a patient's history.

Um, I think that where we are today
is really just the first camp.

The real promise and the early
promise of AI and healthcare is going

to be on that administrative side.

I say that, and it's maybe not as
exciting, like, Oh, why can't we do

like these crazy futuristic things?

I'll get to that.

I think the industry is just not
there yet, but I also wanted to

say with positivity about, or
speak with optimism, I should

say, about those administrative
tasks and the promise there.

And the reason is that part of the reason
that the US healthcare system is just so

complicated, so expensive, so burdensome,
is the tremendous amount of administrative

burden that we have placed on it.

And so actually using AI to alleviate
some of that burden is a great thing.

And I think will actually pay really
meaningful dividends in the end for

patients, providers, the system at large.

Um, so I can go into details there, but
just to say, like, I think about those

two camps, and I really think about the
administrative side is where we're going

to see change in these next few years.

I, you know, I actually
really agree with you.

I see that, I, I see that too.

And I've, I've, I've seen the
statistic that perhaps something

like 40 percent of our healthcare
spend goes into administration.

So it feels like of opportunity there.

That's right.

And I mean, you have to, I think,
believe that a lot of things will

chain together and, and lead to that
trickling down to patients having

cheaper and more affordable care.

But that's what I think
the goal should be.

As an, or as an industry, when we think
about, okay, we're going to really

tackle this administrative problem with
AI, I think that we should say more

than just: let's make the burden lower.

I think that the goal should really
be, and make it cheaper for patients.

Make the whole system, patients included,
less burdened by this extreme cost.

Because you're right, it is up
to 40 percent administrative

costs in some cases.

So at Newfire, I'm heading
up data analytics and AI.

So I have the opportunity to speak to
a lot of our prospects in terms of how

to use AI within their organizations.

And so if, you know, what's your
guidance to them, like what problems

should they look at first in terms
of using AI to, to improve things?

Um, so my advice there is, I think
this is an obvious question for a

lot of business leaders, but it's
worth saying is, you know, what is

the ROI if you solve this problem?

Um, a lot of people, like I said,
they really want to tackle these

really exciting clinical things, and
I don't want to discourage from that,

but we actually don't know what
the ROI is often of solving

some clinical problems with AI.

I think that the ROI is much more
clear for some of these administrative.

So if you're a business leader, I
mean, that really should be the first

question you're asking yourself.

Um, you know, the second question
is kind of this technical

feasibility question, right?

If you are thinking of building and
selling an AI product as a company,

you have to think about do all of
your customers have the technical

infrastructure to support this?

Do they have, like, the
data to support this?

And that is often not the
case right now in healthcare.

There are a lot of health systems
out there who are, like, just

migrating to the cloud, you know.

They, they don't have the tools to
load all their data into some kind of

AI algorithm and get an answer back.

Um.

So that's the second thing.

I think the other thing is
generalizability, you know, is what

you're building something that can
generalize, can safely exist outside

of some very, very small testing
grounds that you've built it in.

That's a trap that a lot of people fall
into, I think, with AI, is they can

build something that performs amazingly,
and then they take it and they try to

have it perform somewhere else and it
doesn't do as well, and I think the

trust starts to really erode in the
industry, um, with AI for those reasons.

And I think the last thing is, like, keep
an eye towards the regulatory concerns

around what you're trying to build.

Those administrative tasks that I
mentioned are often not so, they're

not so prone to regulatory challenges,
whereas the clinical side of things,

that's a big hurdle to get some kind
of AI cleared by the FDA for actual

patient use in the clinical setting.

So, you know, keeping those things
in mind, I think, is really critical.

Now, acknowledging that healthcare
companies, uh, come in lots of different

sizes and different flavors, but I think
about Fisher Price, my first AI project.

Um, which C level officer would
you typically think would be like a

good first customer for a user AI?

That's a great question.

I have, um, I've been talking to some
folks about this recently, and I was

just invited to, like, a roundtable to
listen to chief information officers and

how they're being approached about AI.

And interestingly, I think what's
happening right now is I think

it's the CFOs of most organizations
who are being approached about

AI and the decision-makers

and that comes back to that ROI question
and the fact that a lot of these are

really being targeted for a lot of
these applications are being targeted

for administrative tasks, internal
efficiency gains, things like that.

So what you might think of as like
these CIO, CTO folks who are first

approached decision-makers, I think we're
actually seeing that it's more people

on the financial decision making side.

Yeah, that, that resonates for sure.

You know, part of the pur purpose
of this, uh, this podcast is really

to give our listeners some ideas of
where they can get started, right?

Pragmatic advice.

In that vein, do you have any specific
examples from your recent history

where AI has made an impact in terms
of patient outcomes or operational

efficiency or just reduction of toil?

Definitely.

So some of the applications that I've
seen that I think have been really

exciting in AI, that are, they, they
just are adjacent to clinical, but they

really are administrative applications,
are things that help reduce the

burden of clinical documentation.

So, really solving this problem of
providers are getting burned out,

they're spending way too much time
having to write up patient notes, sift

through patient notes and synthesize
past medical information about a patient.

When you kind of boil that down, those
are information synthesis problems,

which is what AI is really good at.

The applications that I've seen that
do things like ambient documentation.

So for example, a voice recording,
just like you and I are doing right

now, of a patient-doctor conversation,
which gets translated to text.

That's easy.

That's been around for a while, but is
then with an LLM translated into medical

jargon that can go into a patient's chart.

That's incredibly powerful.

That, you know, that saves
the doctor so, so much time.

And I think I've seen amazing reviews
and feedback from clinicians themselves

on how powerful that type of technology
is going to be in the clinic.

I've also seen a lot of applications
of AI that I think are really

promising from like an investment and
investability perspective of revenue

cycle management, billing and coding.

Those are heavily manual processes
right now that are prone to error

that I think AI is excelling at.

And then this is my favorite example, just
kind of throwing back to my imaging days.

This is not AI the most recent example
that, I encountered this years ago,

but I remember at an organization I was
at, we were working on really advanced,

really cutting-edge neuroimaging where
we were developing technologies that

could, like, automatically segment images
of the human brain into all of these

amazing, very descriptive statistics.

And it was hard to sell that because it
was hard to tell a clinician why this

would help improve patient care or like
why this would save them money for many

nuances related to like human brain stuff.

But um, it was amazing technology
that couldn't find a home.

Conversely, there was this technology
coming out at the same time that could

take a cue of like a hundred images in an
emergency room and automatically identify

if any one of them was a brain bleed.

And just for those listening who don't
know, almost any of us could be trained to

a spot a major brain bleed pretty easily.

It's a very visually obvious thing.

But so what this algorithm did, it
wasn't like finding some really hidden

issue, but it would find something that
was incredibly emergent that might be

really low in the radiologist's read
queue and pop it to the top and say,

basically, if you don't read this image,
this patient might die in the next hour.

And that's like an
orchestration task, right?

That's like a clinical workflow
task, but it was such a better at

the time application of AI, if you
really wanted to, like, help patients.

So I just put those two side by side.

It's one of my favorite examples
here of, you know, sometimes the

simpler solution, the better one.

I love that.

I love that.

I mean, it reminds me that we should be
always thinking about kind of the basic

variables that drive outcomes, right?

Time, timeliness, reduction in
effort, all these things, right?

So, you know, uh, stealing from the
product manager book, so looking at

your typical, let's say healthcare
provider, what person or role would

you want to target to, uh, delight
with AI within that organization?

Whose job do you make it easier?

How do you help some be more successful?

That's such a good question.

So, personally, I think we should
be really making things for patients

and physicians, like, the people
delivering and receiving care and

physicians need so much tooling
like this to help with burnout.

So I think delighting the physician,
the ambient documentation, question

or example I gave is perfect there.

However, usually physicians are
not the financial decision-makers,

they're not purchasing your product.

And so, I do think that you really have
to be able to always show delight to

the customer that's paying in the end.

And so, being able to say to them, this
is going to allow your physicians to

spend X amount more time with patients,
or, show some real financial ROI there is

really, really important to keep in mind.

Yeah.

Thank you for that.

I, really, that really does
make a lot of sense to me.

You know, I personally have always
loved the lens of like trying to delight

somebody because we could understand that.

But let's take a, let's get a little
speculative for a second, right?

Think about the future AI in healthcare.

What do you think, what
are you excited about?

What do you think are the big
opportunities and challenges

we might be able to tackle?

So what I'm excited about, this
is, this is a hard question.

I've actually been asked this a
few different times and I'm always

kind of evolving my perspective,
I think, on a regular basis.

Let me zoom out before I answer what
I'm most excited about and tell you what

I think some of the biggest problems
are and we'll, we'll work backwards.

Always lead with problems.

Some of the things in healthcare
that are just, in my mind, so broken

and in need of change really do
come down to what we started talking

about was data infrastructure, right?

Data fragmentation and
data infrastructure.

This, like, shows up in
a lot of different ways.

This shows up as technology not being
able to scale across organizations,

technology not being able to generalize.

And one of the things, I think,
that's becoming more and more

apparent in the US is kind of health
disparities across the country.

There are growing parts of
America that are losing doctors.

What are those people doing?

Like, that, that, that's going to cause
more chronic conditions and everything.

And so I think when we think about like,
where do we really want to be, like,

where's technology really, really going
to help us in the next 20, 30 years?

Personally, I'm not sure that it's going
to be in these super futuristic things of

like, doing a full body scan and finding
like the one cell in your body that is

potentially cancerous that I think is like
a lot of the sci-fi things we think about.

I think the promise is really,
how do we get equal care

to people that live in every single
setting in America across the whole

country when there's obvious resource
deficiencies and the supply and demand

is just like not the same everywhere?

So if someone can figure out how to fit
AI into that problem, like, that is what

I think will really, really drive change.

Really lofty goal, but it's probably the
thing I think about the most when I think

about AI and tech making a real impact.

So, to put words in your mouth, I
think you're reminding us that AI

is a tool, not the end, the outcome.

Yes, I think that's a great way to put it.

I mean, there are so many healthtech
and healthcare companies out there

right now that are using AI internally,
but they're not like AI companies.

And I think that is the right
way to be thinking about it.

You know, use AI to get done
the things that you need to

get done and do them better.

But sometimes when we have a hammer,
everything looks like a nail, and

we, you don't need to become an AI
company to really change healthcare.

Let me share with you some, I'm seeing
a little bit, I'd like to know if

you're having similar experience.

So, you know, in Newfire, we, we
do talk again, we talked to a lot

of clients about implementing AI.

I've, I've noticed a funny thing, it's
because you mentioned data foundations

and fundamentals that made me think
about this, was that as we talk about

AI, frequently the conversation starts
becoming like, oh what's the state of

your data platform, your data warehouse,
your data quality, and it's driving

more awareness of these systems that
have been languishing for a long time.

Are you seeing the same thing?

Yes.

I think that that's a great observation.

People are, to your point, like, have
kind of allowed their systems to deprecate

It just not be as advanced anymore.

And now they're like, Oh no, I can't, I
can't apply the cutting-edge technologies.

And that's okay.

I think that this is a great impetus
for us to all look and say, how

do we build a better foundation?

So we definitely are
seeing the same thing.

And that's why the problems that
excite me are actually really these

infrastructural problems and solutions
so that we can enable these things

that are much farther down the road.

So let's take some of the implementation
challenges off the table for a second.

Where do you wish you
could build an AI solution?

Well, what problem do you wish
you could address with AI?

That's a really good question.

I will, I'll go back to what I said
about like this really future vision.

I wish that we could figure out, and I
mean plenty of people are working on this

just to be clear, and I think plenty of
people like share this view, this is not

something I just thought of, but this
idea of like a person who lives in rural

Arkansas who has a stroke and cannot
get seen by a neuroradiologist in time.

Yet neuroradiologists exist in this
country who could read their scan

and advise, bringing expertise to
places where it currently does not

exist is where I would really want to
focus my time in like an ideal world.

And it's not just telehealth.

It's not just like connecting,
uh, a radiologist who's remote.

It's maybe it's not even connecting a
radiologist, maybe it's actually making

an AI that can automatically read that
image, which people are working on.

Maybe it's creating better surgical
techniques and tooling and suites that

do more robotic tech surgeries, right,
that don't need as much human expertise.

I don't know.

But I do think that the thing
that's most exciting to me is like

fixing this supply chain issue.

Thanks for sharing that because it
makes me, it allows me to sort of

categorize, um, you know, something
we, we see a lot at Newfire.

So we have the privilege to work
with a lot of healthcare startups.

And I realized what a lot of
them are trying to do is bring

expertise to the problem, right?

They're trying to solve for this expert
gap and they're trying to bring it to

the patient, which to me, intuitively
feels like the right thing to do.

I totally agree.

Yes, I've seen some amazing healthtech
startups out there that are, this

is exactly their model, is you know,
there's patients out there that need

a certain type of care and either the
place that they live or the insurance

coverage they have, or maybe both
prohibits them from getting those things.

And I mean, and even at Morgan Health,
we have done research and we've learned

that physical lack of access to care is
a key driver of poor outcomes and, you

know, misdiagnoses more so than not being
able to afford care, which is interesting

because sometimes we think of them as
being the same thing, but physical access

is different than financial access.

I have colleagues in Canada who are
in healthcare and they are trying

to tackle this problem is very,
yeah, a front-of-mind for them

because of course how disparate
Canada is in terms of geography.

Oh, that's really interesting.

Yeah.

I'm sure it's an even bigger
challenge for them there.

Particularly around things
like LGBTQ care, just having a

primary care physician, if you're
in northern British Columbia.

Yeah.

I mean, and you know where we
see this a lot actually is, is

mental health in the US too.

There are, there's kind of like a dearth
of local mental health providers who

will take insurance, and then when
you talk about specific populations,

like LGBTQ populations, it's like you
have an even harder time of finding

somebody near you in your network.

And so connecting expertise
across like geographies is huge.

Emily, similar to me, you've been
both a provider of technology,

also a consumer of technology.

So let me give you an opportunity
as a customer and a consumer of

technology, of solutions, what do you
want companies like Newfire to build?

How can we make a bigger
difference in this industry?

Scalable

Scalable data platforms.

I think something I have experienced, and
you alluded to this earlier, is so many

health tech companies are out there, and
this is, completely not a knock on them.

I I've lived it.

I understand why these decisions are made,
but so many health check companies are out

there building internally from scratch.

They, they're coming in and they're
saying, how do I build this, like,

bespoke something, data platform,
data warehouse for my needs.

And, as I said, there's many reasons
why people choose to build versus

buy, but it really, I think, is not
always a good long-term investment.

And I, I wish that there were an entity
out there, and maybe it's Newfire, who

could build more kind of scalable data
platforms so that these early healthtech

start, healthtech startups that have
a lot of promise don't actually need

to burn all that capital building what
they think they need to do so bespoke.

Um, I could, you know, we could do a
whole podcast on build versus buy data

warehouses, so I won't go further.

I, I, I think that's that, you
know, and I do think the, the

world is moving in that direction.

That's one of the benefits
of having gone remote.

Of course, there's a lot of costs
having gone remote, but there's an

opportunity to pull in expertise of
all types from around the world and,

you know, in different time zones, um,
you're starting to bear some fruit.

Yeah, yeah, definitely.

I mean, the world has changed quite a lot.

I will say I do so many people
are still, you know, building

the same thing in like silos.

Well, so Emily, do you have any other,
uh, any, uh, final words you want to

share as we wrap this conversation up?

30 minutes goes by really fast.

It goes by so fast.

I would say I, I hope everybody like
stays excited, stays hungry about

all of the amazing opportunities
there are right now for AI and just

data in general and healthcare.

I'm a, I'm a first
principles type of woman.

I think a lot about like,
"what is my base case?"

anytime I'm trying to
solve a complex problem.

And I think approaching things that
way when we're thinking about these

like lofty, complex technology
problems is really helpful and

grounding and sort of actually driving
more effective solutions early on.

And, often your base case is going
to come down to infrastructure,

which we talked about a lot.

So that's my, my parting wisdom is to
try to approach problems in that way.

Yeah, this is not a, uh, this
is not a 30-day mission, right?

This is a lifetime mission and
none of us can do it by ourselves.

Yes, definitely.

I, I think that state working in
healthcare data is, um, there will

always be work to be done for sure.

Well, Emily, thank you so much.

This has been an awesome conversation.

I really hope that our listeners
have gotten something out of this.

I think some of the topics
I've heard were AI is here.

It's making a big difference,
but you need to be careful

about what problems you tackle.

Don't be afraid of focusing on
efficiency as that may have the biggest

impact on your patient outcomes.

And finally, look for partners who have
the skill sets to help you be successful

like our clients at Newfire have.

Yes, that's great.

Thanks Gordon.

This was a great conversation.

Thanks Emily.

I hope you have a great day.

Cheers.

Thanks, Gordon.

You too.

Thanks for tuning in.

We hope today's conversation
with Emily Lindemayer, Executive

Director of Data Healthcare
Innovation at JPMorgan Chase &nCo.

has given you a fresh perspective on
how AI is transforming health, starting

with the foundational need to streamline
data and reduce administrative burdens.

Stay tuned for more episodes where
we continue to explore the toughest

challenges and smartest solutions
in business and technology.

Like and subscribe.

Until next time, keep innovating
and solving the hard problems.

This is Hard Problems,
Smart Solutions, The Newfire

Podcast.