April 2, 2025 • 55 mins
Can AI solve the healthcare crisis? Hear from two frontline healthcare executives — Dr. Sanaz Cordes and Dr. Eric Quinones, both Chief Digital Advisors with WWT — on the groundbreaking ways AI is reshaping clinical care, restoring physician focus, and unlocking new levels of operational efficiency. Explore how healthcare organizations are implementing AI today—with use cases including ambient scribing, inbox management, predictive analytics, and automation. Learn how CIOs and clinicians are scaling AI safely and effectively, and what it means for the future of healthcare delivery.
More about this week's guests:
Sanaz Cordes is a physician executive with 20 years of healthcare and healthtech experience. Serial entrepreneur, investor, venture advisor, and industry expert on healthcare innovation and clinical workflow. She is a speaker, blogger, and clinical thought leader that has built and helped scale healthcare companies from idea to acquisition. A pediatrician, Dr. Cordes began her career as a medical director for Providence Health and Services.
Dr. Cordes's top pick: Patient Experience Landscape in 2025
Eric Quinones is a disciplined physician leader driven by the Quintuple Aim; enhancing the patient experience, improving population health, reducing cost of care, improving the work-life of clinicians, and focusing on SDoH through the successful use of digital & data transformational tools, that support clinical and business objectives.
Dr. Quinones's top pick: Key Highlights from HIMSS 2025: A Transformative Year for Digital Health
The AI Proving Ground Podcast leverages the deep AI technical and business expertise from within World Wide Technology's one-of-a-kind AI Proving Ground, which provides unrivaled access to the world's leading AI technologies. This unique lab environment accelerates your ability to learn about, test, train and implement AI solutions.
Learn more about WWT's AI Proving Ground.
The AI Proving Ground is a composable lab environment that features the latest high-performance infrastructure and reference architectures from the world's leading AI companies, such as NVIDIA, Cisco, Dell, F5, AMD, Intel and others.
Developed within our Advanced Technology Center (ATC), this one-of-a-kind lab environment empowers IT teams to evaluate and test AI infrastructure, software and solutions for efficacy, scalability and flexibility — all under one roof. The AI Proving Ground provides visibility into data flows across the entire development pipeline, enabling more informed decision-making while safeguarding production environments.
Mark as Played
Transcript
Episode Transcript
Available transcripts are automatically generated. Complete accuracy is not guaranteed.
Brian Feldt (00:00):
Whether you're a doctor, nurse, caregiver or just
an individual navigating life,we are all part of the
healthcare system, and AI isn'tjust knocking on the door.
It's stepping into exam rooms,hospital corridors and
administrative offices,fundamentally reshaping how care
is delivered Ambient AIcapturing conversations between
patients and physicians, largelanguage models assisting in the
(00:22):
triage to chart summarization,and AI tools helping to reduce
stroke risk or helping eliminateopioid use in post-op pediatric
patients.
This isn't just hype.
This is actually happening andon today's episode, I'll be
talking with a pair of chiefhealthcare advisors from WWT Dr
Sanaz Cordes and Dr EricQuinones.
Between them, decades ofexperience delivering care,
(00:44):
driving toward clinician andpatient outcomes via innovative
technology implementation.
But innovation alone isn'tenough.
This is healthcare.
Trust, intention and humanoversight are essential.
So today, sanaz and Eric aren'tjust talking about what's
possible.
They'll be talking about what'sworking, what's scalable and
what's next.
This is the AI Proving Groundpodcast from Worldwide
(01:07):
Technology.
In today's episode, we'llexplore how AI is moving from
concept to clinical impact,reshaping workflows, empowering
clinicians and transforming thepatient experience across the
healthcare landscape.
Dr Cortes, dr Quinones, thankyou so much for joining us in
(01:32):
the AI Proving Ground podcasttoday.
Sanaz Cordes, MD (01:36):
Hope the two of you are doing well.
Brian Feldt (01:38):
Doing great, excellent.
Before we dive deep into someof the uncoverings that we found
out at the HIMSS conferenceearlier last month, I do want to
ask when it comes down to it,we are all part of the
healthcare system, whether it'sa clinician, whether it's a
healthcare patient.
I'm curious what have you seenover the last couple years in
(02:06):
terms of how AI has altered thestate of healthcare and where
it's going?
Eric, start with you just realquick.
What have you seen from AI inthe landscape?
Eric Quiñones, MD (02:11):
Wow.
So, going back a couple ofyears ago, I would say we've
come a long way.
Ai is not new.
It's been around since the 50s.
But I would say, with thepromises that we'd seen maybe a
decade plus ago with, you know,with Watson and things like that
, I think we fell short of whatthe the intended promise was.
(02:34):
I would say, even in the lastcouple of years, from what 2022
(02:56):
to maybe even last year, it wasa lot of theoretical or hope
what we would see.
But this, this last year, Iwould say we're seeing a lot of
true innovation when it comesdown to various AI technologies
being used in healthcare.
So real, practical, I would say, implications that are being
used today.
Brian Feldt (03:11):
Yes, and as anything to add there in terms
of what you've seen, or maybeeven build upon what Eric just
talked about and give us alittle bit of what we expect
over the next 6, 12, 18 months.
Sanaz Cordes, MD (03:21):
Yeah, absolutely.
And you know, to add on to whatEric was saying, and just even
the last year, I think, with thedemocratization of AI, right,
because now everyone's got chatGPT at home and they're using it
for you know these things andthey're seeing tangible outcomes
from it, and it's puttechnology in the hands of
people that have, you know,quite frankly, in the past been
(03:42):
technology averse, right?
Physicians have always had ahealthy suspicion of tech and AI
of all things.
Is this black box, right?
Like we can't see the code?
We don't know that.
You know, it's just very it's adifferent scenario than even
just regular software, wherethey have a consensus building
process of actually kind ofcreating the tool.
So I do credit chat GPT andthat whole generative AI tsunami
(04:08):
, as Eric likes to call it, aspart of kind of making it
tangible in the last year or twoand having so much excitement
around it as well.
The one thing I will say thatthis HIMSS was evident is maybe
because of that and becausethere are so many options and
shiny object syndrome, right,and CIOs who are used to the
(04:29):
lines of business pushing backon IT.
Now they're seeing thosedifferent departments coming in
with their hey, can we use itfor this.
Hey, can we get that?
That there's now a much moreintentional approach to it.
Where I'm hearing fromleadership at HIMSS the
different organizations sayingwe're interested in AI but know
approach to it.
Where I'm hearing fromleadership at HIMSS, you know
the different organizationssaying we're interested in AI
but we're going to slow it downand make sure we're very
(04:51):
intentional about what tools weimplement, how we implement them
, how we measure, you know,metrics of success from them and
, quite frankly, the cost right,like wanting to kind of let the
noise settle and see who risesto the top, which startups or
others might, you know, go offinto the, into the background,
so that they can really, youknow, have a more intentional
(05:12):
plan for it.
Brian Feldt (05:13):
Yeah, Eric.
What did you see out at HIMSSthat led you to believe that we
are now actually moving fromthat experimentation phase to
the absolutely essential phasewhen it comes to AI in the
health care setting phase?
Eric Quiñones, MD (05:25):
to the absolutely essential phase when
it comes to AI in the healthcaresetting.
Yeah, no, I think there's somereally interesting things.
So, you know, I think there's.
When I think of the clinicalworkflow, where AI is really
hitting the there's, you knowthe rubber's hitting the road
would be when it comes down toambient.
You know technologies right.
So being able to use thosekinds of technologies to record
(05:47):
the conversation that thephysician and the patient are
having in context and putting itin a clinical note, and before
that patient even leaves theroom, that note is ready to be
reviewed by the clinician.
So there's always a person inthe loop, if you will.
So the clinician will bereviewing that note.
Their orders may be teed up andready to go as well.
So it's really impressive to beable to see that and where
(06:10):
they're seeing about a 70%reduction when it comes down to
documentation time.
That's significant.
And then you know other areasthat I think we're seeing it
really being impactful is whenit comes down to, you know,
handling inbox.
You know inundation that we getfrom various clinicians,
(06:31):
patients et cetera, in ourinboxes to be able to use.
You know, large language modelsto help in responding to those
things.
And again, there's always goingto be a.
The physician is still going tobe in the loop and going to be
reviewing that.
But you know the patients havebeen reacting very well to some
of those responses becausethey're more detailed, uh, the
more empathetic and not thatwe're not detailed and
(06:51):
empathetic people, we just don'thave the time sometimes to
really do that.
Um, and then, lastly, I wouldsay chart review.
We spend about a third of ourtime in reviewing charts and
patients come in time andreviewing charts.
When patients come in, what arethey coming in for?
I forgot what's this follow-upbusiness about, and so it will
pull up all that disparate dataand put it into a format where
(07:14):
it's curated and we can seeexactly what they're coming in
for, what may have been missed,et cetera.
So there's these real great usecases when it comes down to
those kinds of things.
So there's these real great usecases when it comes down to
those kinds of things.
I think, on the other hand, too, we're seeing it being applied
when it comes down to justautomation.
Right, it doesn't have to be socomplex, but when it comes down
(07:37):
to automating certain processes, where we're seeing more and
more clinicians being able towork at the top of their license
and when I say clinicians,that's just not physicians,
that's nurses as well and otherfolks on the care team.
Sanaz Cordes, MD (07:51):
Yeah, and Eric , just to circle back to your
first point about the ambient AI, Dr Poon anda few others, dr
Poon from Duke, I think you werethere, a bridge hosted one of
the sessions and something hesaid really kind of resonated
with me.
Where, you know, we introduce alot of technology that can
(08:11):
improve workflows and, to someextent, obviously, the
experience for clinicians.
But this was the first time Iheard somebody put it this way,
where it's like it c ompletelychanged the way doctors just
like the psychology and the waythey're just approaching patient
care.
What I mean by that is when wewere in this whole, you know,
(08:32):
two decades zone of like we'retrying to talk to a patient but
we're also, you know, typingeverything down.
We start thinking astranscriptionists and A it
distracts you from theconversation.
You're not as present as youneed to be transcriptionist and
a.
It distracts you from theconversation.
You're not as present as youneed to be Um and B.
The way you're, you know,interacting um can your goal
directed for the transcriptionright and not necessarily the
(08:54):
way you should be thinking asphysicians.
So giving that back and lettingdoctors just kind of go, I can
like focus on this patient andreally put my doctor brain the
way I was trained at med schoolor residency to work.
You know, that's just to me,you know that's an amazing thing
that technology can, you know,bring back to folks.
And then I think there wasanother physician, dr Mishuras,
(09:18):
just checking it, yeah, fromMass General, and she, I think,
was a part of that same panel.
And when she talked about, youknow, automating the clinical
workflows that Eric wasmentioning, I mean it's just
kind of like a triple win, right, like it's really improving
these clinical outcomes tolevels will readily admit as a
physician, sure, you know, I canprobably come up with a
(09:48):
differential diagnosis and usesome apps and such, but, you
know, humble enough to know thatthe AI could actually, you know
, find a new solution, biggerand better.
So you've got that, you've gotthe cost savings, you've got the
efficiency and, of course, thepatient experience.
So, overall, I mean, I thinkit's just it's.
You know, I have a lot moreexcitement this year because I
think we're seeing actual,tangible results and it's a
(10:09):
great time to be coming backfrom HIMSS riding on that high.
Brian Feldt (10:21):
Yeah, love to hear it.
I'm going to put a pin in that.
Operational workflows here forjust a second Synaz.
I do want to ask you you'retalking a lot about efficiencies
gained through things liketranscription or other use cases
like that that seem prettypractical.
Anything at the HIMSSconference that would speak more
towards the innovation end ofthat spectrum that caught your
eye or that you think will makea real impact either this year
(10:43):
or in the near future?
Sanaz Cordes, MD (10:44):
caught your eye or that you think will make
a real impact either this yearor in the near future.
Yeah, absolutely.
Himss really started off withkind of a bang.
This year we're all kind ofsettling in and getting to our
seats at the keynote and for me,having gone to HIMSS for over I
don't know 15, 16 years now,that opening session was amazing
.
We had the CEO and the CMIOfrom Samsung Medical Center in
(11:06):
Korea and they've won thehighest level of certification
for HIMSS now of any otherorganization, and so they
presented everything thatthey've done at.
It's like the HIMSS indicatorscore.
But they presented everythingthey've done at this hospital
and it's just amazing.
(11:27):
They've, first of all, lessonslearned.
So they created amulti-stakeholder you know,
multidisciplinary approach tohow they're going to roll out
this real connected health model, the smart hospital model, and
so, for example, physicians hada seat at the table saying you
know, when the EHR does X, y andZ, it's disjointed from that.
(11:48):
We would really love it if itwas this way and other folks are
weighing in and they actuallybuilt their own EHR this way and
it's kind of an AI first, ifyou will, kind of EHR model, and
it integrates beyond just thatright, like it's leveraging
large language models to combthrough data and help with
decision making, but it'sautomating tasks downstream.
(12:11):
It's doing like predictive sortof forecasting of you know,
between two and six is when wehave the most MRI orders or
portable, you know, chest strikeorders.
So maybe we forecast how that'sgoing to.
We're going to tee up resourceallocation and optimization,
like in the future.
And then, yeah, I mean it wasjust mind blowing because it's
(12:33):
like if we had a blank slate andcould rewrite healthcare to
make it better, you know thanwhat happened, you know, 20, 15,
20 years ago when everyone madean EHR and now we're dealing
with all these siloed processes.
So, and then there were therobots.
Eric Quiñones, MD (12:51):
I don't know if, eric you want to talk about
the robots those were.
She brought robots.
That was very innovative and sowhat it really helped do, I
would say, you know, especiallyfor the pediatric population
right, that are in the hospitalAgain, hospitals can be very,
you know, just daunting places,and even for adults, so you can
imagine the children.
So these robots are veryinteractive and they were able
(13:14):
to, you know, with the pediatricpopulation and be able to, you
know, do things with them and,you know, I would say, somewhat
distract them from all the otherthings that happen in a
hospital.
But it really helped.
Calm them down is really theend point.
So I thought that was a reallyunique way, or innovative way,
to bring in some, you know,artificial intelligence into the
(13:38):
healthcare space.
Brian Feldt (13:39):
So, Naz, I am curious what took place at HIMSS
in terms of the regulatoryenvironment.
I'm curious what took place atHIMSS in terms of the regulatory
environment.
Lots to unpack in that questionand lots of news and
information could happen overthe next coming weeks or months,
but what were physicians orclinicians talking about when it
comes to the regulatoryenvironment and how that might
impact AI adoption?
Sanaz Cordes, MD (14:00):
Yeah, I mean lots of changes coming for sure
every hour for the nextunforeseeable future, right, I
mean, with political thing, youknow, unprecedented changes that
are being handed down andthere's already a lot of impact
from this.
I think almost everyconversation I had that factored
in.
I'll just give an example of,you know, academic research
organizations.
(14:20):
I worked I spoke with one ofour customers that we're working
with and already they arerestructuring and that starts
from the people, the process andthe technology, right, Because
with the funding change nowthey're having to shuffle data
science dollars here and there.
They're having to figure outwhat tech they can invest in and
what's going to serve morepurposes, so that they're having
(14:42):
to be more strict about theirprioritization.
So it's already there and Ithink there's just so much
unknown and that's kind ofparalyzing some decision making
as well.
So it's already definitelyhaving an impact for sure on the
academic research organizations.
Eric Quiñones, MD (15:01):
Yeah, I'll add really quick.
So just I think the last week,the end of last week, so Johns
Hopkins, they actually announcedthat they're going to be laying
off to 2000 employees,including a significant amount
of the researchers, because thiswas due to what?
An $800 million grant that wascut from what?
(15:22):
The US Agency of InternationalDevelopment.
So you know you're going to beseeing more of that.
And then also I read thatbecause their residency programs
sometimes are federally funded,you're going to be seeing these
residency programs actuallybeing cut as well, in a time
where we need more residencyprograms.
(15:44):
We just can't produce enoughdoctors.
As a matter of fact, the datashows we'll never have enough
doctors ever again.
So you know you could have.
We have a surplus of medicalstudents, graduates, but we
don't have enough residencyprograms to train them.
So you know these kinds ofthings are going to impact
healthcare very significantly.
Brian Feldt (16:06):
I would assume Eric to just build on that.
That's an opportunity as wellfor AI to come in and
potentially supplement, or isthat not necessarily the line of
thinking here?
Eric Quiñones, MD (16:18):
I think, maybe in some ways, but I think
you can't replace some of theseresearchers.
I mean, it's so their work thatthey do is so, so critical.
As a matter of fact, you know,I would say that even in some of
that research you know, thereare, you know, maybe engineers
that are AI data scientists aswell.
So, yeah, they're just notthey're being, you know, I would
(16:42):
say, eliminated in some waysbecause of not because of the
it's more of the politics that'seliminating them and the grant
reductions that are that areeliminating them.
But, yeah, it's going to be avery interesting period, but I
have faith that we're going tofind a way.
We're very good at doing thatas humans, and I think the
(17:07):
research is too important tojust not do it, and so we're
going to figure out ways to doit.
It's going to be very novel,but, yeah, I don't think it'll
replace the scientists in theroom the scientists in the room.
Brian Feldt (17:26):
Yeah, before we move on from HIMSS in general,
eric, any other kind of keytakeaways?
I know Sanaz mentioned a coupleof the keynotes, anything that
you just thought would beworthwhile mentioning or things
that caught your attention.
Eric Quiñones, MD (17:37):
Yeah, I think , going back to real life use
cases, I would say where we'reseeing AI being.
You know the benefits thatwe're seeing today.
One came from Dr Chaudhry.
He is the chief AI officer atSeattle Children's and there was
(17:58):
two use cases that again, hehas the right people process
data technology in place and hereally stresses that it's like
you have to have those things toachieve your goals.
And one goal was they wanted tobe able to reduce opioid
(18:19):
treatment in terms of for thepediatric population to reduce
opioids by 100%, and they use,obviously, opioids for pain and
sedation, but they can have avery addictive quality to them.
So what they did was they usedthese various AI models to help
identify different medicationcombinations to help achieve the
(18:42):
goals that they were trying toachieve in terms of sedation and
pain management, and they wereable to do so, again by 100%.
So that was significant.
And then the other use casethat Dr Chaudhry actually talked
about was reducing pediatricstrokes in post-operative
patients that have had brainsurgery.
So when they go to the ICU,it's not uncommon for a child,
(19:08):
to you know, to have a strokeand no matter what we do, we do
the best we can, but it canhappen, and you can imagine the
outcomes are not very good whenthat happens.
So they were able to applyvarious models to identify which
particular patients are at riskand take those precautions, and
(19:28):
so they were able to reducethose strokes by 50%.
So those are significant usecases.
I think that we weren't able todo before the technology
existed.
Brian Feldt (19:38):
Yeah, I do want to add.
I want to go back to thatoperational workflow idea that
you brought up, sanaz.
I'm wondering as we integratemore AI into the workflows and
the clinician setting, how isthat shifting, how we work
within a hospital, within aphysician's office or anywhere
(19:58):
within the healthcare setting?
What are you seeing as kind ofthose big operational shifts as
we move closer towards that AIadoption?
Sanaz Cordes, MD (20:06):
Yeah, I mean, I think in many ways it's an
operational shift and just abehavioral kind of again use the
word psychological shift,because if you, you know, if you
really think about it, the EHRand a lot of the technologies in
the hospital, the EHR wasn'tinvented to make work better for
anybody, right?
It was a billing and claimsrepository, right, and all of a
(20:27):
sudden, overnight, it createdtwo hours of clicking.
And you guys have all heard mesay this over and over in the
last decade and I really feelthat just in the last couple of
years.
It's like doctors are taking itback, right, clinicians are
taking it back.
It's like, I mean, we have ashortage, right, and the numbers
are all over the place, but forsure, over $60,000, 60,000
physicians shortage, you know,within the next year or two, and
(20:50):
it's not getting any better.
And so I think people finallyare sitting up and sitting up
straight and listening to this.
So I do think it's just thisshift of like we're taking back
medicine, you know, and so byintroducing these automations,
we're just trying to almost,like, go back in time, right, we
want to be able to talk to ourpatients and focus on what's
(21:10):
happening, not be distractedenough to be able to take in the
psychosocial things that areimportant, their social
determinants of health, theirhistory, and not be spending all
this time.
I mean, the rate of imaging thatis redone in ERs is just
staggering.
Right, it's gotten better, butI remember in my days and not so
(21:33):
long ago for others wherepeople come in with a known
underlying problem they had anMRI or whatever done in the same
city maybe, and it's 11pm atnight and you're not going to
get that.
So what do you do?
You repeat the MRI, right?
So I think, by automatingthings and letting AI and other
tools and interoperability workfor you to do that sort of
(21:54):
administrative noise and let youwork top of license, you know,
I really feel like cliniciansare taking healthcare back and
again win-win for the healthsystem, because, because MRIs
cost a lot, you got to pay forall these utilizations and then
that comes back.
That gives back to us as asociety and caring for
increasing aging population, forexample.
(22:15):
Eric talks about that a lot andthat's not going to get any
better, right?
So, yeah, I think it's justfundamentally going to change
how we do everything.
And then there's the wholeoperational piece, right.
I mean claims and billing, andall of that has been a nightmare
from day one, not just for theclinicians and nurses, but for
everybody and for thatdepartment.
And there's so much more theycould be doing right.
(22:38):
Think of what a disasterauthorizations are for things
like referrals and specialtystudies and just price
transparency, pathways andfinancial experience for
patients.
So if we could free those folksup to be able to work and move
the needle on some of that toyou know, long-windedly answer
your question.
I think it's just fundamentallygoing to shift how we do
(22:59):
healthcare and we, quite frankly, we need it because we're at a
crisis point, in my opinion, ifwe don't change things around.
Brian Feldt (23:06):
Agreed.
Yeah, not to play devil'sadvocate here, sinez, but what
about the concept of an AIburnout loop here?
If we are able to be moreefficient and come to
conclusions faster, does thepotential exist that clinicians
are just going through more andmore patients Like?
(23:28):
Is there a balance that needsto be struck in terms of giving
the patients and their familiesor caregivers the time that they
would need or deserve, asopposed to just relying on AI to
get them through the door?
Sanaz Cordes, MD (23:40):
Yeah, I mean, of course there's always risks
with any kind of automation, Iwould say.
My answer is I'm curious tohear Eric's is we are in such a
shortage right now, especiallyof physicians, right where we
are trying to increase other,not mid-level, but the extension
(24:00):
practitioners, pas, nursepractitioners.
We're trying to increase thatpopulation.
We can't do that fast enougheither.
So I think that we are far farfrom having a problem of surplus
AI and freeing up so much timethat we're going to create this
new swing of the pendulum.
I think we're going to finallylet people stop running 100
(24:23):
miles an hour and let them goback to a normal approach, you
know, approach to a workday,which again, can only benefit
everyone.
So I mean, like I said, there'salways a risk.
I don't see that happeningbecause we're just so far behind
and there's so loss of joy inpractice that I think we're
going to give that back in thenear term and maybe even mid to
long term future, before wecross over any other threshold.
Eric Quiñones, MD (24:47):
But, eric, if you want to add to that, yeah,
no, I think I like throwingthese stats out and they vary a
little bit depending on whatyour resource or your resources.
So in the next five years, by2030, the numbers I've seen is
we're looking at a deficit ofabout 140 at the most physicians
(25:08):
across the US, about 200,000nurses across the US in general
as well, and we've been runninga deficit in nursing for a while
now across an average in the US.
In the next five years, everybaby boomer will be 65 and over.
That's 20% of our population,so you're going to have an older
(25:28):
population.
That means a sicker population.
In addition, in 1950, medicalknowledge doubled every 50 years
.
Today it doubles every 74 days.
I cognitively can't keep upwith that.
I would have to read 20 hours aday just to keep top of license
, okay, so things are changingreally fast.
(25:49):
And then the data that we have.
We think an average hospitalproduces about 50 petabytes of
data a year, but now that datais going to be going up even
higher because we're going to beseeing other data sources, such
as genomics, right, and theother biomes as well.
We're going to be seeing otherdata sources, such as genomics,
right and the other biomes aswell.
We're going to be seeing otherdata sources as far as wearable
(26:09):
technologies, right, so thosedata sources will be coming in.
So there's more data to process.
We just can't do it alone, so weneed these technologies to help
us curate and, really, what isthe signal in the noise?
Help us to be proactive andpredictive, to find those
(26:32):
patients that really need ourattention well before a problem
starts.
So, yeah, I think we're goingto be seeing a lot of change
when it comes down to the way wepractice medicine, and I'll
even add that I know new youknow residents that are new
graduates in the residencyprograms.
(26:53):
As they're coming out, whenthey're interviewing in
hospitals, they're asking notabout just their salaries and
their other benefits.
They're asking me what kind oftechnologies do you bring to the
table that actually will helpmake my life different or better
?
So that was never a thing forme.
I know that it was like go towork here and, matter of fact,
(27:17):
it was paper which is not a badthing, but, we.
You know we're not there anymore, but no, those are things that
are happening today and, I think, where we're going to be really
seeing these technologiesreally help and change the way
we're practicing health care.
(27:37):
I'm very hopeful and excitedfor the future, but right now
we're still dealing with a lotof these issues that Dr Cortes
has actually mentioned.
Sanaz Cordes, MD (27:48):
You know, and there's also the patient piece,
right, Like we're talking aboutoperational workflows as it
benefits hospitals andclinicians, but then there's the
patient piece, right, so whatpatients or potential, you know,
patients, healthcare consumersare willing to accept anymore
has definitely changed, likethat bar has raised.
You know patients, healthcareconsumers are willing to accept
anymore has definitely changed,like that bar has raised.
You know, and I liken it to youknow, let's say you're trying
(28:09):
to pick a restaurant to go toand you're between meetings and
one has online booking forreservations and one doesn't.
Which one are you going to?
I mean, I'm obviously alwaysgoing to default to the one that
has.
So patients as well.
Right, they have choices now,and it's not just this mediocre
tech health system versus youknow this slightly less mediocre
(28:30):
, it's my health system.
Or I'm going to pay $29 orwhatever and get instant access
to a physician and aprescription, and you know,
online, or I'm going to walkinto one medical and be seen.
So there's a lot of more.
You know, more options.
And so by implementing thesetechnologies that are often, you
know, dual facing right, so itallows them to be able to do
(28:51):
dynamic appointment schedulingor medication management or, to
Eric's point, as we get moredata and we need to be more
proactive, push more proactivereminders, you know, to the
patients and just a myriad ofother things that are beyond the
table stakes of these types ofthings I'm listing right, like
they want a better overallexperience, access to care 24-7,
have them meet them where theyare right.
(29:13):
Like there are some things thatI'm willing to drive across town
and wait an hour in a waitingroom for, but there are other
things that I'm frustrated thatI can't just dial up, and maybe
even an asynchronous visit whereI don't need to have the doctor
present.
This should be able to be astore and forward type of
engagement where somebody candeliver it.
So I do think that there's thatpatient component and what
people are now, as consumers,willing to do and they'll vote
(29:35):
with their feet if they're notgoing to get it at that hospital
and go somewhere else.
Brian Feldt (29:50):
I think that patient experience is an
interesting topic to dig into alittle bit.
As ambient technologies startto enter the clinician setting
and you know there's kind of analways-on listening aspect to
when you're in the doctor'soffice, does there need to be a
transparency conversation withpatients?
And then, I guess you know,moving that conversation forward
(30:11):
is, where does AI go from herein terms of, like, actual
diagnosis?
Eric Quiñones, MD (30:27):
And how do?
Sanaz Cordes, MD (30:27):
health organizations balance that with
what they tell the patient inthe waiting room Right.
Go ahead, eric, go ahead.
I was just going to share, justto put it in kind of a
perspective.
You know, about 50% of patientsview, you know, ai as a
positive thing, right, andthat's up a lot from the past.
I think it's like 53 percent.
So we're definitely kind ofmoving that needle anyway,
(30:48):
because they again are seeingsome of the benefits of that.
So I just kind of wanted tomake sure we touch on that,
because there's there is demandand acceptance of this as a as a
means to giving them theexperience that they want.
But yeah, eric, if you means togiving them the experience that
they want.
But but yeah, eric, if youwanted to go in more into what.
Eric Quiñones, MD (31:07):
Yeah, no, no, that's I think.
I think we're seeing very, avery positive reaction from
patients.
So what?
The one example I mentionedearlier was inbox management,
right?
So let's say you know you ask aquestion of your physician and
you're waiting for a responseand something that's's before AI
, there was no AI and they'rejust you're waiting and waiting,
(31:28):
and waiting.
You finally get a response.
It could be a very curtresponse and you know to the
point, and not that thephysician was trying to be rude,
it was just the time is notavailable to really take time to
write a very lengthy, detailedresponse.
Available to really take timeto write a very lengthy,
detailed response.
Fast forward where patients aregetting those responses from an
(31:49):
AI assistant and it'll say inthe message this is an AI
assistant for Dr So-and-so andthey would get a very detailed
message.
They would get a very heartfelt, empathetic message and the
responses that they gave whenasked, when they were very, very
(32:10):
happy with what they werereceiving.
So I think you know there's alot of excitement about that,
but I also think that therealways has to be a person in the
loop, right?
This can't just be.
You know this is sent outautomatically when these
messages.
When they were sent out, theywere still reviewed by the
clinician and they're being sentout, but I think that's
(32:32):
important too to know.
But no, it's exciting.
Sanaz Cordes, MD (32:37):
But you bring.
I mean, brian, but you didbring up a good point because,
like to date, I mean, I'm notaware when my doctor's using
clinical decision support andusing one of maybe Epic's AI
algorithms to do something, andto date I think like that's
okay-ish right.
But you bring up a good pointof at what point isn't it okay
(32:59):
without consent right, okaywithout consent right.
At what point, like even withAmbient, as a patient, I am
really happy that I'm beingrecorded because, being on both
sides of that equation, one ofthe biggest frustrations is when
you go and you visit yourphysician and then you know
you're supposed to get thesefollow-up things and then, as
you're getting them, you'rerealizing wait, I already told
(33:21):
him that I have X, y and Z, sowhy isn't there the prescription
there?
So, as a physician, now in thatequation, you know you've got
all that content and so whenyou're getting ready to do your,
you know we call it theassessment and plan, coming up
with your you know treatmentplan.
Like you have all that data,you're not going to miss those
things.
So, again, to date, like it'spositive, positive, and you know
(33:43):
we don't necessarily thinkabout that.
But you know we've got 800 FDAapproved.
You know AI algorithms today.
What do we do, to your point,when it's like 8,000, right,
there has to be some kind ofprocess, and I don't.
I don't have the answer forthat, but it's really a good
point and something worthwatching.
Brian Feldt (34:05):
Yeah, I like the idea of that, the human in the
loop, that aspect, although Ithink it also just begs the
question.
I think a lot of people havethis type of question, which is
how effective is AI in terms ofdiagnosis?
We recognize the need to havethat human in the loop, but at
what point in the future does AIor is it already becoming on
par better than human diagnosis,or is it already becoming on
(34:27):
par better than human diagnosis?
Eric Quiñones, MD (34:30):
I would say it's getting there.
As a matter of fact, I wouldsay it's maybe a year ago it was
at the level of a first-yearresident, but now what I'm
seeing is that it's surpassingthat and again, what matters is
the data.
So we talk about, you know,interoperability is a very
important part of this topic.
(34:50):
If we don't have the data, it'shard to make those, you know,
to run those algorithmsappropriately to get the right
outcome.
But I've seen situations wherethey've had some really rare I
would call it, you know, kind ofreally rare internal medicine
type problems.
(35:10):
I mean, these are real cerebralissues that a patient has.
I mean, they're medicalmysteries and they've applied,
you know, models to.
You know identify what theproblem is.
And differential diagnosis thefirst diagnosis and differential
diagnosis the first diagnosison the differential diagnosis
was the correct diagnosis, soit's getting really good.
(35:30):
Now that doesn't mean that wejust say, hey, let's go with the
first diagnosis and run withthat.
We still have to look at that.
We still have to, you know, askthe question well, what is this
based on?
You know, and I think when youhave that transparency that's
within the model and that maybeyou know, dr Cordes or myself or
any clinician can actuallyclick on and see okay, this is
(35:52):
based on this information, this,this is the you know, the
evidence-based medicine it maybe using, et cetera.
Then it brings more credibilityto what we're seeing.
Nothing will take away the youknow, the clinical mind.
We still have to apply that.
So, but to get to the mysteryquicker, to find out what's
(36:15):
causing it quicker, I think thathelps a tremendous,
tremendously.
You know we can spend a lot oftime trying to figure out a
problem in medicine and I knowI've seen, you know I've seen
patients go through thisparticular process where they've
been misdiagnosed, misdiagnosed, misdiagnosed by multiple,
multiple physicians.
And, as a matter of fact, Ihave a good friend of mine,
(36:37):
she's a nurse, a colleague.
That happened to her and then Itook her information, I put it
in chat, gpt, just to experiment, and it actually kicked out the
differential diagnosis and itwas the first one that they
could not identify.
So it's getting pretty good.
Sanaz Cordes, MD (37:00):
Wow.
I didn't know that story.
Eric Quiñones, MD (37:02):
That's pretty cool.
Sanaz Cordes, MD (37:04):
You know, I think we can use imaging, you
know, as an example, becauseimaging was the first right that
embraced ai and radiologistsembracing ai and that has come
such a long way, I mean, even atworldwide.
You know we've had um, we'vehad our data scientists just you
know, on the bench who wantedto work to solve a problem, be
able to, you know, do a braintumor radiomics.
(37:27):
You know mri project where itcan diagnose the.
You know the end brain tumordiagnosis like 20x faster, and
that's just.
You know a diagnose the.
You know the end brain tumordiagnosis like 20x faster, and
that's just.
You know.
A couple of data scientists,you know, playing on our models.
There's solutions out there thatare just they've been out there
for years and so imaging, Ithink, is a great example of.
You know it started out okayand then people were meticulous
(37:47):
about it and now it's reallyubiquitous.
I mean, most health systems areusing it.
I do think that there's alsothe flip side.
It's not all rosy.
The EHR I mean, let's just callit what it is Epic has a habit
of when these new third-partytechnologies become available
now, most of them AI they havethis habit of telling their
(38:10):
customers oh, we're buildingthat or it's coming in a release
next year.
Oh, don't buy this because andthe health systems also don't
want to be piecemealing a bunchof disparate third-party apps.
So I get it, but they're alsostarting to get a little
impatient and a littlefrustrated with that answer
coming from Epic.
So they do use obviously asmuch of Epic as they can, but
(38:32):
I'm picking on Epic because Iremember it wasn't that long ago
where their Epic sepsis modelwas really faulty and you, that
is not an area you want to be,you know messing with, and so we
can't just kind of blindlytrust it and say that it's
getting as good, um, or better,globally I think it's.
You know very specific um usecases that still require a lot
of.
(38:52):
You know the models gettingmore accurate and smarter and us
holding on to that.
You know cautious skepticismuntil you know we get to maybe
the level, like, of imagingthat's been there now for a
while.
Brian Feldt (39:14):
Well, that's a good segue into the IT conversation,
which is also a very importantcomponent here, recognizing that
AI is moving so incredibly fastand a CIO or an IT team for one
of these healthcareorganizations just have so much
coming at them, whether it's anew vendor or thinking about
(39:34):
their data estates or whateverit might be.
What are some of theirpriorities that they must be
thinking about now so that theycan actually implement and have
a workforce that adopts andscales AI on their behalf?
Sanaz Cordes, MD (39:48):
Yeah, I mean, I'm happy to start with that.
I would say, just in the lastsix months, that's what a lot of
my meetings are about, as youknow, as I meet with some of our
executives at theseorganizations and wherever you
are, however big you are, youknow you could be Northwestern.
Who's on the front end?
I mean, they're doing amazingthings and they've got a whole
system and a workshop.
Or you might be a smallerhealth system.
(40:08):
At some point you're workingtowards an AI center of
excellence health system.
At some point you're workingtowards an AI center of
excellence, and what I mean bythat is, you know, maybe it is
literally a center where you'vegot data scientists and folks
and you know, governancecommittees, or maybe it's just
the CIO, who's got now a hundreddifferent use cases being
demanded of him, and IT teams or, yeah, it folks that are maybe
(40:30):
not skilled in implementing this.
I mean, that's a big factorright now is just getting talent
that can be skilled for AI andfor the cybersecurity component
around AI.
But wherever you are, you'retrying to work to an excellence
model where there's some sort ofstandardization of ingesting
these ideas, standardization ofprioritizing and mapping them to
(40:52):
feasibility versus cost, versusoutcome and then a methodology
of how they're built right.
Does it come off a third party?
And, if so, what are thecriteria for that?
And then Eric brought up themost important point earlier,
which is the data.
Dirty data in, dirty data out.
We've been saying that foryears about software and now
more than ever with AI, we can'temphasize how important that is
(41:14):
.
So then, what's yourmethodology around data?
And then that leads you to thewhole.
Like this is really kind of aplatform play.
These things can't happen insilos, right?
Because if we can rinse andrepeat to some degree all the
effort we've put in to createthis model off of this LLM, we
need to be able to repurposethat.
(41:36):
So it's definitely acoordinated, multidisciplinary,
multi-stakeholder process andthat's where we need to get.
And so our goal mine, when Imeet with customers is don't be
overwhelmed by all that.
Let's just start with one usecase and figure it out.
But each time we're gettingsmarter, just like the AI, so
that as we move forward, we cango from one to five to 10 to 100
(41:57):
without having to, in linearform, increase the effort.
You know, one to 10 to 100.
Brian Feldt (42:02):
Yeah, eric, I do want to ask you about data.
You've mentioned it a number oftimes already and, sanaz, you
just mentioned it as well.
Understanding that the dataestate of these health care
organizations is such a complexarea Some may be sophisticated,
some may be not as much what arewe advising organizations in
terms of how to best leveragetheir data?
(42:24):
Is it start small and work yourway up with momentum, or is it
cleaning everything at once?
How are we talking to clientsabout how they should approach
and leverage their data?
Eric Quiñones, MD (42:35):
Right.
Right, I think one of the firstthings I you know well, this
happens a lot when I'm visitingwith clients and we're having
these discussions it's do theyeven know where their data's at?
You know it's their datagovernance right, starting there
.
And, quite frankly, a lot ofthem can't really answer that
question, sometimes very clearlyor succinctly.
(42:56):
Or, what condition is that dataright?
Is it, has it been?
You know, is it discrete data?
You know, is it just text data?
So you know, again, these arethe conversations I have all the
time and it's like, so that'swhere you kind of have to start
Right.
And then do you need all thedata have and your AI governance
(43:18):
?
You know folks that are in theroom helping to define those
particular use cases, but itreally no matter all of that if
(43:42):
you don't have the data.
You know some of those coredata components you can't do the
.
You know the work that you wantto do.
So really starting there is soimportant and really identifying
where that data is going to be.
So you know, maybe right nowit's in a place where, like I
(44:02):
said, it's disparate, butthey're considering, you know,
moving the data into anenvironment that will give them
more accessibility to the dataand who has accessibility to the
data right?
You know they have their ownresearch teams that they need to
have accessibility, versusothers that don't need to have
that accessibility.
So there's a lot of questionsthat need to be asked and a lot
(44:25):
of you know that come out of alot of these meetings.
And it's good because you knowwe do identify, you know what
next steps need to happen.
And so, because they do have agoal of doing a lot of these
things and they know they haveto do it because they cannot
continue to practice medicinethe way that they're doing it
(44:48):
today they know they need to usethese tools, but they can't use
these tools unless they havethat data governance in place.
So I would even add that theyneed to use these tools, but
they can't use these toolsunless they have that data
governance in place.
Sanaz Cordes, MD (44:55):
So I would even add that they can't use
those tools until they have theinfrastructure modernized in
place, right, so that's.
Another thing is, like you know, when we go in there and we I
mean health care, nonprofithealth care is not a profitable
right, like, obviously, industrythere, but they're very budget
restrained and so a lot of theseorganizations are operating on
(45:17):
legacy tech, end of life tech, alot of stuff's on premises, and
so, again, that's why we kindof say let's just start with one
use case, so that we don'toverwhelm and kind of throw it
all, maybe go with the bathwater, because with that one we'll
figure out what's the mostimportant.
Is it your cloud migrationstatus?
(45:38):
Do we need to focus on a hybridcloud model?
Or is the barrier to success onthat one the fact that you have
a storage challenge?
But there is inevitably goingto be an infrastructure
modernization component thatgoes hand-in-hand with that data
strategy that Eric'sreferencing.
Brian Feldt (45:57):
Yeah, and Eric, you mentioned data governance as
well, and I think I just readmaybe it was last month an ECRI
report about the top 10 threatsright now in healthcare data
governance rated number two.
And that also actually bringsup another issue of just
cybersecurity in general, whichI understand was another
important topic at HIMSS when dowe stand in terms of AI, cyber
(46:21):
data protection within thehealthcare industry?
What are some best practices orwhat are some questions that
leaders need to be askingthemselves?
Eric Quiñones, MD (46:29):
Yeah.
So, Naz, you want to.
You want to take that Cause Iknow that you were very heavily
engaged in some of thoseconversations.
Sanaz Cordes, MD (46:36):
Yeah, yeah, I had a great session on that
actually with Paolo.
We hosted several folks duringHIMSS.
Yeah, I mean, I think AIabsolutely has brought
cybersecurity front and center,not that it hasn't been.
I mean, we had, I think, like720 breaches last year and sure
(46:57):
that's down a couple percentfrom the year before, but we
have to always think about thefact that we're like 2x where we
were before 2018.
So I mean, it's a heat andeveryone knows about the change
healthcare thing, and so nowwe've got AI and there's AI's
role in cyber, got AI and youknow there's AI's role in cyber.
There's AI for cyber, and thenthere's cyber to protect against
(47:18):
, you know, bad actors from AI,and so one of the things with
why we're at such risk inhealthcare is just our
environment.
There's a reason why you knowwe're in the top two costliest
industries for cyber risks.
Right, we've got very, veryvaluable data, right, patient
health information, but thenwe've got these really unique
(47:41):
vulnerabilities.
Right.
We've got all this new MIOTdevices.
You know all these IoT devices,smart devices that are not
nearly ready and the network youknow segmentation policy around
them is not nearly ready, andthe network you know
segmentation policy around themis not nearly ready to be fully
protective.
We've got the very vulnerablesupply chain, because it's
endless and we have blind spotsof you know where these things
(48:02):
are coming from.
And then this is something thatactually I brought up in the
talk that I think we don'temphasize enough.
But then there's the users.
We're introducing cyber toolsvery, very needed cyber tools to
folks that literally.
I mean, if you ask the averagephysician at an organization
(48:22):
what's a CISO, they may not evenknow that answer.
I said that during the sessionand the IT people were like but
I'm telling you they don't.
And and so I mean I havefriends who are practicing
pediatricians.
They don't know what that is,so and I don't say that aren't
secure.
I think that's a gap and AI isjust really accelerating.
Our need to do that anddeepfake and all these things
(49:00):
that they brought up during thesession is really scary, and so
we don't want to get behind thebar on that.
We got to really kind of stayforward.
So there's a lot that we couldbe doing there.
Brian Feldt (49:14):
Yeah, no, absolutely Eric.
Anything to add on that, onthat cyber front?
Eric Quiñones, MD (49:17):
Yeah, no, I think, I think, as as we
continue to move forward, Ithink it's a, it's a
vulnerability that we, we knowwe're going to have and I think
you know, seeing, you know, andjust now has kind of mentioned
it, you know, you know there'sgoing to be this prevalence
where AI is going to be involvedin the surveillance of an
(49:38):
environment, right, so you'regoing to have AI bots, if you
will, you know, really surveyingand being vigilant when it
comes down to protecting ahealthcare system, just as you
have, as bad actors are usingthose same kind of bots to try
to get in.
So you kind of have these, youknow, ai bot wars, I guess, if
(50:00):
you can think of it that way.
But I mean, that's just thereality we're living in and
we're seeing right now.
So it's just we cannot do it onour own.
We need to have the, you know,this intelligence that's working
in the background to be able tohelp be proactive and identify
the threats this conversation ayear from now.
Brian Feldt (50:34):
We're out of the HIMSS conference once again.
What are some of the trends andactivities within healthcare
related to AI, of course, thatyou think we'll be talking about
?
Is it agentic frameworks, andeither physicians or patients
have access to agents, that kindof work on their behalf or is
it more AI diagnosis?
What do you think we'll bespeaking about a year from now?
That'll really be important forall of the above to work
(50:56):
towards better patient outcomes.
Eric Quiñones, MD (51:00):
I'll jump in that really quick.
So my thoughts are, looking atmy crystal ball, it would be
we're going to have more data,right, we're going to have, you
know, more data to deal with,but I think we're going to be
seeing more.
Can I say it will be able to bemore predictive.
(51:20):
Looking at populations, right.
So being able to take in socialdeterminants of health that I
mentioned before genomics data,wearable data, and being able to
survey that population to beable to identify where in my
population of patients are thereal underserved, the
marginalized, the people that weneed to get activated and we
(51:43):
need to get you know, I wouldsay, in the system in a more
proactive way.
So I think we're going to seethat happening more.
We have to be able to scale theresources we don't have enough
of.
(52:03):
So I think we're going to seethese technologies help our
clinicians be able to scale bydoing these kinds of things I
just mentioned, to be able towatch a bigger population and to
identify where those particularproblems could be I shouldn't
say problems, but the signal andthe noise right, so you'd be
able to identify yeah, thispatient is going in the wrong
(52:26):
direction.
The trends are showing that andwe need to get them into the
doctor's office or we need tohave the care team contact them
Again, just having a highertouch with those patients, I
think.
Brian Feldt (52:41):
Yes, and as your crystal ball.
What are you seeing in there?
Sanaz Cordes, MD (52:45):
I think that we're going to start seeing AI
be more patient facing.
I think to date, a lot of it'sbeen behind the scenes, like we
talked about earlier.
You know they go to dosomething and their app or EHR
and maybe there's you know, aihappening on the background.
But I think with things likepardon me virtual human or
agentic AI or you know beingmore aware that they're
interacting with AI, I thinkwe're going to see more
(53:07):
investment and we have hadinterest just in there at WWT
building out solutions for thatpurpose, specifically like
agentic AI.
So I think there will be anincrease in AI in the hands of
patients directly.
Brian Feldt (53:22):
Awesome.
Well, lots happening in thespace, certainly helping out in
terms of clinician burnout andthings of that nature.
So exciting times within thehealthcare industry right now as
it relates to AI, and thank youto the two of you for joining
us here today.
It was an exciting conversation, lots of insights, so thanks
for taking time out of your busyschedules.
Sanaz Cordes, MD (53:43):
Thank you, Brian.
Brian Feldt (53:44):
Yeah, thank you, brian, appreciate it.
Okay, healthcare may be facinga crisis of capacity, but with
the right application of AI, itcould also be on the brink of
(54:05):
its greatest transformation.
Yet, thank you.
Reflect on everything we heardtoday, three big lessons stand
out.
First, ai is no longertheoretical.
Whether it's reducingdocumentation time through
ambient listening tools orlowering pediatric stroke risk,
ai is moving from pilot projectsto essential tools in the care
delivery process.
Second, healthcare must leadwith intentionality.
(54:29):
With so many new tools floodingthe space, leaders are pushing
pause, not on innovation, but onuncoordinated adoption.
The name of the game isgovernance, infrastructure
readiness and clear ROI.
And third, the human elementremains irreplaceable.
From patient trust to clinicianadoption, ai only works when
people are in the loop, usingthese tools to augment, not
(54:51):
replace, the irreplaceable humanaspects of care.
If you liked this episode ofthe AI Proving Ground podcast,
please consider leaving a reviewor rating us, and sharing with
friends and colleagues is alwaysappreciated.
This episode of the AI ProvingGround podcast was co-produced
by Mallory Schaffran, naz Bakerand Stephanie Hammond.
Our audio and video engineer isJohn Knobloch and my name is
(55:12):
Brian Felt.
We'll see you next time.
Whether you're a doctor, nurse, caregiver or just
an individual navigating life,we are all part of the
healthcare system, and AI isn'tjust knocking on the door.
It's stepping into exam rooms,hospital corridors and
administrative offices,fundamentally reshaping how care
is delivered Ambient AIcapturing conversations between
patients and physicians, largelanguage models assisting in the
(00:22):
triage to chart summarization,and AI tools helping to reduce
stroke risk or helping eliminateopioid use in post-op pediatric
patients.
This isn't just hype.
This is actually happening andon today's episode, I'll be
talking with a pair of chiefhealthcare advisors from WWT Dr
Sanaz Cordes and Dr EricQuinones.
Between them, decades ofexperience delivering care,
(00:44):
driving toward clinician andpatient outcomes via innovative
technology implementation.
But innovation alone isn'tenough.
This is healthcare.
Trust, intention and humanoversight are essential.
So today, sanaz and Eric aren'tjust talking about what's
possible.
They'll be talking about what'sworking, what's scalable and
what's next.
This is the AI Proving Groundpodcast from Worldwide
(01:07):
Technology.
In today's episode, we'llexplore how AI is moving from
concept to clinical impact,reshaping workflows, empowering
clinicians and transforming thepatient experience across the
healthcare landscape.
Dr Cortes, dr Quinones, thankyou so much for joining us in
(01:32):
the AI Proving Ground podcasttoday.
Sanaz Cordes, MD (01:36):
Hope the two of you are doing well.
Brian Feldt (01:38):
Doing great, excellent.
Before we dive deep into someof the uncoverings that we found
out at the HIMSS conferenceearlier last month, I do want to
ask when it comes down to it,we are all part of the
healthcare system, whether it'sa clinician, whether it's a
healthcare patient.
I'm curious what have you seenover the last couple years in
(02:06):
terms of how AI has altered thestate of healthcare and where
it's going?
Eric, start with you just realquick.
What have you seen from AI inthe landscape?
Eric Quiñones, MD (02:11):
Wow.
So, going back a couple ofyears ago, I would say we've
come a long way.
Ai is not new.
It's been around since the 50s.
But I would say, with thepromises that we'd seen maybe a
decade plus ago with, you know,with Watson and things like that
, I think we fell short of whatthe the intended promise was.
(02:34):
I would say, even in the lastcouple of years, from what 2022
(02:56):
to maybe even last year, it wasa lot of theoretical or hope
what we would see.
But this, this last year, Iwould say we're seeing a lot of
true innovation when it comesdown to various AI technologies
being used in healthcare.
So real, practical, I would say, implications that are being
used today.
Brian Feldt (03:11):
Yes, and as anything to add there in terms
of what you've seen, or maybeeven build upon what Eric just
talked about and give us alittle bit of what we expect
over the next 6, 12, 18 months.
Sanaz Cordes, MD (03:21):
Yeah, absolutely.
And you know, to add on to whatEric was saying, and just even
the last year, I think, with thedemocratization of AI, right,
because now everyone's got chatGPT at home and they're using it
for you know these things andthey're seeing tangible outcomes
from it, and it's puttechnology in the hands of
people that have, you know,quite frankly, in the past been
(03:42):
technology averse, right?
Physicians have always had ahealthy suspicion of tech and AI
of all things.
Is this black box, right?
Like we can't see the code?
We don't know that.
You know, it's just very it's adifferent scenario than even
just regular software, wherethey have a consensus building
process of actually kind ofcreating the tool.
So I do credit chat GPT andthat whole generative AI tsunami
(04:08):
, as Eric likes to call it, aspart of kind of making it
tangible in the last year or twoand having so much excitement
around it as well.
The one thing I will say thatthis HIMSS was evident is maybe
because of that and becausethere are so many options and
shiny object syndrome, right,and CIOs who are used to the
(04:29):
lines of business pushing backon IT.
Now they're seeing thosedifferent departments coming in
with their hey, can we use itfor this.
Hey, can we get that?
That there's now a much moreintentional approach to it.
Where I'm hearing fromleadership at HIMSS the
different organizations sayingwe're interested in AI but know
approach to it.
Where I'm hearing fromleadership at HIMSS, you know
the different organizationssaying we're interested in AI
but we're going to slow it downand make sure we're very
(04:51):
intentional about what tools weimplement, how we implement them
, how we measure, you know,metrics of success from them and
, quite frankly, the cost right,like wanting to kind of let the
noise settle and see who risesto the top, which startups or
others might, you know, go offinto the, into the background,
so that they can really, youknow, have a more intentional
(05:12):
plan for it.
Brian Feldt (05:13):
Yeah, Eric.
What did you see out at HIMSSthat led you to believe that we
are now actually moving fromthat experimentation phase to
the absolutely essential phasewhen it comes to AI in the
health care setting phase?
Eric Quiñones, MD (05:25):
to the absolutely essential phase when
it comes to AI in the healthcaresetting.
Yeah, no, I think there's somereally interesting things.
So, you know, I think there's.
When I think of the clinicalworkflow, where AI is really
hitting the there's, you knowthe rubber's hitting the road
would be when it comes down toambient.
You know technologies right.
So being able to use thosekinds of technologies to record
(05:47):
the conversation that thephysician and the patient are
having in context and putting itin a clinical note, and before
that patient even leaves theroom, that note is ready to be
reviewed by the clinician.
So there's always a person inthe loop, if you will.
So the clinician will bereviewing that note.
Their orders may be teed up andready to go as well.
So it's really impressive to beable to see that and where
(06:10):
they're seeing about a 70%reduction when it comes down to
documentation time.
That's significant.
And then you know other areasthat I think we're seeing it
really being impactful is whenit comes down to, you know,
handling inbox.
You know inundation that we getfrom various clinicians,
(06:31):
patients et cetera, in ourinboxes to be able to use.
You know, large language modelsto help in responding to those
things.
And again, there's always goingto be a.
The physician is still going tobe in the loop and going to be
reviewing that.
But you know the patients havebeen reacting very well to some
of those responses becausethey're more detailed, uh, the
more empathetic and not thatwe're not detailed and
(06:51):
empathetic people, we just don'thave the time sometimes to
really do that.
Um, and then, lastly, I wouldsay chart review.
We spend about a third of ourtime in reviewing charts and
patients come in time andreviewing charts.
When patients come in, what arethey coming in for?
I forgot what's this follow-upbusiness about, and so it will
pull up all that disparate dataand put it into a format where
(07:14):
it's curated and we can seeexactly what they're coming in
for, what may have been missed,et cetera.
So there's these real great usecases when it comes down to
those kinds of things.
So there's these real great usecases when it comes down to
those kinds of things.
I think, on the other hand, too, we're seeing it being applied
when it comes down to justautomation.
Right, it doesn't have to be socomplex, but when it comes down
(07:37):
to automating certain processes, where we're seeing more and
more clinicians being able towork at the top of their license
and when I say clinicians,that's just not physicians,
that's nurses as well and otherfolks on the care team.
Sanaz Cordes, MD (07:51):
Yeah, and Eric , just to circle back to your
first point about the ambient AI, Dr Poon anda few others, dr
Poon from Duke, I think you werethere, a bridge hosted one of
the sessions and something hesaid really kind of resonated
with me.
Where, you know, we introduce alot of technology that can
(08:11):
improve workflows and, to someextent, obviously, the
experience for clinicians.
But this was the first time Iheard somebody put it this way,
where it's like it c ompletelychanged the way doctors just
like the psychology and the waythey're just approaching patient
care.
What I mean by that is when wewere in this whole, you know,
(08:32):
two decades zone of like we'retrying to talk to a patient but
we're also, you know, typingeverything down.
We start thinking astranscriptionists and A it
distracts you from theconversation.
You're not as present as youneed to be transcriptionist and
a.
It distracts you from theconversation.
You're not as present as youneed to be Um and B.
The way you're, you know,interacting um can your goal
directed for the transcriptionright and not necessarily the
(08:54):
way you should be thinking asphysicians.
So giving that back and lettingdoctors just kind of go, I can
like focus on this patient andreally put my doctor brain the
way I was trained at med schoolor residency to work.
You know, that's just to me,you know that's an amazing thing
that technology can, you know,bring back to folks.
And then I think there wasanother physician, dr Mishuras,
(09:18):
just checking it, yeah, fromMass General, and she, I think,
was a part of that same panel.
And when she talked about, youknow, automating the clinical
workflows that Eric wasmentioning, I mean it's just
kind of like a triple win, right, like it's really improving
these clinical outcomes tolevels will readily admit as a
physician, sure, you know, I canprobably come up with a
(09:48):
differential diagnosis and usesome apps and such, but, you
know, humble enough to know thatthe AI could actually, you know
, find a new solution, biggerand better.
So you've got that, you've gotthe cost savings, you've got the
efficiency and, of course, thepatient experience.
So, overall, I mean, I thinkit's just it's.
You know, I have a lot moreexcitement this year because I
think we're seeing actual,tangible results and it's a
(10:09):
great time to be coming backfrom HIMSS riding on that high.
Brian Feldt (10:21):
Yeah, love to hear it.
I'm going to put a pin in that.
Operational workflows here forjust a second Synaz.
I do want to ask you you'retalking a lot about efficiencies
gained through things liketranscription or other use cases
like that that seem prettypractical.
Anything at the HIMSSconference that would speak more
towards the innovation end ofthat spectrum that caught your
eye or that you think will makea real impact either this year
(10:43):
or in the near future?
Sanaz Cordes, MD (10:44):
caught your eye or that you think will make
a real impact either this yearor in the near future.
Yeah, absolutely.
Himss really started off withkind of a bang.
This year we're all kind ofsettling in and getting to our
seats at the keynote and for me,having gone to HIMSS for over I
don't know 15, 16 years now,that opening session was amazing
.
We had the CEO and the CMIOfrom Samsung Medical Center in
(11:06):
Korea and they've won thehighest level of certification
for HIMSS now of any otherorganization, and so they
presented everything thatthey've done at.
It's like the HIMSS indicatorscore.
But they presented everythingthey've done at this hospital
and it's just amazing.
(11:27):
They've, first of all, lessonslearned.
So they created amulti-stakeholder you know,
multidisciplinary approach tohow they're going to roll out
this real connected health model, the smart hospital model, and
so, for example, physicians hada seat at the table saying you
know, when the EHR does X, y andZ, it's disjointed from that.
(11:48):
We would really love it if itwas this way and other folks are
weighing in and they actuallybuilt their own EHR this way and
it's kind of an AI first, ifyou will, kind of EHR model, and
it integrates beyond just thatright, like it's leveraging
large language models to combthrough data and help with
decision making, but it'sautomating tasks downstream.
(12:11):
It's doing like predictive sortof forecasting of you know,
between two and six is when wehave the most MRI orders or
portable, you know, chest strikeorders.
So maybe we forecast how that'sgoing to.
We're going to tee up resourceallocation and optimization,
like in the future.
And then, yeah, I mean it wasjust mind blowing because it's
(12:33):
like if we had a blank slate andcould rewrite healthcare to
make it better, you know thanwhat happened, you know, 20, 15,
20 years ago when everyone madean EHR and now we're dealing
with all these siloed processes.
So, and then there were therobots.
Eric Quiñones, MD (12:51):
I don't know if, eric you want to talk about
the robots those were.
She brought robots.
That was very innovative and sowhat it really helped do, I
would say, you know, especiallyfor the pediatric population
right, that are in the hospitalAgain, hospitals can be very,
you know, just daunting places,and even for adults, so you can
imagine the children.
So these robots are veryinteractive and they were able
(13:14):
to, you know, with the pediatricpopulation and be able to, you
know, do things with them and,you know, I would say, somewhat
distract them from all the otherthings that happen in a
hospital.
But it really helped.
Calm them down is really theend point.
So I thought that was a reallyunique way, or innovative way,
to bring in some, you know,artificial intelligence into the
(13:38):
healthcare space.
Brian Feldt (13:39):
So, Naz, I am curious what took place at HIMSS
in terms of the regulatoryenvironment.
I'm curious what took place atHIMSS in terms of the regulatory
environment.
Lots to unpack in that questionand lots of news and
information could happen overthe next coming weeks or months,
but what were physicians orclinicians talking about when it
comes to the regulatoryenvironment and how that might
impact AI adoption?
Sanaz Cordes, MD (14:00):
Yeah, I mean lots of changes coming for sure
every hour for the nextunforeseeable future, right, I
mean, with political thing, youknow, unprecedented changes that
are being handed down andthere's already a lot of impact
from this.
I think almost everyconversation I had that factored
in.
I'll just give an example of,you know, academic research
organizations.
(14:20):
I worked I spoke with one ofour customers that we're working
with and already they arerestructuring and that starts
from the people, the process andthe technology, right, Because
with the funding change nowthey're having to shuffle data
science dollars here and there.
They're having to figure outwhat tech they can invest in and
what's going to serve morepurposes, so that they're having
(14:42):
to be more strict about theirprioritization.
So it's already there and Ithink there's just so much
unknown and that's kind ofparalyzing some decision making
as well.
So it's already definitelyhaving an impact for sure on the
academic research organizations.
Eric Quiñones, MD (15:01):
Yeah, I'll add really quick.
So just I think the last week,the end of last week, so Johns
Hopkins, they actually announcedthat they're going to be laying
off to 2000 employees,including a significant amount
of the researchers, because thiswas due to what?
An $800 million grant that wascut from what?
(15:22):
The US Agency of InternationalDevelopment.
So you know you're going to beseeing more of that.
And then also I read thatbecause their residency programs
sometimes are federally funded,you're going to be seeing these
residency programs actuallybeing cut as well, in a time
where we need more residencyprograms.
(15:44):
We just can't produce enoughdoctors.
As a matter of fact, the datashows we'll never have enough
doctors ever again.
So you know you could have.
We have a surplus of medicalstudents, graduates, but we
don't have enough residencyprograms to train them.
So you know these kinds ofthings are going to impact
healthcare very significantly.
Brian Feldt (16:06):
I would assume Eric to just build on that.
That's an opportunity as wellfor AI to come in and
potentially supplement, or isthat not necessarily the line of
thinking here?
Eric Quiñones, MD (16:18):
I think, maybe in some ways, but I think
you can't replace some of theseresearchers.
I mean, it's so their work thatthey do is so, so critical.
As a matter of fact, you know,I would say that even in some of
that research you know, thereare, you know, maybe engineers
that are AI data scientists aswell.
So, yeah, they're just notthey're being, you know, I would
(16:42):
say, eliminated in some waysbecause of not because of the
it's more of the politics that'seliminating them and the grant
reductions that are that areeliminating them.
But, yeah, it's going to be avery interesting period, but I
have faith that we're going tofind a way.
We're very good at doing thatas humans, and I think the
(17:07):
research is too important tojust not do it, and so we're
going to figure out ways to doit.
It's going to be very novel,but, yeah, I don't think it'll
replace the scientists in theroom the scientists in the room.
Brian Feldt (17:26):
Yeah, before we move on from HIMSS in general,
eric, any other kind of keytakeaways?
I know Sanaz mentioned a coupleof the keynotes, anything that
you just thought would beworthwhile mentioning or things
that caught your attention.
Eric Quiñones, MD (17:37):
Yeah, I think , going back to real life use
cases, I would say where we'reseeing AI being.
You know the benefits thatwe're seeing today.
One came from Dr Chaudhry.
He is the chief AI officer atSeattle Children's and there was
(17:58):
two use cases that again, hehas the right people process
data technology in place and hereally stresses that it's like
you have to have those things toachieve your goals.
And one goal was they wanted tobe able to reduce opioid
(18:19):
treatment in terms of for thepediatric population to reduce
opioids by 100%, and they use,obviously, opioids for pain and
sedation, but they can have avery addictive quality to them.
So what they did was they usedthese various AI models to help
identify different medicationcombinations to help achieve the
(18:42):
goals that they were trying toachieve in terms of sedation and
pain management, and they wereable to do so, again by 100%.
So that was significant.
And then the other use casethat Dr Chaudhry actually talked
about was reducing pediatricstrokes in post-operative
patients that have had brainsurgery.
So when they go to the ICU,it's not uncommon for a child,
(19:08):
to you know, to have a strokeand no matter what we do, we do
the best we can, but it canhappen, and you can imagine the
outcomes are not very good whenthat happens.
So they were able to applyvarious models to identify which
particular patients are at riskand take those precautions, and
(19:28):
so they were able to reducethose strokes by 50%.
So those are significant usecases.
I think that we weren't able todo before the technology
existed.
Brian Feldt (19:38):
Yeah, I do want to add.
I want to go back to thatoperational workflow idea that
you brought up, sanaz.
I'm wondering as we integratemore AI into the workflows and
the clinician setting, how isthat shifting, how we work
within a hospital, within aphysician's office or anywhere
(19:58):
within the healthcare setting?
What are you seeing as kind ofthose big operational shifts as
we move closer towards that AIadoption?
Sanaz Cordes, MD (20:06):
Yeah, I mean, I think in many ways it's an
operational shift and just abehavioral kind of again use the
word psychological shift,because if you, you know, if you
really think about it, the EHRand a lot of the technologies in
the hospital, the EHR wasn'tinvented to make work better for
anybody, right?
It was a billing and claimsrepository, right, and all of a
(20:27):
sudden, overnight, it createdtwo hours of clicking.
And you guys have all heard mesay this over and over in the
last decade and I really feelthat just in the last couple of
years.
It's like doctors are taking itback, right, clinicians are
taking it back.
It's like, I mean, we have ashortage, right, and the numbers
are all over the place, but forsure, over $60,000, 60,000
physicians shortage, you know,within the next year or two, and
(20:50):
it's not getting any better.
And so I think people finallyare sitting up and sitting up
straight and listening to this.
So I do think it's just thisshift of like we're taking back
medicine, you know, and so byintroducing these automations,
we're just trying to almost,like, go back in time, right, we
want to be able to talk to ourpatients and focus on what's
(21:10):
happening, not be distractedenough to be able to take in the
psychosocial things that areimportant, their social
determinants of health, theirhistory, and not be spending all
this time.
I mean, the rate of imaging thatis redone in ERs is just
staggering.
Right, it's gotten better, butI remember in my days and not so
(21:33):
long ago for others wherepeople come in with a known
underlying problem they had anMRI or whatever done in the same
city maybe, and it's 11pm atnight and you're not going to
get that.
So what do you do?
You repeat the MRI, right?
So I think, by automatingthings and letting AI and other
tools and interoperability workfor you to do that sort of
(21:54):
administrative noise and let youwork top of license, you know,
I really feel like cliniciansare taking healthcare back and
again win-win for the healthsystem, because, because MRIs
cost a lot, you got to pay forall these utilizations and then
that comes back.
That gives back to us as asociety and caring for
increasing aging population, forexample.
(22:15):
Eric talks about that a lot andthat's not going to get any
better, right?
So, yeah, I think it's justfundamentally going to change
how we do everything.
And then there's the wholeoperational piece, right.
I mean claims and billing, andall of that has been a nightmare
from day one, not just for theclinicians and nurses, but for
everybody and for thatdepartment.
And there's so much more theycould be doing right.
(22:38):
Think of what a disasterauthorizations are for things
like referrals and specialtystudies and just price
transparency, pathways andfinancial experience for
patients.
So if we could free those folksup to be able to work and move
the needle on some of that toyou know, long-windedly answer
your question.
I think it's just fundamentallygoing to shift how we do
(22:59):
healthcare and we, quite frankly, we need it because we're at a
crisis point, in my opinion, ifwe don't change things around.
Brian Feldt (23:06):
Agreed.
Yeah, not to play devil'sadvocate here, sinez, but what
about the concept of an AIburnout loop here?
If we are able to be moreefficient and come to
conclusions faster, does thepotential exist that clinicians
are just going through more andmore patients Like?
(23:28):
Is there a balance that needsto be struck in terms of giving
the patients and their familiesor caregivers the time that they
would need or deserve, asopposed to just relying on AI to
get them through the door?
Sanaz Cordes, MD (23:40):
Yeah, I mean, of course there's always risks
with any kind of automation, Iwould say.
My answer is I'm curious tohear Eric's is we are in such a
shortage right now, especiallyof physicians, right where we
are trying to increase other,not mid-level, but the extension
(24:00):
practitioners, pas, nursepractitioners.
We're trying to increase thatpopulation.
We can't do that fast enougheither.
So I think that we are far farfrom having a problem of surplus
AI and freeing up so much timethat we're going to create this
new swing of the pendulum.
I think we're going to finallylet people stop running 100
(24:23):
miles an hour and let them goback to a normal approach, you
know, approach to a workday,which again, can only benefit
everyone.
So I mean, like I said, there'salways a risk.
I don't see that happeningbecause we're just so far behind
and there's so loss of joy inpractice that I think we're
going to give that back in thenear term and maybe even mid to
long term future, before wecross over any other threshold.
Eric Quiñones, MD (24:47):
But, eric, if you want to add to that, yeah,
no, I think I like throwingthese stats out and they vary a
little bit depending on whatyour resource or your resources.
So in the next five years, by2030, the numbers I've seen is
we're looking at a deficit ofabout 140 at the most physicians
(25:08):
across the US, about 200,000nurses across the US in general
as well, and we've been runninga deficit in nursing for a while
now across an average in the US.
In the next five years, everybaby boomer will be 65 and over.
That's 20% of our population,so you're going to have an older
(25:28):
population.
That means a sicker population.
In addition, in 1950, medicalknowledge doubled every 50 years
.
Today it doubles every 74 days.
I cognitively can't keep upwith that.
I would have to read 20 hours aday just to keep top of license
, okay, so things are changingreally fast.
(25:49):
And then the data that we have.
We think an average hospitalproduces about 50 petabytes of
data a year, but now that datais going to be going up even
higher because we're going to beseeing other data sources, such
as genomics, right, and theother biomes as well.
We're going to be seeing otherdata sources, such as genomics,
right and the other biomes aswell.
We're going to be seeing otherdata sources as far as wearable
(26:09):
technologies, right, so thosedata sources will be coming in.
So there's more data to process.
We just can't do it alone, so weneed these technologies to help
us curate and, really, what isthe signal in the noise?
Help us to be proactive andpredictive, to find those
(26:32):
patients that really need ourattention well before a problem
starts.
So, yeah, I think we're goingto be seeing a lot of change
when it comes down to the way wepractice medicine, and I'll
even add that I know new youknow residents that are new
graduates in the residencyprograms.
(26:53):
As they're coming out, whenthey're interviewing in
hospitals, they're asking notabout just their salaries and
their other benefits.
They're asking me what kind oftechnologies do you bring to the
table that actually will helpmake my life different or better
?
So that was never a thing forme.
I know that it was like go towork here and, matter of fact,
(27:17):
it was paper which is not a badthing, but, we.
You know we're not there anymore, but no, those are things that
are happening today and, I think, where we're going to be really
seeing these technologiesreally help and change the way
we're practicing health care.
(27:37):
I'm very hopeful and excitedfor the future, but right now
we're still dealing with a lotof these issues that Dr Cortes
has actually mentioned.
Sanaz Cordes, MD (27:48):
You know, and there's also the patient piece,
right, Like we're talking aboutoperational workflows as it
benefits hospitals andclinicians, but then there's the
patient piece, right, so whatpatients or potential, you know,
patients, healthcare consumersare willing to accept anymore
has definitely changed, likethat bar has raised.
You know patients, healthcareconsumers are willing to accept
anymore has definitely changed,like that bar has raised.
You know, and I liken it to youknow, let's say you're trying
(28:09):
to pick a restaurant to go toand you're between meetings and
one has online booking forreservations and one doesn't.
Which one are you going to?
I mean, I'm obviously alwaysgoing to default to the one that
has.
So patients as well.
Right, they have choices now,and it's not just this mediocre
tech health system versus youknow this slightly less mediocre
(28:30):
, it's my health system.
Or I'm going to pay $29 orwhatever and get instant access
to a physician and aprescription, and you know,
online, or I'm going to walkinto one medical and be seen.
So there's a lot of more.
You know, more options.
And so by implementing thesetechnologies that are often, you
know, dual facing right, so itallows them to be able to do
(28:51):
dynamic appointment schedulingor medication management or, to
Eric's point, as we get moredata and we need to be more
proactive, push more proactivereminders, you know, to the
patients and just a myriad ofother things that are beyond the
table stakes of these types ofthings I'm listing right, like
they want a better overallexperience, access to care 24-7,
have them meet them where theyare right.
(29:13):
Like there are some things thatI'm willing to drive across town
and wait an hour in a waitingroom for, but there are other
things that I'm frustrated thatI can't just dial up, and maybe
even an asynchronous visit whereI don't need to have the doctor
present.
This should be able to be astore and forward type of
engagement where somebody candeliver it.
So I do think that there's thatpatient component and what
people are now, as consumers,willing to do and they'll vote
(29:35):
with their feet if they're notgoing to get it at that hospital
and go somewhere else.
Brian Feldt (29:50):
I think that patient experience is an
interesting topic to dig into alittle bit.
As ambient technologies startto enter the clinician setting
and you know there's kind of analways-on listening aspect to
when you're in the doctor'soffice, does there need to be a
transparency conversation withpatients?
And then, I guess you know,moving that conversation forward
(30:11):
is, where does AI go from herein terms of, like, actual
diagnosis?
Eric Quiñones, MD (30:27):
And how do?
Sanaz Cordes, MD (30:27):
health organizations balance that with
what they tell the patient inthe waiting room Right.
Go ahead, eric, go ahead.
I was just going to share, justto put it in kind of a
perspective.
You know, about 50% of patientsview, you know, ai as a
positive thing, right, andthat's up a lot from the past.
I think it's like 53 percent.
So we're definitely kind ofmoving that needle anyway,
(30:48):
because they again are seeingsome of the benefits of that.
So I just kind of wanted tomake sure we touch on that,
because there's there is demandand acceptance of this as a as a
means to giving them theexperience that they want.
But yeah, eric, if you means togiving them the experience that
they want.
But but yeah, eric, if youwanted to go in more into what.
Eric Quiñones, MD (31:07):
Yeah, no, no, that's I think.
I think we're seeing very, avery positive reaction from
patients.
So what?
The one example I mentionedearlier was inbox management,
right?
So let's say you know you ask aquestion of your physician and
you're waiting for a responseand something that's's before AI
, there was no AI and they'rejust you're waiting and waiting,
(31:28):
and waiting.
You finally get a response.
It could be a very curtresponse and you know to the
point, and not that thephysician was trying to be rude,
it was just the time is notavailable to really take time to
write a very lengthy, detailedresponse.
Available to really take timeto write a very lengthy,
detailed response.
Fast forward where patients aregetting those responses from an
(31:49):
AI assistant and it'll say inthe message this is an AI
assistant for Dr So-and-so andthey would get a very detailed
message.
They would get a very heartfelt, empathetic message and the
responses that they gave whenasked, when they were very, very
(32:10):
happy with what they werereceiving.
So I think you know there's alot of excitement about that,
but I also think that therealways has to be a person in the
loop, right?
This can't just be.
You know this is sent outautomatically when these
messages.
When they were sent out, theywere still reviewed by the
clinician and they're being sentout, but I think that's
(32:32):
important too to know.
But no, it's exciting.
Sanaz Cordes, MD (32:37):
But you bring.
I mean, brian, but you didbring up a good point because,
like to date, I mean, I'm notaware when my doctor's using
clinical decision support andusing one of maybe Epic's AI
algorithms to do something, andto date I think like that's
okay-ish right.
But you bring up a good pointof at what point isn't it okay
(32:59):
without consent right, okaywithout consent right.
At what point, like even withAmbient, as a patient, I am
really happy that I'm beingrecorded because, being on both
sides of that equation, one ofthe biggest frustrations is when
you go and you visit yourphysician and then you know
you're supposed to get thesefollow-up things and then, as
you're getting them, you'rerealizing wait, I already told
(33:21):
him that I have X, y and Z, sowhy isn't there the prescription
there?
So, as a physician, now in thatequation, you know you've got
all that content and so whenyou're getting ready to do your,
you know we call it theassessment and plan, coming up
with your you know treatmentplan.
Like you have all that data,you're not going to miss those
things.
So, again, to date, like it'spositive, positive, and you know
(33:43):
we don't necessarily thinkabout that.
But you know we've got 800 FDAapproved.
You know AI algorithms today.
What do we do, to your point,when it's like 8,000, right,
there has to be some kind ofprocess, and I don't.
I don't have the answer forthat, but it's really a good
point and something worthwatching.
Brian Feldt (34:05):
Yeah, I like the idea of that, the human in the
loop, that aspect, although Ithink it also just begs the
question.
I think a lot of people havethis type of question, which is
how effective is AI in terms ofdiagnosis?
We recognize the need to havethat human in the loop, but at
what point in the future does AIor is it already becoming on
par better than human diagnosis,or is it already becoming on
(34:27):
par better than human diagnosis?
Eric Quiñones, MD (34:30):
I would say it's getting there.
As a matter of fact, I wouldsay it's maybe a year ago it was
at the level of a first-yearresident, but now what I'm
seeing is that it's surpassingthat and again, what matters is
the data.
So we talk about, you know,interoperability is a very
important part of this topic.
(34:50):
If we don't have the data, it'shard to make those, you know,
to run those algorithmsappropriately to get the right
outcome.
But I've seen situations wherethey've had some really rare I
would call it, you know, kind ofreally rare internal medicine
type problems.
(35:10):
I mean, these are real cerebralissues that a patient has.
I mean, they're medicalmysteries and they've applied,
you know, models to.
You know identify what theproblem is.
And differential diagnosis thefirst diagnosis and differential
diagnosis the first diagnosison the differential diagnosis
was the correct diagnosis, soit's getting really good.
(35:30):
Now that doesn't mean that wejust say, hey, let's go with the
first diagnosis and run withthat.
We still have to look at that.
We still have to, you know, askthe question well, what is this
based on?
You know, and I think when youhave that transparency that's
within the model and that maybeyou know, dr Cordes or myself or
any clinician can actuallyclick on and see okay, this is
(35:52):
based on this information, this,this is the you know, the
evidence-based medicine it maybe using, et cetera.
Then it brings more credibilityto what we're seeing.
Nothing will take away the youknow, the clinical mind.
We still have to apply that.
So, but to get to the mysteryquicker, to find out what's
(36:15):
causing it quicker, I think thathelps a tremendous,
tremendously.
You know we can spend a lot oftime trying to figure out a
problem in medicine and I knowI've seen, you know I've seen
patients go through thisparticular process where they've
been misdiagnosed, misdiagnosed, misdiagnosed by multiple,
multiple physicians.
And, as a matter of fact, Ihave a good friend of mine,
(36:37):
she's a nurse, a colleague.
That happened to her and then Itook her information, I put it
in chat, gpt, just to experiment, and it actually kicked out the
differential diagnosis and itwas the first one that they
could not identify.
So it's getting pretty good.
Sanaz Cordes, MD (37:00):
Wow.
I didn't know that story.
Eric Quiñones, MD (37:02):
That's pretty cool.
Sanaz Cordes, MD (37:04):
You know, I think we can use imaging, you
know, as an example, becauseimaging was the first right that
embraced ai and radiologistsembracing ai and that has come
such a long way, I mean, even atworldwide.
You know we've had um, we'vehad our data scientists just you
know, on the bench who wantedto work to solve a problem, be
able to, you know, do a braintumor radiomics.
(37:27):
You know mri project where itcan diagnose the.
You know the end brain tumordiagnosis like 20x faster, and
that's just.
You know a diagnose the.
You know the end brain tumordiagnosis like 20x faster, and
that's just.
You know.
A couple of data scientists,you know, playing on our models.
There's solutions out there thatare just they've been out there
for years and so imaging, Ithink, is a great example of.
You know it started out okayand then people were meticulous
(37:47):
about it and now it's reallyubiquitous.
I mean, most health systems areusing it.
I do think that there's alsothe flip side.
It's not all rosy.
The EHR I mean, let's just callit what it is Epic has a habit
of when these new third-partytechnologies become available
now, most of them AI they havethis habit of telling their
(38:10):
customers oh, we're buildingthat or it's coming in a release
next year.
Oh, don't buy this because andthe health systems also don't
want to be piecemealing a bunchof disparate third-party apps.
So I get it, but they're alsostarting to get a little
impatient and a littlefrustrated with that answer
coming from Epic.
So they do use obviously asmuch of Epic as they can, but
(38:32):
I'm picking on Epic because Iremember it wasn't that long ago
where their Epic sepsis modelwas really faulty and you, that
is not an area you want to be,you know messing with, and so we
can't just kind of blindlytrust it and say that it's
getting as good, um, or better,globally I think it's.
You know very specific um usecases that still require a lot
of.
(38:52):
You know the models gettingmore accurate and smarter and us
holding on to that.
You know cautious skepticismuntil you know we get to maybe
the level, like, of imagingthat's been there now for a
while.
Brian Feldt (39:14):
Well, that's a good segue into the IT conversation,
which is also a very importantcomponent here, recognizing that
AI is moving so incredibly fastand a CIO or an IT team for one
of these healthcareorganizations just have so much
coming at them, whether it's anew vendor or thinking about
(39:34):
their data estates or whateverit might be.
What are some of theirpriorities that they must be
thinking about now so that theycan actually implement and have
a workforce that adopts andscales AI on their behalf?
Sanaz Cordes, MD (39:48):
Yeah, I mean, I'm happy to start with that.
I would say, just in the lastsix months, that's what a lot of
my meetings are about, as youknow, as I meet with some of our
executives at theseorganizations and wherever you
are, however big you are, youknow you could be Northwestern.
Who's on the front end?
I mean, they're doing amazingthings and they've got a whole
system and a workshop.
Or you might be a smallerhealth system.
(40:08):
At some point you're workingtowards an AI center of
excellence health system.
At some point you're workingtowards an AI center of
excellence, and what I mean bythat is, you know, maybe it is
literally a center where you'vegot data scientists and folks
and you know, governancecommittees, or maybe it's just
the CIO, who's got now a hundreddifferent use cases being
demanded of him, and IT teams or, yeah, it folks that are maybe
(40:30):
not skilled in implementing this.
I mean, that's a big factorright now is just getting talent
that can be skilled for AI andfor the cybersecurity component
around AI.
But wherever you are, you'retrying to work to an excellence
model where there's some sort ofstandardization of ingesting
these ideas, standardization ofprioritizing and mapping them to
(40:52):
feasibility versus cost, versusoutcome and then a methodology
of how they're built right.
Does it come off a third party?
And, if so, what are thecriteria for that?
And then Eric brought up themost important point earlier,
which is the data.
Dirty data in, dirty data out.
We've been saying that foryears about software and now
more than ever with AI, we can'temphasize how important that is
(41:14):
.
So then, what's yourmethodology around data?
And then that leads you to thewhole.
Like this is really kind of aplatform play.
These things can't happen insilos, right?
Because if we can rinse andrepeat to some degree all the
effort we've put in to createthis model off of this LLM, we
need to be able to repurposethat.
(41:36):
So it's definitely acoordinated, multidisciplinary,
multi-stakeholder process andthat's where we need to get.
And so our goal mine, when Imeet with customers is don't be
overwhelmed by all that.
Let's just start with one usecase and figure it out.
But each time we're gettingsmarter, just like the AI, so
that as we move forward, we cango from one to five to 10 to 100
(41:57):
without having to, in linearform, increase the effort.
You know, one to 10 to 100.
Brian Feldt (42:02):
Yeah, eric, I do want to ask you about data.
You've mentioned it a number oftimes already and, sanaz, you
just mentioned it as well.
Understanding that the dataestate of these health care
organizations is such a complexarea Some may be sophisticated,
some may be not as much what arewe advising organizations in
terms of how to best leveragetheir data?
(42:24):
Is it start small and work yourway up with momentum, or is it
cleaning everything at once?
How are we talking to clientsabout how they should approach
and leverage their data?
Eric Quiñones, MD (42:35):
Right.
Right, I think one of the firstthings I you know well, this
happens a lot when I'm visitingwith clients and we're having
these discussions it's do theyeven know where their data's at?
You know it's their datagovernance right, starting there
.
And, quite frankly, a lot ofthem can't really answer that
question, sometimes very clearlyor succinctly.
(42:56):
Or, what condition is that dataright?
Is it, has it been?
You know, is it discrete data?
You know, is it just text data?
So you know, again, these arethe conversations I have all the
time and it's like, so that'swhere you kind of have to start
Right.
And then do you need all thedata have and your AI governance
(43:18):
?
You know folks that are in theroom helping to define those
particular use cases, but itreally no matter all of that if
(43:42):
you don't have the data.
You know some of those coredata components you can't do the
.
You know the work that you wantto do.
So really starting there is soimportant and really identifying
where that data is going to be.
So you know, maybe right nowit's in a place where, like I
(44:02):
said, it's disparate, butthey're considering, you know,
moving the data into anenvironment that will give them
more accessibility to the dataand who has accessibility to the
data right?
You know they have their ownresearch teams that they need to
have accessibility, versusothers that don't need to have
that accessibility.
So there's a lot of questionsthat need to be asked and a lot
(44:25):
of you know that come out of alot of these meetings.
And it's good because you knowwe do identify, you know what
next steps need to happen.
And so, because they do have agoal of doing a lot of these
things and they know they haveto do it because they cannot
continue to practice medicinethe way that they're doing it
(44:48):
today they know they need to usethese tools, but they can't use
these tools unless they havethat data governance in place.
So I would even add that theyneed to use these tools, but
they can't use these toolsunless they have that data
governance in place.
Sanaz Cordes, MD (44:55):
So I would even add that they can't use
those tools until they have theinfrastructure modernized in
place, right, so that's.
Another thing is, like you know, when we go in there and we I
mean health care, nonprofithealth care is not a profitable
right, like, obviously, industrythere, but they're very budget
restrained and so a lot of theseorganizations are operating on
(45:17):
legacy tech, end of life tech, alot of stuff's on premises, and
so, again, that's why we kindof say let's just start with one
use case, so that we don'toverwhelm and kind of throw it
all, maybe go with the bathwater, because with that one we'll
figure out what's the mostimportant.
Is it your cloud migrationstatus?
(45:38):
Do we need to focus on a hybridcloud model?
Or is the barrier to success onthat one the fact that you have
a storage challenge?
But there is inevitably goingto be an infrastructure
modernization component thatgoes hand-in-hand with that data
strategy that Eric'sreferencing.
Brian Feldt (45:57):
Yeah, and Eric, you mentioned data governance as
well, and I think I just readmaybe it was last month an ECRI
report about the top 10 threatsright now in healthcare data
governance rated number two.
And that also actually bringsup another issue of just
cybersecurity in general, whichI understand was another
important topic at HIMSS when dowe stand in terms of AI, cyber
(46:21):
data protection within thehealthcare industry?
What are some best practices orwhat are some questions that
leaders need to be askingthemselves?
Eric Quiñones, MD (46:29):
Yeah.
So, Naz, you want to.
You want to take that Cause Iknow that you were very heavily
engaged in some of thoseconversations.
Sanaz Cordes, MD (46:36):
Yeah, yeah, I had a great session on that
actually with Paolo.
We hosted several folks duringHIMSS.
Yeah, I mean, I think AIabsolutely has brought
cybersecurity front and center,not that it hasn't been.
I mean, we had, I think, like720 breaches last year and sure
(46:57):
that's down a couple percentfrom the year before, but we
have to always think about thefact that we're like 2x where we
were before 2018.
So I mean, it's a heat andeveryone knows about the change
healthcare thing, and so nowwe've got AI and there's AI's
role in cyber, got AI and youknow there's AI's role in cyber.
There's AI for cyber, and thenthere's cyber to protect against
(47:18):
, you know, bad actors from AI,and so one of the things with
why we're at such risk inhealthcare is just our
environment.
There's a reason why you knowwe're in the top two costliest
industries for cyber risks.
Right, we've got very, veryvaluable data, right, patient
health information, but thenwe've got these really unique
(47:41):
vulnerabilities.
Right.
We've got all this new MIOTdevices.
You know all these IoT devices,smart devices that are not
nearly ready and the network youknow segmentation policy around
them is not nearly ready, andthe network you know
segmentation policy around themis not nearly ready to be fully
protective.
We've got the very vulnerablesupply chain, because it's
endless and we have blind spotsof you know where these things
(48:02):
are coming from.
And then this is something thatactually I brought up in the
talk that I think we don'temphasize enough.
But then there's the users.
We're introducing cyber toolsvery, very needed cyber tools to
folks that literally.
I mean, if you ask the averagephysician at an organization
(48:22):
what's a CISO, they may not evenknow that answer.
I said that during the sessionand the IT people were like but
I'm telling you they don't.
And and so I mean I havefriends who are practicing
pediatricians.
They don't know what that is,so and I don't say that aren't
secure.
I think that's a gap and AI isjust really accelerating.
Our need to do that anddeepfake and all these things
(49:00):
that they brought up during thesession is really scary, and so
we don't want to get behind thebar on that.
We got to really kind of stayforward.
So there's a lot that we couldbe doing there.
Brian Feldt (49:14):
Yeah, no, absolutely Eric.
Anything to add on that, onthat cyber front?
Eric Quiñones, MD (49:17):
Yeah, no, I think, I think, as as we
continue to move forward, Ithink it's a, it's a
vulnerability that we, we knowwe're going to have and I think
you know, seeing, you know, andjust now has kind of mentioned
it, you know, you know there'sgoing to be this prevalence
where AI is going to be involvedin the surveillance of an
(49:38):
environment, right, so you'regoing to have AI bots, if you
will, you know, really surveyingand being vigilant when it
comes down to protecting ahealthcare system, just as you
have, as bad actors are usingthose same kind of bots to try
to get in.
So you kind of have these, youknow, ai bot wars, I guess, if
(50:00):
you can think of it that way.
But I mean, that's just thereality we're living in and
we're seeing right now.
So it's just we cannot do it onour own.
We need to have the, you know,this intelligence that's working
in the background to be able tohelp be proactive and identify
the threats this conversation ayear from now.
Brian Feldt (50:34):
We're out of the HIMSS conference once again.
What are some of the trends andactivities within healthcare
related to AI, of course, thatyou think we'll be talking about
?
Is it agentic frameworks, andeither physicians or patients
have access to agents, that kindof work on their behalf or is
it more AI diagnosis?
What do you think we'll bespeaking about a year from now?
That'll really be important forall of the above to work
(50:56):
towards better patient outcomes.
Eric Quiñones, MD (51:00):
I'll jump in that really quick.
So my thoughts are, looking atmy crystal ball, it would be
we're going to have more data,right, we're going to have, you
know, more data to deal with,but I think we're going to be
seeing more.
Can I say it will be able to bemore predictive.
(51:20):
Looking at populations, right.
So being able to take in socialdeterminants of health that I
mentioned before genomics data,wearable data, and being able to
survey that population to beable to identify where in my
population of patients are thereal underserved, the
marginalized, the people that weneed to get activated and we
(51:43):
need to get you know, I wouldsay, in the system in a more
proactive way.
So I think we're going to seethat happening more.
We have to be able to scale theresources we don't have enough
of.
(52:03):
So I think we're going to seethese technologies help our
clinicians be able to scale bydoing these kinds of things I
just mentioned, to be able towatch a bigger population and to
identify where those particularproblems could be I shouldn't
say problems, but the signal andthe noise right, so you'd be
able to identify yeah, thispatient is going in the wrong
(52:26):
direction.
The trends are showing that andwe need to get them into the
doctor's office or we need tohave the care team contact them
Again, just having a highertouch with those patients, I
think.
Brian Feldt (52:41):
Yes, and as your crystal ball.
What are you seeing in there?
Sanaz Cordes, MD (52:45):
I think that we're going to start seeing AI
be more patient facing.
I think to date, a lot of it'sbeen behind the scenes, like we
talked about earlier.
You know they go to dosomething and their app or EHR
and maybe there's you know, aihappening on the background.
But I think with things likepardon me virtual human or
agentic AI or you know beingmore aware that they're
interacting with AI, I thinkwe're going to see more
(53:07):
investment and we have hadinterest just in there at WWT
building out solutions for thatpurpose, specifically like
agentic AI.
So I think there will be anincrease in AI in the hands of
patients directly.
Brian Feldt (53:22):
Awesome.
Well, lots happening in thespace, certainly helping out in
terms of clinician burnout andthings of that nature.
So exciting times within thehealthcare industry right now as
it relates to AI, and thank youto the two of you for joining
us here today.
It was an exciting conversation, lots of insights, so thanks
for taking time out of your busyschedules.
Sanaz Cordes, MD (53:43):
Thank you, Brian.
Brian Feldt (53:44):
Yeah, thank you, brian, appreciate it.
Okay, healthcare may be facinga crisis of capacity, but with
the right application of AI, itcould also be on the brink of
(54:05):
its greatest transformation.
Yet, thank you.
Reflect on everything we heardtoday, three big lessons stand
out.
First, ai is no longertheoretical.
Whether it's reducingdocumentation time through
ambient listening tools orlowering pediatric stroke risk,
ai is moving from pilot projectsto essential tools in the care
delivery process.
Second, healthcare must leadwith intentionality.
(54:29):
With so many new tools floodingthe space, leaders are pushing
pause, not on innovation, but onuncoordinated adoption.
The name of the game isgovernance, infrastructure
readiness and clear ROI.
And third, the human elementremains irreplaceable.
From patient trust to clinicianadoption, ai only works when
people are in the loop, usingthese tools to augment, not
(54:51):
replace, the irreplaceable humanaspects of care.
If you liked this episode ofthe AI Proving Ground podcast,
please consider leaving a reviewor rating us, and sharing with
friends and colleagues is alwaysappreciated.
This episode of the AI ProvingGround podcast was co-produced
by Mallory Schaffran, naz Bakerand Stephanie Hammond.
Our audio and video engineer isJohn Knobloch and my name is
(55:12):
Brian Felt.
We'll see you next time.
Popular Podcasts
Crime Junkie
Does hearing about a true crime case always leave you scouring the internet for the truth behind the story? Dive into your next mystery with Crime Junkie. Every Monday, join your host Ashley Flowers as she unravels all the details of infamous and underreported true crime cases with her best friend Brit Prawat. From cold cases to missing persons and heroes in our community who seek justice, Crime Junkie is your destination for theories and stories you won’t hear anywhere else. Whether you're a seasoned true crime enthusiast or new to the genre, you'll find yourself on the edge of your seat awaiting a new episode every Monday. If you can never get enough true crime... Congratulations, you’ve found your people. Follow to join a community of Crime Junkies! Crime Junkie is presented by audiochuck Media Company.
24/7 News: The Latest
The latest news in 4 minutes updated every hour, every day.
Stuff You Should Know
If you've ever wanted to know about champagne, satanism, the Stonewall Uprising, chaos theory, LSD, El Nino, true crime and Rosa Parks, then look no further. Josh and Chuck have you covered.