August 20, 2025 • 42 mins
Dr. Jeremy Friese knows medicine from both sides. A practicing radiologist and technology executive, he’s seen firsthand how administrative burden undermines care. In this episode of NEJM AI Grand Rounds, he walks through the origins of prior authorization, explains why he believes artificial intelligence can close the gap between patients and payers, and argues that real reform means showing your work—just like in math class. At Humata, he’s combining human oversight, LLMs, and interoperability to try to fix a broken system. For clinicians overwhelmed by back-office complexity, this conversation offers both urgency and optimism.
Mark as Played
Transcript
Episode Transcript
Available transcripts are automatically generated. Complete accuracy is not guaranteed.
(00:04):
It is unequivocal, no doubt in mymind that the velocity of the huge
immediate opportunity is on theadministrative side of medicine.
It's back-office BS that we're tackling.
It's documentation.
It's the administrative stuff,and that's a massive problem.
Doctors are burnt out.
Patients aren't getting the care theydeserve, like we have to deploy it here.
And that's where the opportunity sitsin the near term. And in the long term,
(00:29):
giving patients access to goodquality care at their fingertips
and being able to do that 24/7,
and ensuring that the right informationand the right care is getting delivered.
Like, that is such a much largeropportunity than we're tackling on
the administrative side, and I firmlybelieve that that will take shape
(00:50):
over the coming years and decades,and I can't wait to be a part of that.
Hi, and welcome to anotherepisode of NEJM AI Grand Rounds.
I'm your co-host Raj Manrai and todaywe are delighted to bring you our
conversation with Dr. Jeremy Friese.
Jeremy is a radiologist andhe's a CEO of Humata Health
(01:12):
where he's working on buildingtechnology for prior authorization.
We had a wide ranging conversation withJeremy digging into both his career as
a physician and now as a leader of acompany. And we had a chance to discuss
the impact of artificial intelligence
technology more broadly in health care.
I learned a lot from this conversation,and as with Dr. Shiv Rao from Abridge
(01:32):
who we had on the podcast a fewepisodes ago, this is a chance to hear
the unique perspective of a clinicianleader of a medical AI company.
The NEJM AI Grand Rounds podcastis brought to you by Microsoft,
Viz.ai, Lyric, and Elevance Health.
We thank them for their support.
(01:56):
And with that, we bring you ourconversation with Dr. Jeremy Friese.
Well,
Jeremy, thank you for joiningus on AI Grand Rounds today.
We're super excited to have you.
Great to be here.
Jeremy,
great to have you on AI Grand Rounds.
So, this is a question thatwe always get started with.
Could you tell us about the trainingprocedure for your own neural network?
(02:16):
How you got interested in AI,and what data and experiences
led you to where you are today?
I anticipated this question and love it.
It's such an interesting one.
I grew up on a farm in South Dakota,and that was really the beginning
of my interest in health care.
And then that evolved into technology.
So, my learnings from working on thefarm, frankly, is just the power of
(02:39):
people and the power of hard work.
And when you put those two thingstogether, really great things can happen.
One of the other things that I sawliterally every single day in my community
is just how people help each other out.
And when you do that from a good place,really impactful things can happen
on an individual level, but also onscale. From there went on to undergrad
(03:01):
as a business major, and thought I wasgonna take over my family businesses.
Not health care related in any way.
And then I saw my mom die at thehands of our health care system.
And she was diagnosed with breast cancerand then died three years later at age 50.
And that was really, nodoctors in my family.
That was the first time I experiencedhealth care, the U.S. health care system.
(03:22):
And I was silly enough to thinkthat I had to go to medical
school to have an impact on that.
So, then I spent the next 10 yearsdoing what you have to do to do
that, and I was good at taking tests.
So, I got into the places that Iwanted to get into, ended up going
to Mayo Clinic for medical school,and then I stayed there for 20 years.
And my time at Mayo was
impactful in so many ways.
I knew that when I went intomedical school, I wanted to do
(03:45):
something on a systemic level.
What I didn't anticipate is howmuch I'd love the operating room.
And a few key things that reallystood out for my Mayo time is
the power of company culture andhow when you can align a whole group
of people around a mission, youcan really do spectacular things.
(04:06):
And the Mayo culture is renowned,truly does put the patient first
at everything that they do.
And I was just really fortunateto be a part of that for a couple
of decades, and that really had animpact on how I think about myself as
a technologist and company builder.
I remember a story that our CEO told atone of our grand rounds of he was coming
(04:26):
into the operating room one morning andthe janitor was cleaning the walls and
was just doing it really diligently.
And he said, I see you every singlemorning doing this with such passion.
And his response was, well, of course,because it's my job to take care of the
patient and by cleaning these walls,I'm keeping them from getting sick.
And it really solidified for mehow everybody in the organization,
(04:48):
when you're working on somethingand moving in the same direction,
you can have really huge impact.
And that that's had an impact on theway I think about building businesses
and solving problems in health care.
And then also it helped run ourlarge radiology department and
nobody teaches you about the rev cycle inmedical school or in training in any way.
And it's frankly not a very seemingly, verysexy topic for physicians and caregivers.
(05:12):
But frankly, what I learned there isno money, no mission. And even at a
very well-resourced place like MayoClinic, if you don't run the business
of health care you don't earn theopportunity to take care of patients.
And so that was really impactful for me.
I'm an interventional radiologist, and soI chose radiology because, A, it impacts
virtually every patient that toucheshealth care and B, it's where the most
(05:35):
cutting-edge technology was taking place
when I was going through training. Andit was really clear to me through seeing
those innovations, how this was the futureof medicine, and both imaging, diagnostics,
as well as therapy, that was really thetouchstone for the combination of using
technology to solve patient problems.
And by allowing doctors to be doctors,and nurses to be nurses, by solving this
(06:00):
sort of back-office BS problem, we gaveour physicians and nurses the opportunity
to actually take care of patients.
And so, I'd say those things combinedwere, are the reason that right now I'm
trying to solve this silly problem ofprior authorization using technology.
Yeah.
So, what strikes me is that, I think thisis a recurring theme for so many of our
(06:21):
guests and these conversations that wehave on AI Grand Rounds, that there's a
personal experience with either you ora loved one that sort of is the genesis
of your interest in going into medicineand really having impact in health care.
And so, thank you for sharing that.
And thank you for also just kind ofwalking us through your career arc there.
(06:41):
We're gonna dig into what you're doingat Humata in a few moments, but I
think, maybe, if we can just spend alittle bit more time on your career as
a practicing interventional radiologist.
You got your start at Mayo Clinic afteryour formal training, and I think a lot of
our listeners are always very interestedin what was going through your mind in
(07:01):
choosing a place to start your careerafter that formal training and how you
prioritize different aspects of the job.
What kind of led you to, youcould just spend a couple more minutes,
what led you to, to Mayo and whatwas appealing about starting your
career after formal training there?
The initial answer is, I chose Mayoinstead of several of the East Coast
(07:22):
folks out of the gate because I knew mymom was gonna die during medical school.
Mm-hmm.
And you know, Mayo's a wonderfulplace, but it was for family reasons.
And I think that's a really importantpoint to note that there's a lot
of really wonderful places that youcan work and practice and train.
And because we see this as a calling,sometimes we don't think about taking
care of ourselves and our familyas we're making these decisions.
(07:44):
And frankly,
I would've got phenomenal trainingif I would've went East Coast, West
Coast, or somewhere else in the middle.
But that family decision was also areally big part of it for me, and I think
should be for everybody to some degree.
I had a wonderful time there inmedical school, learned a ton.
Was then there for surgery,internship, as well as residency.
(08:05):
And then spent a couple years in Bostonas a fellow and going to business school.
And so, my decision to go back to Mayo
was in part because Mayo has thisreally wonderful opportunity where
they would say (08:16):
we identify you as someone that we want to have on staff.
Now go to Boston, or go to Paris, or goto San Francisco and learn how to be the
best surgeon of the left pinky or whateveryour thing is and then come back.
And so, that was an opportunitythat I took advantage of.
And then I stayed there for adecade. And so, had the opportunity
to both work at Mayo, but then alsowork at Brigman Women's in Boston.
(08:37):
And so, the reason to come back wasbecause of that initial opportunity.
And the reason I stayed for so long isbecause it was just so many different
opportunities to do different things.
And I think one of the thingsthat academics allows you to do
is to scratch several itches.
And so, not only can you practicemedicine and do an extremely
high level. And even as a youngperson recently out of training,
(08:59):
people are flying from all overthe world to come get care at your
institution. And you really hone your craftand get to be one of the world's best,
which is a pretty special opportunity.
And so, it was that opportunity that keptme there and the opportunity to say,
okay, I know that I don't wanna practice
full-time because I have theseother things that I'm interested in.
(09:20):
I'm interested in technology.I'm interested in investing.
I'm interested in the administrative side,the business side of medicine, and
you get that opportunity in academics.
And so that was really thekey driving factors for me,
Raj.
Yeah.
Maybe before we go to what you're up tonow, you practiced for several decades.
Are you still practicing right now?
(09:41):
I loved interventional radiology.
Unfortunately, it's not oneof those specialties— Right?
—that you can do very well part-time.
You can't dabble.
Yeah,
you can't dabble.
And so, I've had to give it up, nowthat I've shifted full-time to the
technologist, entrepreneur side, butI can't wait for the opportunity to
get back when that opportunity arises.
Right.
So, maybe you could just reflectfor a few minutes here on how your
(10:06):
practice changed as while youwere practicing as an interventional
radiologist over the last few decades.
'Cause you saw this veryinteresting boom of
AI right around the same time,specifically AI for imaging.
And we were hearing sometimesthese hyperbolic quotes around
(10:26):
"we no longer need to trainradiologists" and things like this.
And during this time, you're practicing,you're honing your skills, right?
You're working with a lot of trainees.
Surely, you're getting questionsfrom residents, fellows in training.
Like, wow, is AI gonna automatewhat I'm doing in the next
five, 10 years?
And so, from that perspective you'rehearing all of this chatter, right?
(10:47):
But then you're actually practicing.
And what I hear from a lot of doctorsis things haven't really changed that
much over the same timescale, eventhough the research has really advanced.
And so, I'm curious if that resonatesat all, or if you could describe what
has changed over those couple decades.
Absolutely.
Yes.
And it's also super interesting becausewe're obviously a training program. And
(11:08):
so, at Mayo, where you normally have awhole bunch of applicants, but also
in radiology, which is one of themost competitive, to see that ebb and
flow of the applicants. And frankly,how that ebb and flow corresponded
with is AI gonna take all the jobs?
No, it's not.
Is AI gonna take all the jobs?
No, it's not.
And frankly, saw thatcycle a couple of times.
I'm a firm believer that all of thesetechnologies that we have today, that
(11:31):
we had over the last decade, and thatwe're gonna get over the next decade,
are going to augment humans, augmentclinicians, and augment patients to
only do all of these things better.
And as we've continued to see, especiallyin imaging, as the technology evolves,
we're not seeing fewer images, we'reseeing dramatically more images.
(11:54):
And even with technology, even withartificial intelligence, to augment
even, maybe take over some of those,
I don't see a world in the nearfuture where you're gonna need
fewer radiologists in my mind.
Radiology imaging in general, whetherit's radiology or cardiology, et cetera,
there's only increasing need for humansto sit on top of, or beneath the technology
(12:19):
to really help take care of patients.
And frankly, the best radiologists arenot just technologists, they're also
clinicians, and they're the ones thatgo beyond the sort of pixels and both
have conversations and consultation withradiologists or the doctor's doctor.
So, how do you also then interpretthis for the gastroenterologists
(12:40):
or whatever specialty?
And, and that role isn'tgoing away anytime soon.
Cool, thanks.
Maybe I wanna follow up on that beforewe transition to your work at Humata.
So the reference Raj was alluding to,there was this Jeff Hinton quote in 2016.
Jeff Hinton, Nobel Laureate touringaward winner, godfather of AI, said,
(13:01):
it's just completely obvious now thatwe should stop training radiologists.
It's like the coyote who's gone overthe cliff but hasn't looked down yet.
That didn't come to pass.
And I'd love to understand a littlebit more about why it didn't come to pass,
because clearly like vision capabilitiesfor AI have gotten very, very good.
Lots and lots of studies show thatat least in controlled settings,
(13:22):
they're competitive with radiologists,so there's a capability there.
On the flip side though, there's thisthing called Jevon's paradox, which is,
as a tool makes it moreefficient to use a resource,
the demand for thatresource goes up, not down.
So, I don't know if that's what's going onhere, but, like, I also have had experience
with radiologists where I ask them ifAI is replacing them and they'll say no.
(13:44):
And they'll be like, but you know what?
We actually didn't hirea radiologist last year.
We actually can do more withthe staff that we have now.
So, I don't know exactlyhow to square the circle.
Is there like this
silent replacement where you just needa few lower head count at a practice.
Is that wrong?
Yeah.
I'd love your thoughts on, inwhat way was Jeff Hinton wrong?
I think clearly with hindsight now,he was wrong, but, like, it's not
(14:06):
clear to me in what way he was wrong.
He was wrong looking backward over sincethat time he made the comment, it's
also not gonna happen over the sametimeframe going forward, in my opinion.
And it stems from a bunchof different reasons.
One of them is just
the pace with which we adopt thistechnology in health care is maybe not at
(14:26):
the same level as other industries, andthere's some very good reasons for that.
And there are things that computers justcan't do. And that's stemming back to
my own neural network of the need forhumans in the delivery of care, and the
explanation of care, and the ability to
decipher and understandall of these things.
(14:50):
So, I think that's anotherreally critical component of
the reason why that's the case.
And then the other is just thetechnology continues to advance on
how imaging and how other diagnosticsare being used, that there's always
gonna be this sort of catch up
of how the interpretivetechnology is gonna be used
versus the cutting-edge stuff.
And again, I don't see that slowingdown in health care anytime soon.
(15:13):
If anything, it only acceleratesto come back to your point of does
an individual practice not hiresomeone for the year versus next
year because they're more efficient.
There's no doubt.
Radiologists are dramatically moreefficient today than they were five
years ago, 10 years ago, 20 years ago.
And we continue to see a dramatic need andshortage for radiologists in particular.
(15:36):
But the same thing is true across otherspecialists that are being impacted by
technology, and it's because the need forthese specialties only continues to grow.
And
maybe one more before we move on.
So, if the radiologists are moreproductive and there's increased demand
for radiologists, is there more vol?
Like, where is this additionalvolume of work coming from?
(15:57):
Or do more people have accessto radiological services?
Or if each worker is more productiveand you need more workers, there
has to be an increased volumeof work that is being demanded.
I mean, you're absolutely right.
So, it's a combination of there are morestudies being done, there's also a whole
lot more data in each of those studies,and so the answer is both of those.
(16:19):
And so, there's just awhole lot more images.
There's also a whole lot more studies,and we continue to see that grow.
Three to 5% per year.
And again, imaging is central toalmost everything in medicine.
And there's maybe bad jokes or badwhatever it is, that there's no more
physical exam, it's all go get a CT scan.
And that is in part the way we're trainingour doctors today is imaging is central.
(16:46):
It will continue to be central and onlyincreasingly so over the coming years.
Cool.
Thanks.
Um. I might be biased, but yeah.
Yeah.
It's a radiographic. I think thefacts are true, uh, centric, you know.
I think, I think probably true.
Um, so now I wanna talkabout your work at Humata.
And I'm gonna first tryand state succinctly,
like, what you guys are working on. Andthen maybe break down each one of the
(17:09):
mission statement of Humata down so thatwe can, because I myself am not wonky
on the health care administration side.
So, I think for my own, like, education,I'd love to walk through some of the
pieces here that you're working on.
So, if I was gonna state itsuccinctly, it would be using
AI to solve prior authorization.
So, feel free to hop in thereand editorially edit any of
that mission statement that mayor may not have gotten wrong.
(17:32):
Yes and is what I would say.
I would say we're on a mission to solveprior authorization for patients, and
that means that to actually be ableto do that, you have to solve it on
both sides of the fax machine for bothproviders and payers because there's
a problem on both sides, frankly.
And if patients are gonna get ayes, you need to have an efficient
submission and you need to have anefficient decision on the other side.
(17:55):
And then on the first partof your comment, we're an AI
company solving prior auth.
I would say we're a technologycompany solving prior auth, and that
includes the tools of artificialintelligence that also includes
a bunch of other tools.
And so, one thing that I feel verystrongly about as a problem solver
is you need to start with theproblem, not with the technology.
(18:16):
And you need to use all the toolsat your disposal to solve it.
And those tools will evolve asthey do, and you'll hone and refine
what those tools look like tospecifically solve your problem.
So, that's a long-winded way to saywe are a software company solving
prior authorization for patients.
Got it.
Cool.
(18:36):
Okay, so super helpful.
Now I'm gonna ask you toexplain this like I'm five.
What is prior authorization?
This is a, a phrase that I think lotsof well-educated people hear a lot, and
we understand what the two English wordsmean together and have some vague sense
of what it is, but what is prior auth?
Why does it exist?
And set up the problem for us.
(18:57):
So, I would see a patient.
And decide in my medical opinion,that they need to have a surgery.
Before I could go do that surgery, I wouldhave to contact a payer and an insurance
company to say, Dr. Friese believes thatAndy should have this procedure, CPT code
(19:20):
77450, whatever the number is.
And then you have to prove why you needto have that procedure, and then the
insurance company will take it backinto the bowels of the insurance company
and come back to you with a decision.
Now, when I was practicing full-time, theyhad 30 days to give you that decision.
Now, some of the things that you'reseeing, they have to give you
(19:41):
an answer faster, but the shortanswer is I asked for approval
to do that procedure, and they say,yes, you can do that surgery for
Andy given his health insuranceplan and we'll pay for it.
And so, while it sounds like a reallysimple, short thing you can say, the
short answer is, I have to get approvalbefore I deliver surgery, before I do
(20:04):
an imaging report, before I prescribe adrug to confirm that it will be paid for
by the insurance company.
And this is, I think, simplified, likea piece of friction in the system.
Does the piece of frictionexist to reduce medical waste?
To reduce fraud?
Like, what is this policy, which introducesa friction point in the health care
(20:25):
system? What's it designed to be doing?
Yeah, it, I mean, it's been aroundfor decades, so this like, while it's
kind of become a sexy topic lately,it's been around for decades and
the reason it was put in place is,
A, to stop fraud, as you've said, butthat's only a small portion of it.
The other piece is all of these healthplans are designed uniquely, and so
(20:47):
in one plan, it might be covered. Inanother plan, it might not be covered.
And a way to help control costsof the individual plans is to say
in this plan, you need to get approval.
In this plan, you don't.
But it really is a measure to helpcontrol costs and not just let doctors
(21:09):
do whatever they feel is right.
And presumably there's a patientprotection aspect of this, too,
where a patient doesn't get stuckwith a bill for procedure that
they actually weren't covered for.
There's probably a component of that, too.
Nailed it.
Absolutely
right.
The, the other component of that is that
we do a lot of things in health carethat may not be necessary.
(21:29):
And so, there's a whole lot of thingsthat are done off typical care paradigms.
And so, if you're a glass half fullperson, you would say this is to help
make sure that patients are gettingcare along those care paradigms.
If you're a glass half empty,you know, you would say it's
because the insurance company istrying to inflict friction so that
they don't have to pay for stuff.
And the answer is, it's
(21:50):
somewhere in the middle.
Got
it.
Cool.
So, I feel like I understand priorauth now, so, so thanks for that.
So, based on your framing, I am guessingthat this is not an easy if then else
kind of authorization procedure, that it'svery complicated and context dependent.
So, what are you guys building from atechnological and AI perspective to
(22:11):
make this check to be lower friction?
You're absolutely right.
It's a very complex problem.
There are multiple steps through it that Iwon't bore everybody with, but the crux of
the problem is every single one of theseplans has a different set of rules, a
different set of policies that need to be
met in the clinical record
in order to show that for thisparticular plan, for this particular
(22:35):
patient, you fall in the guidelinesof what would be expected.
So, A, these rules aren't necessarilyeasily known by the providers.
And then B, they're changing a lot.
And at least in our experience,we see that these policies
change upwards of 8% per month.
And so, imagine being a doctor
(22:55):
doing a procedure. The rulesof the game for one particular
plan are changing that often.
That's just not somethinga human can keep track of.
And so, why artificial intelligence andtechnology more broadly is so intimately
perfect for this particular problem isbecause computers are just better at
understanding that huge body of knowledge.
(23:17):
And then
sifting through the medical informationto help you build better submissions.
But here's the other problem, you guys.
I said this is a two sidesof the fax machine problem.
You've got literally armies ofhumans sitting on the payer side.
They're doing the exact same work.
They're understanding these medicalpolicies, they're understanding
the clinicals that get sent over tothem, and literally reading through
(23:39):
a hundred-page document of clinicalsto try to say yes or no, this fits.
That's exactly whatcomputers are better at.
And can do dramaticallymore efficient than humans.
Great, thanks.
So, maybe, if you could talk abouthow interventional radiology has
changed over the past two decades.
How has this prior auth problemchanged in the wake of AI?
(24:02):
So, I can imagine if you go back 10 years,
there's a different set of technologicalsolutions that you might have in 2025.
Is this just an LLM problem now?
Do LLMs hallucinate and thereforeproof things like what are the,
what's the technological state-of-the-art look like
in 2025 for this? Yeah, there.
So, to really solve this problem, thereare really two pieces of the puzzle.
And the first
(24:23):
is a less sexy problem thanusing artificial intelligence.
The first piece is interoperability,and so how do you get providers and
payers talking to each other througha standard set of pipes and rules that
you can actually share information?
And so, what's different, let'scall it this year from even
five years ago, is there's nowa huge push from CMS and others.
(24:44):
Non-CMS bodies to say, let's try tostandardize some of the pipes so we can
share information back and forth, notonly for prior auth but for other things.
And so, we're seeing a significant tailwindfor those pieces starting to take shape.
And, specifically, CMS has a mandate
around that.
By 2027, it continues to get bumpeda little bit, but that's progress.
So, huge progress on theinteroperability front.
(25:06):
Still a lot more work to be done,but really excited about that.
And then the second piece is aroundthe sexy technology side of the house.
And absolutely what we were able todo five years ago to understand a
medical record and match it up withmedical policies was a whole lot
more brute force than it is today.
And LLMs have allowed usto synthesize, summarize,
(25:30):
and do that for both medical policiesas well as for clinical information. And
then match those two up in ways that weresort of unimaginable even two years ago.
And so, our ability to get to the rightclinical information and get that
over to the payer is dramaticallyimproved because of these technologies.
And then the same thingis true on the payer side.
Our ability to understand andsynthesize these things today and
(25:52):
make that process faster for payersis exactly what CMS is pushing for
and Dr. Oz is pushing forward.
I mean, I just feel like we're reallyat a interesting time in the industry
that because of both the interest andappetite from the industry to solve
this problem, but also because thetechnology is at a place where we can
really do this in a meaningful way.
It's pretty spectacular.
(26:13):
Yeah, it's a great transition tothe question that I wanted to ask,
which was a topic you just mentioneda moment ago, large language models.
And so, I feel like we, this is actually,I don't know, we're 20-something
minutes into the conversation, and thismight be a record, before ChatGPT was
brought up, or large language modelswas brought up, just 'cause still such a
dominant topic on these conversations.
(26:34):
I think the less eloquent version,but more direct version of my question
is (26:38):
how have large language models
completely changed your pipelines orthe way you were doing things before
a couple years ago when ChatGPT andnow all of its cousins were introduced?
Yes and no.
And so, as I started with, there'sa whole bunch of other pieces of
this problem of understanding isa prior auth needed, yes or no?
(27:00):
What's the status ofthe prior authorization?
And those other questionsare not necessarily large
language model questions.
They're questions that you have.
That's exactly
what
other systems work reasonably well.
Or maybe even withhigher fidelity than LLM.
Like, you don't need tothrow an LLM at everything.
You just nailed it.
That's what I was trying to say.
The other pieces of the problemreally haven't been changed by LLMs.
(27:21):
The key problem that we've now been ableto dramatically improve our speed and
ability to tackle is the ability to useLLMs, understand the medical policies.
Match that up with the clinicals so thatyou can use that technology to both submit
better clinicals and make better decisionson the payer side. Like, that has been
(27:42):
a complete game changer, which is onereally critical, important piece of the
puzzle, but it is one piece of the puzzle.
And as I think about otherfolks that are in the space.
You can't just throw an LLMat this problem and solve it.
It's part of the puzzle. An importantpart of the puzzle that's changed the way
we've handled some of the clinical pieces.
But it is one important,uh, tool in our toolbox.
(28:02):
Where you use LLMs,
you know, we hear a lot about
hallucinations.
And I think early on we were, I thinkeveryone was, you know, you saw those
New York Times stories with a lawyerwho used ChatGPT and it just fabricated
case law and it did so very confidently.
Right.
And we're hearing sort of mixed reports.
Now we have the newreasoning models, right?
(28:25):
The O series from OpenAI and then theequivalents from Google and others.
And so, we're hearing mixedreports about the sort of
rate of hallucinations.
I think a lot of folks saythey're going down in general.
Others are saying there's still a problem.
Other reports altogethersaying they're going up.
Right?
And with some of the new advanced modelsthat are more capable, but also still
(28:45):
hallucinating, potentially even morethan some of the early incarnations.
And so, I imagine, this is a question thatI think comes up with any of the companies
that are using LLMs, but I think also aswe're thinking about outside of companies,
just physicians using products, right?
That they're using now,
in practice and I think atvery large scale already.
How do you think about hallucinations?
(29:07):
How do you guard against them fromyour perspective where you use LLMs?
Is this a big concern andwhat can we do about it?
Lemme step back, lemmegive you an analogy.
I've got teenage kids.
When they're in their math class,they have to show their work.
They can't just show the final output.
I've always believed in usingartificial intelligence,
you need to show your work.
(29:29):
And so, that's exactly the sameapproach we're taking with LLMs.
So, as an example, we will use an LLMto summarize and to build questions.
We will always submit the source work.
And so as you're submitting something to apayer, you're not just submitting the LLM
output, you're submitting the LLM output
and the actual doctor's notes.
(29:52):
'Cause frankly, that source of truth isthe only thing that is being considered
on the flip side of the fax machine.
And so,
but yeah, what if the doctor'snote was written by an LLM?
So, even in that scenario,they have to approve it.
So, we are making the assumptionthat what is in the medical record,
whether it's written by an LLM ornot, there's still a human oversight
that says yes, this is the fact ofwhat was discussed with the patient
(30:17):
and it's true. If there's a fallacyin the medical record, I, that would
give me great consternation, butI trust that Shiv and the Abridge
team and the like in Microsoft andthe Ambient and the other folks are
making sure that, that those sort ofhallucinations are not showing up there.
And we're saying we're Switzerland.
(30:37):
What's documented in the medical recordis fact, and we're gonna use that and
get that over and, and show our work.
Cool.
Thanks.
The other thing I'll add there, thesame thing is true on the payer side.
So, under no circumstance willour computer ever be used and
say no for care, like, because ofhallucinations. And frankly, because
(30:57):
I think this is a human endeavor.
A human, a trained physician, or nurse, orsomebody, needs to be the one that says:
I'm sorry, Andy.
I'm sorry, Raj.
You are not approved for this carebecause of A, B, and C. And that
needs to be a human, not a computer.
The computers can help the human get toan answer, but you need to show the work
of where in the medical record is thatdocumented to answer this question or
(31:20):
where there's contradictory information.
And again, we use LLMs to showwhere that information is, but the
summary cannot be the final answer.
Cool.
Maybe one more question on thisbefore we move to the lightning round.
So, you had this great way offraming the area of health care
that you operate in, and it's likeon both sides of the fax machine.
(31:40):
And I think that that, like, really nicelyhighlights for me— Like, it might be the
title of this episode on both sidesof the—. Yeah, on both sides
of the fax machine.
It nicely highlights how like there arethese different technological strata in
the health care system and you're workingin the one that happened to be ossified
around 1985 or something like that.
Uh, and when fax machines werestill like cutting-edge technology.
(32:01):
But I wanted to talk a littlebit about the business model. Because
I think one of the things that haschanged a lot in health care over the
last five years is there a very viableand lucrative business models for
technology companies. Like it used to be
impossible as a startup in the health carespace to find a viable business model.
I think you mentioned Abridge,they've obviously unlocked one.
Some other health carestartups are doing it too.
(32:22):
I think you guys are onto it.
So, which side of the fax machine do youconsider to be your primary customer?
Is it the doctor to help get stuffapproved, or is it the payer to
help them deny unnecessary care?
Or is it both?
I like to say we're inthe business of yes.
And so, I want to help doctors get to yes,
(32:42):
so that patients can getthe care they deserve.
And I want payers to get to yes
more efficiently, so that they cando that more efficiently and cheaper
than what they're currently doing
and get the patient to yes.
I mean, our business model is wehelp both providers and payers.
And both of them have an ROI.
To do that using technology justlike in the ambient dictation space.
(33:05):
I think prior auth is another area whereboth health systems and payers are saying,
we cannot continue to do this with humans.
It's just silliness and they'readopting artificial intelligence and
our software, and broadly at a ratewith which I haven't seen in the past.
And it's a pretty excitingthing to experience.
Cool.
Awesome.
(33:25):
Thanks Jeremy.
So, I think if it's okay with you, we'regonna move to the lightning round.
Let's do it.
So, lightning ground, it sounds likeyou've listened to the show before.
The goal are sort of succinctanswers to a grab bag of questions.
You can decide which ones areserious, which ones are non-serious,
(33:47):
and we'll just hop into it.
Let's do it.
So, the first one and thephrasing of this one matters:
In what ways will AI change medicinethe least over the next five years?
Doctors
seeing patients and it beinga truly human endeavor.
Oh, interesting.
I might come back tothat in big picture, but
(34:07):
we'll, we'll keep it moving.
Alright, Jeremy. This, this is thesecond lightning round question.
If you weren't in medicine,what job would you be doing?
I'd be a professor.
A professor of...? Business.
Nice.
I think.
Sorry, I know this is lightning round, but you,
you got me going.
(34:27):
You know, I think there are severalthings in this world that can have
dramatic impact at scale and Ithink capitalism deployed in the right
way can have a huge impact on humans.
And so that's, uh, that'show I see the world.
Alright.
If you could have dinner with oneperson, dead or alive, who would it be?
Teddy Roosevelt.
Oh, nice.
(34:48):
Probably some bourbon.
Bourbon would be had.
Amen
to that.
First modern president, developedthe national parks, interesting chap.
Nice.
Do you think things createdby AI can be considered
art?
absolutely.
Alright.
Last one.
Uh, you, you touched on this, at thebeginning, but we're gonna revisit it now.
(35:10):
Should medicine be considered just a jobor should it be considered a calling?
I think that that's anartificially challenging question.
I think the answer is yes.
There are components of it where it's ajob and you have a, you're a human outside
of this, and if you're going to actuallybe able to practice it as a calling, you
(35:30):
also gotta take care of yourself outsideof that, meaning that you have to be able
to partition it and seeing it as a job.
Cool.
Thanks.
Well, Jeremy, you survived.
Yeah. I was gonna say
in the business of yes.
Yes, that's
right.
Well, Jeremy, you survivedthe lightning round.
Nicely done.
Thank you.
Thank you.
Alright Jeremy, so we just havea few last sort of big picture
(35:52):
concluding questions for you.
And I think we've touched onboth of these various points
of the conversation already.
But you are a rare breed of physician.
Executive leader of a technology company.
And to be honest, I wishthere were more, right?
We had Shiv on the podcast a coupleepisodes ago, and there are a few other
(36:14):
great examples, but I think there reallyis something different in my mind about
a company that's being led by a physicianworking on problems in health care.
And so, you really understand
what doctors do, how theywork as part of care teams, and
what patients are looking for.
And I think it, it shapes a lotof the way you design the company
and your mission. With that setup,
this is a bit of a sort of interestingquestion to ask then, uh, but will AI and
(36:38):
medicine be driven in your opinion moreby computer scientists or by clinicians?
I guess neither is also an answer.
So, you know, one of the things thatreally impacted my neural net and the
way I see the world, both care delivery,but also building technology businesses.
Again, at my days at Mayo,every leadership post had two people
(37:04):
that were equally yoked to tacklewhatever the leadership position was.
It was a physician and administrator.
And when I think about
building technology businesses
and the answer to your question,I don't think that you can expect
a physician to have the same levelof technical skill as someone that
(37:26):
lives and breathes it every day.
And I don't think that a technologist
can truly understand the ethos of theproblem of health care and the delivery
of taking care of a patient as well assomeone that has put hands on a patient
and done surgery or done whatever.
And so, it really needs to be amarriage of the two, and the companies
that will have the biggest impact
are the ones that do that the best.And I think that same thing is true
(37:48):
when you look at care delivery.
The ones that do that thebest are the ones that
keep the mission aligned around theright thing and then build solutions
that really solve that problem.
So, I, I guess I'm in the business ofyes, I think the answer has gotta be both.
And the companies that havethe biggest impact are gonna be
the ones that do that the best.
Cool.
I maybe want to ask a slightlydifferent variation on that question
(38:13):
about AI versus human-led medicine.
So, one of the big themes we explore onthe show is obviously AI and medicine.
We tend to have this artificial
notion of what that looks like, which islike a doctor-in-a-box kind of construct.
But really what we've learned throughmany episodes on this podcast, including
this conversation, is that actually AI isalready mediating a lot of interactions
(38:34):
in the health care system already.
Going back to Abridge, going back towhat you guys are doing, going back
to people who talk with ChatGPTbefore they even see their doctor.
So how
far do you think this sort of AIexpansion goes? Are there certain
interactions in the health care systemthat you think in principle will
never be mediated primarily by AI?
Is there a human carve out that you'dlike to preregister here that you
(38:56):
think that is just like not able tobe serviced by AI or do you think we
have essentially full self-drivingmedicine in the fullness of time?
That's a big question.
I know, but we're to thebig picture section, so.
There's no doubt in my mind therewill be components of medicine.
The world I love,
I love the analogy of fullself-driving AI medicine.
You're already seeing components ofthat today, and there's no doubt that
(39:20):
will only take shape to a greater degreeover the coming years and decades. And
humans, since the onset of Google, havebeen using technology to try to augment
their interaction with physicians
and other care providers. I think thatonly becomes greater and maybe some of the
(39:41):
simple stuff continues to then become moreAI-led or computer-led versus human-led.
And frankly, it should, like, we'vegot, it's difficult to get access,
it's difficult to get information.
It's difficult to get accessto, to me and other physicians.
So, you should be ableto get these answers.
In my former life, was annoying whensome, someone asked Dr. Google a
question before they would ask me aquestion because they would usually
(40:03):
go, like down to the third page andthen think that was the gospel truth.
It is only gonna become more.There's no doubt that AI will carve
out pieces of the puzzle and humanswill not need to be involved.
And I welcome that opportunity.
I think every physician and nurse shouldwelcome that opportunity. And there
will be parts of the doctor-patientinteraction that are, are and need to
(40:26):
remain deeply human. And those connections
maybe only become more importantas technology plays a role in
the other pieces of diagnosisand treatment to allow the human
caregiver to be a human caregiver.
Got it.
Cool.
Maybe one question before we wrap here.
(40:47):
You've had a very impressive career asan interventional radiologist, like doing
the practice of medicine, seeing patients.
You're now the CEO
of a successful health care startup.
So, you've done lots of work in diagnosticmedicine, the practice of medicine,
and now on the administrative side.
So, could you give us a sense of which areaadministration or, like, medical practice
(41:10):
has the more, higher potential for AI?
I have a weakly held opinion, butI'd love to hear your thoughts
which side of the house you thinkactually has the most potential.
Two answers to the question.
The first is on the velocity andwhere you're gonna see the impact
first, and then the, the second comesto what the bigger opportunity is.
It's unequivocal, no doubt in mymind that the velocity of the huge
(41:33):
immediate opportunity is on theadministrative side of medicine.
It's back-office BS that we'retackling. It's documentation.
It's the administrative stuff,and that's a massive problem.
Doctors are burnt out.
Patients aren't getting the care theydeserve. Like, we have to deploy it here,
and that's where the opportunity sitsin the near term. And in the long term,
(41:54):
giving patients access to good qualitycare at their fingertips, and being able
to do that 24/7. And ensuring that theright information and the right care
is getting delivered like that is sucha much larger opportunity than we're
tackling on the administrative side.
(42:15):
And I firmly believe that that will takeshape over the coming years and decades,
and I can't wait to be a part of that.
Cool.
Awesome.
Well, Jeremy, this has beena super fun conversation.
Uh, thanks for taking thetime to sit down with us.
Great
to spend time with you guys.
Thank you.
Thanks so much, Jeremy.
That was great.
This copyrighted podcast from theMassachusetts Medical Society may
(42:37):
not be reproduced, distributed,or used for commercial purposes
without prior written permission ofthe Massachusetts Medical Society.
For information on reusing NEJM Grouppodcasts, please visit the permissions
and licensing page at the NEJM website.
It is unequivocal, no doubt in mymind that the velocity of the huge
immediate opportunity is on theadministrative side of medicine.
It's back-office BS that we're tackling.
It's documentation.
It's the administrative stuff,and that's a massive problem.
Doctors are burnt out.
Patients aren't getting the care theydeserve, like we have to deploy it here.
And that's where the opportunity sitsin the near term. And in the long term,
(00:29):
giving patients access to goodquality care at their fingertips
and being able to do that 24/7,
and ensuring that the right informationand the right care is getting delivered.
Like, that is such a much largeropportunity than we're tackling on
the administrative side, and I firmlybelieve that that will take shape
(00:50):
over the coming years and decades,and I can't wait to be a part of that.
Hi, and welcome to anotherepisode of NEJM AI Grand Rounds.
I'm your co-host Raj Manrai and todaywe are delighted to bring you our
conversation with Dr. Jeremy Friese.
Jeremy is a radiologist andhe's a CEO of Humata Health
(01:12):
where he's working on buildingtechnology for prior authorization.
We had a wide ranging conversation withJeremy digging into both his career as
a physician and now as a leader of acompany. And we had a chance to discuss
the impact of artificial intelligence
technology more broadly in health care.
I learned a lot from this conversation,and as with Dr. Shiv Rao from Abridge
(01:32):
who we had on the podcast a fewepisodes ago, this is a chance to hear
the unique perspective of a clinicianleader of a medical AI company.
The NEJM AI Grand Rounds podcastis brought to you by Microsoft,
Viz.ai, Lyric, and Elevance Health.
We thank them for their support.
(01:56):
And with that, we bring you ourconversation with Dr. Jeremy Friese.
Well,
Jeremy, thank you for joiningus on AI Grand Rounds today.
We're super excited to have you.
Great to be here.
Jeremy,
great to have you on AI Grand Rounds.
So, this is a question thatwe always get started with.
Could you tell us about the trainingprocedure for your own neural network?
(02:16):
How you got interested in AI,and what data and experiences
led you to where you are today?
I anticipated this question and love it.
It's such an interesting one.
I grew up on a farm in South Dakota,and that was really the beginning
of my interest in health care.
And then that evolved into technology.
So, my learnings from working on thefarm, frankly, is just the power of
(02:39):
people and the power of hard work.
And when you put those two thingstogether, really great things can happen.
One of the other things that I sawliterally every single day in my community
is just how people help each other out.
And when you do that from a good place,really impactful things can happen
on an individual level, but also onscale. From there went on to undergrad
(03:01):
as a business major, and thought I wasgonna take over my family businesses.
Not health care related in any way.
And then I saw my mom die at thehands of our health care system.
And she was diagnosed with breast cancerand then died three years later at age 50.
And that was really, nodoctors in my family.
That was the first time I experiencedhealth care, the U.S. health care system.
(03:22):
And I was silly enough to thinkthat I had to go to medical
school to have an impact on that.
So, then I spent the next 10 yearsdoing what you have to do to do
that, and I was good at taking tests.
So, I got into the places that Iwanted to get into, ended up going
to Mayo Clinic for medical school,and then I stayed there for 20 years.
And my time at Mayo was
impactful in so many ways.
I knew that when I went intomedical school, I wanted to do
(03:45):
something on a systemic level.
What I didn't anticipate is howmuch I'd love the operating room.
And a few key things that reallystood out for my Mayo time is
the power of company culture andhow when you can align a whole group
of people around a mission, youcan really do spectacular things.
(04:06):
And the Mayo culture is renowned,truly does put the patient first
at everything that they do.
And I was just really fortunateto be a part of that for a couple
of decades, and that really had animpact on how I think about myself as
a technologist and company builder.
I remember a story that our CEO told atone of our grand rounds of he was coming
(04:26):
into the operating room one morning andthe janitor was cleaning the walls and
was just doing it really diligently.
And he said, I see you every singlemorning doing this with such passion.
And his response was, well, of course,because it's my job to take care of the
patient and by cleaning these walls,I'm keeping them from getting sick.
And it really solidified for mehow everybody in the organization,
(04:48):
when you're working on somethingand moving in the same direction,
you can have really huge impact.
And that that's had an impact on theway I think about building businesses
and solving problems in health care.
And then also it helped run ourlarge radiology department and
nobody teaches you about the rev cycle inmedical school or in training in any way.
And it's frankly not a very seemingly, verysexy topic for physicians and caregivers.
(05:12):
But frankly, what I learned there isno money, no mission. And even at a
very well-resourced place like MayoClinic, if you don't run the business
of health care you don't earn theopportunity to take care of patients.
And so that was really impactful for me.
I'm an interventional radiologist, and soI chose radiology because, A, it impacts
virtually every patient that toucheshealth care and B, it's where the most
(05:35):
cutting-edge technology was taking place
when I was going through training. Andit was really clear to me through seeing
those innovations, how this was the futureof medicine, and both imaging, diagnostics,
as well as therapy, that was really thetouchstone for the combination of using
technology to solve patient problems.
And by allowing doctors to be doctors,and nurses to be nurses, by solving this
(06:00):
sort of back-office BS problem, we gaveour physicians and nurses the opportunity
to actually take care of patients.
And so, I'd say those things combinedwere, are the reason that right now I'm
trying to solve this silly problem ofprior authorization using technology.
Yeah.
So, what strikes me is that, I think thisis a recurring theme for so many of our
(06:21):
guests and these conversations that wehave on AI Grand Rounds, that there's a
personal experience with either you ora loved one that sort of is the genesis
of your interest in going into medicineand really having impact in health care.
And so, thank you for sharing that.
And thank you for also just kind ofwalking us through your career arc there.
(06:41):
We're gonna dig into what you're doingat Humata in a few moments, but I
think, maybe, if we can just spend alittle bit more time on your career as
a practicing interventional radiologist.
You got your start at Mayo Clinic afteryour formal training, and I think a lot of
our listeners are always very interestedin what was going through your mind in
(07:01):
choosing a place to start your careerafter that formal training and how you
prioritize different aspects of the job.
What kind of led you to, youcould just spend a couple more minutes,
what led you to, to Mayo and whatwas appealing about starting your
career after formal training there?
The initial answer is, I chose Mayoinstead of several of the East Coast
(07:22):
folks out of the gate because I knew mymom was gonna die during medical school.
Mm-hmm.
And you know, Mayo's a wonderfulplace, but it was for family reasons.
And I think that's a really importantpoint to note that there's a lot
of really wonderful places that youcan work and practice and train.
And because we see this as a calling,sometimes we don't think about taking
care of ourselves and our familyas we're making these decisions.
(07:44):
And frankly,
I would've got phenomenal trainingif I would've went East Coast, West
Coast, or somewhere else in the middle.
But that family decision was also areally big part of it for me, and I think
should be for everybody to some degree.
I had a wonderful time there inmedical school, learned a ton.
Was then there for surgery,internship, as well as residency.
(08:05):
And then spent a couple years in Bostonas a fellow and going to business school.
And so, my decision to go back to Mayo
was in part because Mayo has thisreally wonderful opportunity where
they would say (08:16):
we identify you as someone that we want to have on staff.
Now go to Boston, or go to Paris, or goto San Francisco and learn how to be the
best surgeon of the left pinky or whateveryour thing is and then come back.
And so, that was an opportunitythat I took advantage of.
And then I stayed there for adecade. And so, had the opportunity
to both work at Mayo, but then alsowork at Brigman Women's in Boston.
(08:37):
And so, the reason to come back wasbecause of that initial opportunity.
And the reason I stayed for so long isbecause it was just so many different
opportunities to do different things.
And I think one of the thingsthat academics allows you to do
is to scratch several itches.
And so, not only can you practicemedicine and do an extremely
high level. And even as a youngperson recently out of training,
(08:59):
people are flying from all overthe world to come get care at your
institution. And you really hone your craftand get to be one of the world's best,
which is a pretty special opportunity.
And so, it was that opportunity that keptme there and the opportunity to say,
okay, I know that I don't wanna practice
full-time because I have theseother things that I'm interested in.
(09:20):
I'm interested in technology.I'm interested in investing.
I'm interested in the administrative side,the business side of medicine, and
you get that opportunity in academics.
And so that was really thekey driving factors for me,
Raj.
Yeah.
Maybe before we go to what you're up tonow, you practiced for several decades.
Are you still practicing right now?
(09:41):
I loved interventional radiology.
Unfortunately, it's not oneof those specialties— Right?
—that you can do very well part-time.
You can't dabble.
Yeah,
you can't dabble.
And so, I've had to give it up, nowthat I've shifted full-time to the
technologist, entrepreneur side, butI can't wait for the opportunity to
get back when that opportunity arises.
Right.
So, maybe you could just reflectfor a few minutes here on how your
(10:06):
practice changed as while youwere practicing as an interventional
radiologist over the last few decades.
'Cause you saw this veryinteresting boom of
AI right around the same time,specifically AI for imaging.
And we were hearing sometimesthese hyperbolic quotes around
(10:26):
"we no longer need to trainradiologists" and things like this.
And during this time, you're practicing,you're honing your skills, right?
You're working with a lot of trainees.
Surely, you're getting questionsfrom residents, fellows in training.
Like, wow, is AI gonna automatewhat I'm doing in the next
five, 10 years?
And so, from that perspective you'rehearing all of this chatter, right?
(10:47):
But then you're actually practicing.
And what I hear from a lot of doctorsis things haven't really changed that
much over the same timescale, eventhough the research has really advanced.
And so, I'm curious if that resonatesat all, or if you could describe what
has changed over those couple decades.
Absolutely.
Yes.
And it's also super interesting becausewe're obviously a training program. And
(11:08):
so, at Mayo, where you normally have awhole bunch of applicants, but also
in radiology, which is one of themost competitive, to see that ebb and
flow of the applicants. And frankly,how that ebb and flow corresponded
with is AI gonna take all the jobs?
No, it's not.
Is AI gonna take all the jobs?
No, it's not.
And frankly, saw thatcycle a couple of times.
I'm a firm believer that all of thesetechnologies that we have today, that
(11:31):
we had over the last decade, and thatwe're gonna get over the next decade,
are going to augment humans, augmentclinicians, and augment patients to
only do all of these things better.
And as we've continued to see, especiallyin imaging, as the technology evolves,
we're not seeing fewer images, we'reseeing dramatically more images.
(11:54):
And even with technology, even withartificial intelligence, to augment
even, maybe take over some of those,
I don't see a world in the nearfuture where you're gonna need
fewer radiologists in my mind.
Radiology imaging in general, whetherit's radiology or cardiology, et cetera,
there's only increasing need for humansto sit on top of, or beneath the technology
(12:19):
to really help take care of patients.
And frankly, the best radiologists arenot just technologists, they're also
clinicians, and they're the ones thatgo beyond the sort of pixels and both
have conversations and consultation withradiologists or the doctor's doctor.
So, how do you also then interpretthis for the gastroenterologists
(12:40):
or whatever specialty?
And, and that role isn'tgoing away anytime soon.
Cool, thanks.
Maybe I wanna follow up on that beforewe transition to your work at Humata.
So the reference Raj was alluding to,there was this Jeff Hinton quote in 2016.
Jeff Hinton, Nobel Laureate touringaward winner, godfather of AI, said,
(13:01):
it's just completely obvious now thatwe should stop training radiologists.
It's like the coyote who's gone overthe cliff but hasn't looked down yet.
That didn't come to pass.
And I'd love to understand a littlebit more about why it didn't come to pass,
because clearly like vision capabilitiesfor AI have gotten very, very good.
Lots and lots of studies show thatat least in controlled settings,
(13:22):
they're competitive with radiologists,so there's a capability there.
On the flip side though, there's thisthing called Jevon's paradox, which is,
as a tool makes it moreefficient to use a resource,
the demand for thatresource goes up, not down.
So, I don't know if that's what's going onhere, but, like, I also have had experience
with radiologists where I ask them ifAI is replacing them and they'll say no.
(13:44):
And they'll be like, but you know what?
We actually didn't hirea radiologist last year.
We actually can do more withthe staff that we have now.
So, I don't know exactlyhow to square the circle.
Is there like this
silent replacement where you just needa few lower head count at a practice.
Is that wrong?
Yeah.
I'd love your thoughts on, inwhat way was Jeff Hinton wrong?
I think clearly with hindsight now,he was wrong, but, like, it's not
(14:06):
clear to me in what way he was wrong.
He was wrong looking backward over sincethat time he made the comment, it's
also not gonna happen over the sametimeframe going forward, in my opinion.
And it stems from a bunchof different reasons.
One of them is just
the pace with which we adopt thistechnology in health care is maybe not at
(14:26):
the same level as other industries, andthere's some very good reasons for that.
And there are things that computers justcan't do. And that's stemming back to
my own neural network of the need forhumans in the delivery of care, and the
explanation of care, and the ability to
decipher and understandall of these things.
(14:50):
So, I think that's anotherreally critical component of
the reason why that's the case.
And then the other is just thetechnology continues to advance on
how imaging and how other diagnosticsare being used, that there's always
gonna be this sort of catch up
of how the interpretivetechnology is gonna be used
versus the cutting-edge stuff.
And again, I don't see that slowingdown in health care anytime soon.
(15:13):
If anything, it only acceleratesto come back to your point of does
an individual practice not hiresomeone for the year versus next
year because they're more efficient.
There's no doubt.
Radiologists are dramatically moreefficient today than they were five
years ago, 10 years ago, 20 years ago.
And we continue to see a dramatic need andshortage for radiologists in particular.
(15:36):
But the same thing is true across otherspecialists that are being impacted by
technology, and it's because the need forthese specialties only continues to grow.
And
maybe one more before we move on.
So, if the radiologists are moreproductive and there's increased demand
for radiologists, is there more vol?
Like, where is this additionalvolume of work coming from?
(15:57):
Or do more people have accessto radiological services?
Or if each worker is more productiveand you need more workers, there
has to be an increased volumeof work that is being demanded.
I mean, you're absolutely right.
So, it's a combination of there are morestudies being done, there's also a whole
lot more data in each of those studies,and so the answer is both of those.
(16:19):
And so, there's just awhole lot more images.
There's also a whole lot more studies,and we continue to see that grow.
Three to 5% per year.
And again, imaging is central toalmost everything in medicine.
And there's maybe bad jokes or badwhatever it is, that there's no more
physical exam, it's all go get a CT scan.
And that is in part the way we're trainingour doctors today is imaging is central.
(16:46):
It will continue to be central and onlyincreasingly so over the coming years.
Cool.
Thanks.
Um. I might be biased, but yeah.
Yeah.
It's a radiographic. I think thefacts are true, uh, centric, you know.
I think, I think probably true.
Um, so now I wanna talkabout your work at Humata.
And I'm gonna first tryand state succinctly,
like, what you guys are working on. Andthen maybe break down each one of the
(17:09):
mission statement of Humata down so thatwe can, because I myself am not wonky
on the health care administration side.
So, I think for my own, like, education,I'd love to walk through some of the
pieces here that you're working on.
So, if I was gonna state itsuccinctly, it would be using
AI to solve prior authorization.
So, feel free to hop in thereand editorially edit any of
that mission statement that mayor may not have gotten wrong.
(17:32):
Yes and is what I would say.
I would say we're on a mission to solveprior authorization for patients, and
that means that to actually be ableto do that, you have to solve it on
both sides of the fax machine for bothproviders and payers because there's
a problem on both sides, frankly.
And if patients are gonna get ayes, you need to have an efficient
submission and you need to have anefficient decision on the other side.
(17:55):
And then on the first partof your comment, we're an AI
company solving prior auth.
I would say we're a technologycompany solving prior auth, and that
includes the tools of artificialintelligence that also includes
a bunch of other tools.
And so, one thing that I feel verystrongly about as a problem solver
is you need to start with theproblem, not with the technology.
(18:16):
And you need to use all the toolsat your disposal to solve it.
And those tools will evolve asthey do, and you'll hone and refine
what those tools look like tospecifically solve your problem.
So, that's a long-winded way to saywe are a software company solving
prior authorization for patients.
Got it.
Cool.
(18:36):
Okay, so super helpful.
Now I'm gonna ask you toexplain this like I'm five.
What is prior authorization?
This is a, a phrase that I think lotsof well-educated people hear a lot, and
we understand what the two English wordsmean together and have some vague sense
of what it is, but what is prior auth?
Why does it exist?
And set up the problem for us.
(18:57):
So, I would see a patient.
And decide in my medical opinion,that they need to have a surgery.
Before I could go do that surgery, I wouldhave to contact a payer and an insurance
company to say, Dr. Friese believes thatAndy should have this procedure, CPT code
(19:20):
77450, whatever the number is.
And then you have to prove why you needto have that procedure, and then the
insurance company will take it backinto the bowels of the insurance company
and come back to you with a decision.
Now, when I was practicing full-time, theyhad 30 days to give you that decision.
Now, some of the things that you'reseeing, they have to give you
(19:41):
an answer faster, but the shortanswer is I asked for approval
to do that procedure, and they say,yes, you can do that surgery for
Andy given his health insuranceplan and we'll pay for it.
And so, while it sounds like a reallysimple, short thing you can say, the
short answer is, I have to get approvalbefore I deliver surgery, before I do
(20:04):
an imaging report, before I prescribe adrug to confirm that it will be paid for
by the insurance company.
And this is, I think, simplified, likea piece of friction in the system.
Does the piece of frictionexist to reduce medical waste?
To reduce fraud?
Like, what is this policy, which introducesa friction point in the health care
(20:25):
system? What's it designed to be doing?
Yeah, it, I mean, it's been aroundfor decades, so this like, while it's
kind of become a sexy topic lately,it's been around for decades and
the reason it was put in place is,
A, to stop fraud, as you've said, butthat's only a small portion of it.
The other piece is all of these healthplans are designed uniquely, and so
(20:47):
in one plan, it might be covered. Inanother plan, it might not be covered.
And a way to help control costsof the individual plans is to say
in this plan, you need to get approval.
In this plan, you don't.
But it really is a measure to helpcontrol costs and not just let doctors
(21:09):
do whatever they feel is right.
And presumably there's a patientprotection aspect of this, too,
where a patient doesn't get stuckwith a bill for procedure that
they actually weren't covered for.
There's probably a component of that, too.
Nailed it.
Absolutely
right.
The, the other component of that is that
we do a lot of things in health carethat may not be necessary.
(21:29):
And so, there's a whole lot of thingsthat are done off typical care paradigms.
And so, if you're a glass half fullperson, you would say this is to help
make sure that patients are gettingcare along those care paradigms.
If you're a glass half empty,you know, you would say it's
because the insurance company istrying to inflict friction so that
they don't have to pay for stuff.
And the answer is, it's
(21:50):
somewhere in the middle.
Got
it.
Cool.
So, I feel like I understand priorauth now, so, so thanks for that.
So, based on your framing, I am guessingthat this is not an easy if then else
kind of authorization procedure, that it'svery complicated and context dependent.
So, what are you guys building from atechnological and AI perspective to
(22:11):
make this check to be lower friction?
You're absolutely right.
It's a very complex problem.
There are multiple steps through it that Iwon't bore everybody with, but the crux of
the problem is every single one of theseplans has a different set of rules, a
different set of policies that need to be
met in the clinical record
in order to show that for thisparticular plan, for this particular
(22:35):
patient, you fall in the guidelinesof what would be expected.
So, A, these rules aren't necessarilyeasily known by the providers.
And then B, they're changing a lot.
And at least in our experience,we see that these policies
change upwards of 8% per month.
And so, imagine being a doctor
(22:55):
doing a procedure. The rulesof the game for one particular
plan are changing that often.
That's just not somethinga human can keep track of.
And so, why artificial intelligence andtechnology more broadly is so intimately
perfect for this particular problem isbecause computers are just better at
understanding that huge body of knowledge.
(23:17):
And then
sifting through the medical informationto help you build better submissions.
But here's the other problem, you guys.
I said this is a two sidesof the fax machine problem.
You've got literally armies ofhumans sitting on the payer side.
They're doing the exact same work.
They're understanding these medicalpolicies, they're understanding
the clinicals that get sent over tothem, and literally reading through
(23:39):
a hundred-page document of clinicalsto try to say yes or no, this fits.
That's exactly whatcomputers are better at.
And can do dramaticallymore efficient than humans.
Great, thanks.
So, maybe, if you could talk abouthow interventional radiology has
changed over the past two decades.
How has this prior auth problemchanged in the wake of AI?
(24:02):
So, I can imagine if you go back 10 years,
there's a different set of technologicalsolutions that you might have in 2025.
Is this just an LLM problem now?
Do LLMs hallucinate and thereforeproof things like what are the,
what's the technological state-of-the-art look like
in 2025 for this? Yeah, there.
So, to really solve this problem, thereare really two pieces of the puzzle.
And the first
(24:23):
is a less sexy problem thanusing artificial intelligence.
The first piece is interoperability,and so how do you get providers and
payers talking to each other througha standard set of pipes and rules that
you can actually share information?
And so, what's different, let'scall it this year from even
five years ago, is there's nowa huge push from CMS and others.
(24:44):
Non-CMS bodies to say, let's try tostandardize some of the pipes so we can
share information back and forth, notonly for prior auth but for other things.
And so, we're seeing a significant tailwindfor those pieces starting to take shape.
And, specifically, CMS has a mandate
around that.
By 2027, it continues to get bumpeda little bit, but that's progress.
So, huge progress on theinteroperability front.
(25:06):
Still a lot more work to be done,but really excited about that.
And then the second piece is aroundthe sexy technology side of the house.
And absolutely what we were able todo five years ago to understand a
medical record and match it up withmedical policies was a whole lot
more brute force than it is today.
And LLMs have allowed usto synthesize, summarize,
(25:30):
and do that for both medical policiesas well as for clinical information. And
then match those two up in ways that weresort of unimaginable even two years ago.
And so, our ability to get to the rightclinical information and get that
over to the payer is dramaticallyimproved because of these technologies.
And then the same thingis true on the payer side.
Our ability to understand andsynthesize these things today and
(25:52):
make that process faster for payersis exactly what CMS is pushing for
and Dr. Oz is pushing forward.
I mean, I just feel like we're reallyat a interesting time in the industry
that because of both the interest andappetite from the industry to solve
this problem, but also because thetechnology is at a place where we can
really do this in a meaningful way.
It's pretty spectacular.
(26:13):
Yeah, it's a great transition tothe question that I wanted to ask,
which was a topic you just mentioneda moment ago, large language models.
And so, I feel like we, this is actually,I don't know, we're 20-something
minutes into the conversation, and thismight be a record, before ChatGPT was
brought up, or large language modelswas brought up, just 'cause still such a
dominant topic on these conversations.
(26:34):
I think the less eloquent version,but more direct version of my question
is (26:38):
how have large language models
completely changed your pipelines orthe way you were doing things before
a couple years ago when ChatGPT andnow all of its cousins were introduced?
Yes and no.
And so, as I started with, there'sa whole bunch of other pieces of
this problem of understanding isa prior auth needed, yes or no?
(27:00):
What's the status ofthe prior authorization?
And those other questionsare not necessarily large
language model questions.
They're questions that you have.
That's exactly
what
other systems work reasonably well.
Or maybe even withhigher fidelity than LLM.
Like, you don't need tothrow an LLM at everything.
You just nailed it.
That's what I was trying to say.
The other pieces of the problemreally haven't been changed by LLMs.
(27:21):
The key problem that we've now been ableto dramatically improve our speed and
ability to tackle is the ability to useLLMs, understand the medical policies.
Match that up with the clinicals so thatyou can use that technology to both submit
better clinicals and make better decisionson the payer side. Like, that has been
(27:42):
a complete game changer, which is onereally critical, important piece of the
puzzle, but it is one piece of the puzzle.
And as I think about otherfolks that are in the space.
You can't just throw an LLMat this problem and solve it.
It's part of the puzzle. An importantpart of the puzzle that's changed the way
we've handled some of the clinical pieces.
But it is one important,uh, tool in our toolbox.
(28:02):
Where you use LLMs,
you know, we hear a lot about
hallucinations.
And I think early on we were, I thinkeveryone was, you know, you saw those
New York Times stories with a lawyerwho used ChatGPT and it just fabricated
case law and it did so very confidently.
Right.
And we're hearing sort of mixed reports.
Now we have the newreasoning models, right?
(28:25):
The O series from OpenAI and then theequivalents from Google and others.
And so, we're hearing mixedreports about the sort of
rate of hallucinations.
I think a lot of folks saythey're going down in general.
Others are saying there's still a problem.
Other reports altogethersaying they're going up.
Right?
And with some of the new advanced modelsthat are more capable, but also still
(28:45):
hallucinating, potentially even morethan some of the early incarnations.
And so, I imagine, this is a question thatI think comes up with any of the companies
that are using LLMs, but I think also aswe're thinking about outside of companies,
just physicians using products, right?
That they're using now,
in practice and I think atvery large scale already.
How do you think about hallucinations?
(29:07):
How do you guard against them fromyour perspective where you use LLMs?
Is this a big concern andwhat can we do about it?
Lemme step back, lemmegive you an analogy.
I've got teenage kids.
When they're in their math class,they have to show their work.
They can't just show the final output.
I've always believed in usingartificial intelligence,
you need to show your work.
(29:29):
And so, that's exactly the sameapproach we're taking with LLMs.
So, as an example, we will use an LLMto summarize and to build questions.
We will always submit the source work.
And so as you're submitting something to apayer, you're not just submitting the LLM
output, you're submitting the LLM output
and the actual doctor's notes.
(29:52):
'Cause frankly, that source of truth isthe only thing that is being considered
on the flip side of the fax machine.
And so,
but yeah, what if the doctor'snote was written by an LLM?
So, even in that scenario,they have to approve it.
So, we are making the assumptionthat what is in the medical record,
whether it's written by an LLM ornot, there's still a human oversight
that says yes, this is the fact ofwhat was discussed with the patient
(30:17):
and it's true. If there's a fallacyin the medical record, I, that would
give me great consternation, butI trust that Shiv and the Abridge
team and the like in Microsoft andthe Ambient and the other folks are
making sure that, that those sort ofhallucinations are not showing up there.
And we're saying we're Switzerland.
(30:37):
What's documented in the medical recordis fact, and we're gonna use that and
get that over and, and show our work.
Cool.
Thanks.
The other thing I'll add there, thesame thing is true on the payer side.
So, under no circumstance willour computer ever be used and
say no for care, like, because ofhallucinations. And frankly, because
(30:57):
I think this is a human endeavor.
A human, a trained physician, or nurse, orsomebody, needs to be the one that says:
I'm sorry, Andy.
I'm sorry, Raj.
You are not approved for this carebecause of A, B, and C. And that
needs to be a human, not a computer.
The computers can help the human get toan answer, but you need to show the work
of where in the medical record is thatdocumented to answer this question or
(31:20):
where there's contradictory information.
And again, we use LLMs to showwhere that information is, but the
summary cannot be the final answer.
Cool.
Maybe one more question on thisbefore we move to the lightning round.
So, you had this great way offraming the area of health care
that you operate in, and it's likeon both sides of the fax machine.
(31:40):
And I think that that, like, really nicelyhighlights for me— Like, it might be the
title of this episode on both sidesof the—. Yeah, on both sides
of the fax machine.
It nicely highlights how like there arethese different technological strata in
the health care system and you're workingin the one that happened to be ossified
around 1985 or something like that.
Uh, and when fax machines werestill like cutting-edge technology.
(32:01):
But I wanted to talk a littlebit about the business model. Because
I think one of the things that haschanged a lot in health care over the
last five years is there a very viableand lucrative business models for
technology companies. Like it used to be
impossible as a startup in the health carespace to find a viable business model.
I think you mentioned Abridge,they've obviously unlocked one.
Some other health carestartups are doing it too.
(32:22):
I think you guys are onto it.
So, which side of the fax machine do youconsider to be your primary customer?
Is it the doctor to help get stuffapproved, or is it the payer to
help them deny unnecessary care?
Or is it both?
I like to say we're inthe business of yes.
And so, I want to help doctors get to yes,
(32:42):
so that patients can getthe care they deserve.
And I want payers to get to yes
more efficiently, so that they cando that more efficiently and cheaper
than what they're currently doing
and get the patient to yes.
I mean, our business model is wehelp both providers and payers.
And both of them have an ROI.
To do that using technology justlike in the ambient dictation space.
(33:05):
I think prior auth is another area whereboth health systems and payers are saying,
we cannot continue to do this with humans.
It's just silliness and they'readopting artificial intelligence and
our software, and broadly at a ratewith which I haven't seen in the past.
And it's a pretty excitingthing to experience.
Cool.
Awesome.
(33:25):
Thanks Jeremy.
So, I think if it's okay with you, we'regonna move to the lightning round.
Let's do it.
So, lightning ground, it sounds likeyou've listened to the show before.
The goal are sort of succinctanswers to a grab bag of questions.
You can decide which ones areserious, which ones are non-serious,
(33:47):
and we'll just hop into it.
Let's do it.
So, the first one and thephrasing of this one matters:
In what ways will AI change medicinethe least over the next five years?
Doctors
seeing patients and it beinga truly human endeavor.
Oh, interesting.
I might come back tothat in big picture, but
(34:07):
we'll, we'll keep it moving.
Alright, Jeremy. This, this is thesecond lightning round question.
If you weren't in medicine,what job would you be doing?
I'd be a professor.
A professor of...? Business.
Nice.
I think.
Sorry, I know this is lightning round, but you,
you got me going.
(34:27):
You know, I think there are severalthings in this world that can have
dramatic impact at scale and Ithink capitalism deployed in the right
way can have a huge impact on humans.
And so that's, uh, that'show I see the world.
Alright.
If you could have dinner with oneperson, dead or alive, who would it be?
Teddy Roosevelt.
Oh, nice.
(34:48):
Probably some bourbon.
Bourbon would be had.
Amen
to that.
First modern president, developedthe national parks, interesting chap.
Nice.
Do you think things createdby AI can be considered
art?
absolutely.
Alright.
Last one.
Uh, you, you touched on this, at thebeginning, but we're gonna revisit it now.
(35:10):
Should medicine be considered just a jobor should it be considered a calling?
I think that that's anartificially challenging question.
I think the answer is yes.
There are components of it where it's ajob and you have a, you're a human outside
of this, and if you're going to actuallybe able to practice it as a calling, you
(35:30):
also gotta take care of yourself outsideof that, meaning that you have to be able
to partition it and seeing it as a job.
Cool.
Thanks.
Well, Jeremy, you survived.
Yeah. I was gonna say
in the business of yes.
Yes, that's
right.
Well, Jeremy, you survivedthe lightning round.
Nicely done.
Thank you.
Thank you.
Alright Jeremy, so we just havea few last sort of big picture
(35:52):
concluding questions for you.
And I think we've touched onboth of these various points
of the conversation already.
But you are a rare breed of physician.
Executive leader of a technology company.
And to be honest, I wishthere were more, right?
We had Shiv on the podcast a coupleepisodes ago, and there are a few other
(36:14):
great examples, but I think there reallyis something different in my mind about
a company that's being led by a physicianworking on problems in health care.
And so, you really understand
what doctors do, how theywork as part of care teams, and
what patients are looking for.
And I think it, it shapes a lotof the way you design the company
and your mission. With that setup,
this is a bit of a sort of interestingquestion to ask then, uh, but will AI and
(36:38):
medicine be driven in your opinion moreby computer scientists or by clinicians?
I guess neither is also an answer.
So, you know, one of the things thatreally impacted my neural net and the
way I see the world, both care delivery,but also building technology businesses.
Again, at my days at Mayo,every leadership post had two people
(37:04):
that were equally yoked to tacklewhatever the leadership position was.
It was a physician and administrator.
And when I think about
building technology businesses
and the answer to your question,I don't think that you can expect
a physician to have the same levelof technical skill as someone that
(37:26):
lives and breathes it every day.
And I don't think that a technologist
can truly understand the ethos of theproblem of health care and the delivery
of taking care of a patient as well assomeone that has put hands on a patient
and done surgery or done whatever.
And so, it really needs to be amarriage of the two, and the companies
that will have the biggest impact
are the ones that do that the best.And I think that same thing is true
(37:48):
when you look at care delivery.
The ones that do that thebest are the ones that
keep the mission aligned around theright thing and then build solutions
that really solve that problem.
So, I, I guess I'm in the business ofyes, I think the answer has gotta be both.
And the companies that havethe biggest impact are gonna be
the ones that do that the best.
Cool.
I maybe want to ask a slightlydifferent variation on that question
(38:13):
about AI versus human-led medicine.
So, one of the big themes we explore onthe show is obviously AI and medicine.
We tend to have this artificial
notion of what that looks like, which islike a doctor-in-a-box kind of construct.
But really what we've learned throughmany episodes on this podcast, including
this conversation, is that actually AI isalready mediating a lot of interactions
(38:34):
in the health care system already.
Going back to Abridge, going back towhat you guys are doing, going back
to people who talk with ChatGPTbefore they even see their doctor.
So how
far do you think this sort of AIexpansion goes? Are there certain
interactions in the health care systemthat you think in principle will
never be mediated primarily by AI?
Is there a human carve out that you'dlike to preregister here that you
(38:56):
think that is just like not able tobe serviced by AI or do you think we
have essentially full self-drivingmedicine in the fullness of time?
That's a big question.
I know, but we're to thebig picture section, so.
There's no doubt in my mind therewill be components of medicine.
The world I love,
I love the analogy of fullself-driving AI medicine.
You're already seeing components ofthat today, and there's no doubt that
(39:20):
will only take shape to a greater degreeover the coming years and decades. And
humans, since the onset of Google, havebeen using technology to try to augment
their interaction with physicians
and other care providers. I think thatonly becomes greater and maybe some of the
(39:41):
simple stuff continues to then become moreAI-led or computer-led versus human-led.
And frankly, it should, like, we'vegot, it's difficult to get access,
it's difficult to get information.
It's difficult to get accessto, to me and other physicians.
So, you should be ableto get these answers.
In my former life, was annoying whensome, someone asked Dr. Google a
question before they would ask me aquestion because they would usually
(40:03):
go, like down to the third page andthen think that was the gospel truth.
It is only gonna become more.There's no doubt that AI will carve
out pieces of the puzzle and humanswill not need to be involved.
And I welcome that opportunity.
I think every physician and nurse shouldwelcome that opportunity. And there
will be parts of the doctor-patientinteraction that are, are and need to
(40:26):
remain deeply human. And those connections
maybe only become more importantas technology plays a role in
the other pieces of diagnosisand treatment to allow the human
caregiver to be a human caregiver.
Got it.
Cool.
Maybe one question before we wrap here.
(40:47):
You've had a very impressive career asan interventional radiologist, like doing
the practice of medicine, seeing patients.
You're now the CEO
of a successful health care startup.
So, you've done lots of work in diagnosticmedicine, the practice of medicine,
and now on the administrative side.
So, could you give us a sense of which areaadministration or, like, medical practice
(41:10):
has the more, higher potential for AI?
I have a weakly held opinion, butI'd love to hear your thoughts
which side of the house you thinkactually has the most potential.
Two answers to the question.
The first is on the velocity andwhere you're gonna see the impact
first, and then the, the second comesto what the bigger opportunity is.
It's unequivocal, no doubt in mymind that the velocity of the huge
(41:33):
immediate opportunity is on theadministrative side of medicine.
It's back-office BS that we'retackling. It's documentation.
It's the administrative stuff,and that's a massive problem.
Doctors are burnt out.
Patients aren't getting the care theydeserve. Like, we have to deploy it here,
and that's where the opportunity sitsin the near term. And in the long term,
(41:54):
giving patients access to good qualitycare at their fingertips, and being able
to do that 24/7. And ensuring that theright information and the right care
is getting delivered like that is sucha much larger opportunity than we're
tackling on the administrative side.
(42:15):
And I firmly believe that that will takeshape over the coming years and decades,
and I can't wait to be a part of that.
Cool.
Awesome.
Well, Jeremy, this has beena super fun conversation.
Uh, thanks for taking thetime to sit down with us.
Great
to spend time with you guys.
Thank you.
Thanks so much, Jeremy.
That was great.
This copyrighted podcast from theMassachusetts Medical Society may
(42:37):
not be reproduced, distributed,or used for commercial purposes
without prior written permission ofthe Massachusetts Medical Society.
For information on reusing NEJM Grouppodcasts, please visit the permissions
and licensing page at the NEJM website.
Popular Podcasts
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.
Dateline NBC
Current and classic episodes, featuring compelling true-crime mysteries, powerful documentaries and in-depth investigations. Follow now to get the latest episodes of Dateline NBC completely free, or subscribe to Dateline Premium for ad-free listening and exclusive bonus content: DatelinePremium.com
CrimeLess: Hillbilly Heist
It’s 1996 in rural North Carolina, and an oddball crew makes history when they pull off America’s third largest cash heist. But it’s all downhill from there. Join host Johnny Knoxville as he unspools a wild and woolly tale about a group of regular ‘ol folks who risked it all for a chance at a better life. CrimeLess: Hillbilly Heist answers the question: what would you do with 17.3 million dollars? The answer includes diamond rings, mansions, velvet Elvis paintings, plus a run for the border, murder-for-hire-plots, and FBI busts.