– English Lyrics

Lyrics & Translation

[English]

I actually applied immediately. I got

rejected because I was in high school.

And then I applied again. I got rejected

again. And then the third time we came

armed with a prototype that actually

worked. The market is a great I mean

it's just like punch in the face. Like

it teaches you a lot like very quickly.

I think our biggest learning is get in

front of the customer and then just

solve problems for them and and start

with one specific problem regardless of

how small it looks. We went from having

no revenue to having 3 or 4 million in

revenue in like a year. And it was like

well we're done. like we solve all the

all the problems that are to solve and

then and you're like well no actually

the game starts now.

[Music]

>> I'm thrilled to be joined today by Tana

Tanden the CEO of Kamir. We're very

excited to have him because he's one of

the leading providers of enterprise

software to healthcare. Thanks so much

for joining us.

>> Thanks for having me.

>> Why don't we kick things off? Why don't

you tell me what Athellis is and what

are you guys working on?

>> Yeah. Uh, Athellis Camir builds software

products for providers and so we have

tools that automate their day-to-day

workflows like Camircribe which is an

ambient documentation tool revenue cycle

which is a payment stack. It's basically

Stripe for healthcare and doctors. It

submits their claims, uses LLMs to

negotiate denials with insurance

companies, appeal those denials, and

then really render all of the data viz

and business intelligence to run your

practice from a financial standpoint.

And then also a whole host of clinical

tools. come engage which does back

office and front office interactions

with the patient, explaining their bills

to them, uh explaining what types of

tasks they need to do to get prepared

for a procedure like a colonoscopy,

sending them appointment reminders, do

on the order of a couple billion dollars

worth of payments volume every year. Um,

help document using our ambient AI tool

20 million appointments every year and

then uh probably 150 million touch

points uh annually on on on our patient

engagement piece. um spread in large

health systems and private practices.

>> And so this is sort of a bundle of

pieces of software that are each solving

kind of independent problems for a

practice and collectively sort of serve

as the software operating system for

this entire practice.

>> That's right. And so you know our

solutions are used by the physicians and

the nurses in the practice or the

hospital and then also by the back

office. So your accountants, your

billing team, what we call revenue cycle

in healthcare. everything from uh you

know submitting the claim to actually

you know ledgering everything up for the

uh CFO's office

>> and you can actually do all of this with

technology today. You can have an AI

model that'll call up an insurance

company and negotiate with them and that

actually works

>> 100%. I mean I think the ability for

language models to both interact via

voice and text with humans and we just

got there a year ago and so we're just

at the precipice of what it can

accomplish. Um, and when you walk into a

health system, there are tens of

thousands of people whose sole job it is

to do these types of tasks. And if you

can free up their time, you can bring

that productivity back onto what matters

the most, which is patient care.

>> And I imagine a lot of those tasks are

already outsourced to some degree. So,

you're kind of just slotting in to where

they've already outsourced it to someone

else.

>> Exactly. They they use vendors like R1,

which is an offshoring RCM provider,

Axis, Omega. Many of them have their own

in-house shops where they actually, you

know, it's effectively outsourced, but

they actually have the individuals on

site and the the direction of the

industry for the last 20 years has been

offshoring. It works okay for what it's

worth. I mean, these became big

businesses, but I think LLM's there's an

opportunity to reimagine all of that

where it's it's pure software.

>> And this ultimately becomes a product

offering for all types of hospitals,

both small and large.

>> Yep. We work with the HCAS of the world

which is a hundred billion dollar

hospital empire you know 3,000 sites of

care 186 hospitals all the way down to

your private practice owned by a single

physician and you know they use varying

parts of the solution but it's provided

as one platform which is revenue cycle

workflow automation and patient

engagement in one

>> what does that actually mean like for

someone who's not in healthcare like

what do these roles actually look like

in terms of where technology fits into

them

>> yeah the way I think about it is the

physicians dayto-day day is really

defined as interacting with a patient

and treating that patient uh and then

often generating the relevant

documentation so that their back office

which is think of it like their

accounting team their AR team invoicing

can actually submit those claims to

insurance.

>> One of the challenges in healthcare is

the person paying for the for the

service is not the person receiving that

service. So a person paying as like an

insurance company or Medicare or

whatever receiving it is the patient

>> is the patient themselves and that

individual the beneficiary you know

their incentives are to get treated to

get treated fast to get the best

treatment in the world and often the

insurance company's incentive is to pay

out as little as possible and hopefully

treat as little as possible which is a

counter incentive to really what the

patient is trying to accomplish. And so

when it comes to then building

technology then you guys build

technology for all parts of this from

the patient experience itself as well as

the kind of front and back office

accounting and kind of administration of

that patient experience. We we see it as

you know there are two key protagonists

in our story which is the patients and

the providers and we build software

solutions and hardware solutions for

that group of individuals and there's an

effective enemy that's created as a

result which is the insurance company

and all of our tools are pointed at the

payers in in terms of trying to make

their lives harder and make the lives of

our customers easier.

>> Okay. Very interesting. Okay. I want to

dive into your technology, but before

that I'd love to hear a little bit about

the backstory of so why don't we rewind

the clock uh to the very early days like

how did this company come about? Like

what was the very early days like? Um

what was your experience in YC like and

kind of how did you even begin to arrive

at this idea that you're working on now?

So I I was very fortunate to have grown

up in the Bay Area and when you're in

the Bay Area, Y Cominator is just in the

ether like people like Casar and Justin

Khan and Sam Alman were like our heroes

and the I remember it was like 2013 or

2014 that we somehow snuck into YC

startup school because it was happening

in Certino. Um

>> I think I was there too actually also

sort of snuck into it.

>> It was a great start. I think Zuck came,

Jeff Dorsey was there.

>> Yep. Same one. Um, Flexport did their

interview on stage and I remember I was,

you know, I was like teenager in high

school. I was just in awe of of of these

people that had built such amazing

technology and YC is this ecosystem that

had enabled a lot of it uh and was

starting to enable more and more of it.

So, you know, fast forward a couple

years, Y Cominator hosted this hackathon

called YC hacks and uh I had read a

couple papers about how you could use

computer vision to potentially analyze

blood cells and also use really cheap

pieces of glass to enhance a smartphone

camera to basically turn into a

microscope.

>> So, this is like in 2016 or 174. 2014.

So, this is like early computer vision.

Like the deep learning is like just

starting to happen to some degree.

Computer vision is just barely starting.

the the models that we were using them

were like even like for for computer

vision were you know you're using like

segmentation algorithms like watershed

and like you know we like random forest

is what like were like the best models

for for often NLP and even in some task

in computer vision. And so armed with

those two pieces of information, I went

to YC hacks um and built this little

really hacked together smartphone camera

with a random force that had been

trained to classify malarial cells

versus non-malarial cells um based off

of a training set on the internet. And

it worked like pretty well like you

could you could segment these two. Um

and and then you know I was a senior in

high school at the time but it became my

science for project that year and I got

very obsessed with this this idea of man

we can use computer vision to automate a

task that someone like a pathologist a

very trained person is doing in in a in

a expensive laboratory and bring that

care directly to the patient um and

simplify the provider's life and that

eventually became a which was started as

a med device diagnostics company. Oh,

interesting. Okay. So, when you when you

applied to YC then, so this is shortly

after that, maybe a year or two after

you were a freshman in college or

something like that.

>> So, I actually applied immediately. I

got rejected because I was in high

school and and then I applied again. I

got rejected again. And then the third

time we came armed with a prototype that

actually worked. And it was after Id

spent about six months at Stanford and

you know the resources that you have

access to with the med school and the uh

you know in the the mechanical

engineering department like we were able

to put something pretty serious

together. Uh and that's when we got in

for summer 16.

>> And this was with a med device that was

doing some type of kind of image based

analysis of was it was this original

idea of like blood sample analysis?

>> That's right. So it was basically taking

a small volume of blood, turning it into

a monollayer, so you could see, you

know, single cells basically like

they're on a on a smear. Um, and then

basically a a microscope with an imager

attached to it, but was hacked in a way

that it was very uh large field of view

and still high resolution. And then we

trained computer vision algorithms to

recognize and segment those various

cells.

>> And so at this point, you know, your

background is primarily in computer

science. I assume you had some

experience in this application area in

biology in high school and early

college. You know, what gave you the

conviction to be like, I'm going to

presumably drop out of college, start

this company that is in a regulated

industry with medical devices. Like how

did you actually arrive at that

conviction? I

>> I would say two key sources. one my

co-founder Deepka who's also one of my

closest friends uh we used to compete in

science spheres against each other and

her research was always way more uh call

it med device or bioengineering focused

and she worked through high school on

this uh micrfluidic test strip that

could detect salmonella very quickly

from you know produce and you know I

thought it was remarkable that you could

build something like that in in high

school because it was you know was

hardware there was bio involved you have

to have a good understanding of you know

how these det protection strips were

actually fabbed and cut and whatnot. And

so it was clear that okay, we could

build things in the real world with uh

with micrfluidics. And then number two,

I think the, you know, when you're at

Stanford, you get to meet these amazing

professors. For me, Chris Manning was

was someone who gave me my first shot in

the Stanford AI lab under Richard

Socher, um, who started Metammind where

I was a research intern as well. And I

think they were just it was really

motivating because for them it was like

there's there's no boundaries to what

machine learning can do. you should go

after these complex industries. Um, and

they, you know, Richard was ended up

being the first check right after Y

Cominator 50K check and and I think that

support really encouraged us as well.

>> Yeah, totally. Okay. So, so let's now

talk about your time in YC. You know, at

this point it's just the two of you. Um,

you're working on this regulated medical

device. What did you set as kind of your

demo day goal and like how did you what

did you accomplish in those few months

of YC that ultimately then put you on a

trajectory to build something somewhat

different now from what you describe

now? We'll kind of talk about that that

transition. But yeah, what what did that

first few months look like?

>> I I think the first week itself of YC

was total shock because we came in with

these traditional timelines for how long

a med device should take or how long

anything in healthcare should take,

>> which is like years or something.

>> And and that seemed pretty good to me. I

was like, cool. Like we're going to be

on this for a long time. It's, you know,

we're going to go super slow. Remember I

had a conversation with with Casar, who

was our group partner and is also now on

our board. Um, and he, you know, he was

basically like, he made me walk through

every assumption on, well, why is that

months? That that just seems like it's

like a paperwork. You should be able to

do it in a week. Why are you waiting for

Stanford IRB to approve this trial where

you could go to some local lab and

probably get it done in a week? Uh, and

I think because of that, Deep and I sat

down and just like reimagined how

quickly you could build this stuff and

and really went from almost like

zerobased budgeting, but for a clinical

trial. And the other piece was, you

know, if you're a competitive person, in

our batch, we had companies like, you

know, Scale AI was in our batch and and

this was a company that pivoted mid

batch and got to like millions in

revenue by the end of the batch. You

know, for us, we never wanted to just be

the best healthcare company in the

batch. We wanted to be the best company

in the batch. And and so it was it was

this this pressure that we sort of

created for ourselves and that group

partners created that really caused us

to pace uh more aggressively than if we

were doing it outside of of of YC. And

so we set this goal of we will finish a

clinical trial and have a ready to go

FDA submission by the end of YC. Um and

it was crazy but we actually got the

clinical trial done. We needed to redo

some of the experiments per the FDA's

request. So the the approval itself took

a little longer, but we we we got the

the core basis of of the clinicals done

in those 10 weeks.

>> I mean, that's pretty absurd pace,

right? I mean, I think like when you

think about a normal regulated medical

device, we're talking about many years

in that process. And yeah, I mean, I

think this ability to kind of break down

into its only what the critical

components of it are, what parts you can

actually shortcut versus not is it's a

very challenging head space to get into.

And then you guys as as 19-year-olds, I

mean, I can only imagine some number of

people you talk to are like, "Who are

these guys to be making a medical

device?" Like, how do you go up against

that? Like, how did you get yourself to

not care? Basically,

>> we had many conversations where we got a

quote from Stanford where it was like,

"We can run this trial and it'll be

$120,000 and it will take 3 years."

>> And and I mean, when you see something

like that, it's almost so absurd that it

wheels you into action a little bit,

which is like, if this is the default

system, I mean, something is broken. and

and and I think um because of that we

thought to ourselves it's time to

rethink first principles and it really

came down to one identifying a partner

hospital that was small enough and

willing to work with us very quickly and

that ended up being in Huarez Mexico so

we flew out there mid batch uh ran the

trial with them number two is really

deleting as many parts as possible and

running the simplest form of the trial

so instead of hiring traditional

clinical trial operators deep and I ran

the trial ourselves she took samples in

the front room. I was in the back room

running them on the device. Uh, and then

we had a nurse that we had hired for the

day who was running them on the

traditional system called the

Sysmex5000.

And we would then compare those results

and show that they were accurate and and

equivalent to each other. And the whole

process from finding the hospital to

making sure the devices were ready to

actually running the trial and then

getting the results end to end was

probably 6 weeks of the batch. Once we

hit those milestones, that's when we

raised our seed round.

>> Okay. So you started there though. You

started with this medical device. Today

you're not a medical device company for

the most part at least or at all I

think. What was that arc like? I'm sure

there was many steps in which the

company changed. But maybe tell me a

little bit about how you segment your

brain into like the few eras of this

company and how you went from one of

these to perhaps pivoting to uh kind of

another version of that company over the

last few years. you know, we did YC, we

raised our seed round from Sequoia,

Alfred, um, over at Sequoia, and the

next 18 to 24 months just became brutal

focus on getting an FDA clearance. Um,

get the device to a state where it can

replicably run in in in clinical trials,

uh, where you can, you know, all kinds

of crazy tests like we have to drop the

device and kick the device and, you

know, things break and you have to be

able to continue to pass the the

relevant tests. And that was it. We were

laser focused on this one goal. And I

think that was very freeing in some ways

because it was the only thing we had to

get done. We also only had $3 million to

do it. Um, and you know, the the other

piece of it was to start proving out the

commercials. And that's when I started

engaging with uh what became our first

customer, which was actually a

pharmaceutical company that manufactured

a drug called Cloopene, which is a

refractory schizophrenia treatment that

causes neutropenia, white blood cell

count reduction as a side effect. And

you know, they told us, "We've been

dreaming about a device like this." and

and they signed that first million-doll

contract even before we had approval

just based off of the results of the

clinical. So, we had that commercial

proof. We had early clinics that were

adopting it. We were, you know, seeing

them use it and then we got the FDA

clearance and that was really this like

the first like epoch of the company the

first like two two and a half years.

>> Once we got that, you know, the the

market is a great I mean it's just like

punch in the face. Like it teaches you a

lot like very quickly.

>> You must have felt like you were on

cloud n then you're like, "Oh, we're

we're chilling. We we figured this out

now. Now we're going to go to the moon

with this." I I mean we got the

clearance and it was like well we're

done like we solved all the all the

problems that are to solve and then and

you're like well no actually the game

starts now and the exciting thing was

that we went from having no revenue to

having 3 or 4 million in revenue in like

a year. So as far as traditional series

A metrics and pacing like we were there.

The thing that we realized in parallel

was okay we can probably scale this

business to tens of millions in revenue

and build a business that a ro or an

Abbott or someone will buy for you know

200 million bucks or whatever that

number is like a good med device

outcome. At the same time we had this

team that we felt could run through

walls and had built machine learning

software had gone through regulatory

done the operations of device

manufacturing and we were in these

clinics like our devices were deployed

and we saw all kinds of problems.

>> So you got in there by deploying the

devices we were solving all sorts of

other problems they were facing and

along the way you discovered there's

this massive other set of problems that

you can solve too.

>> Yep. We we wanted to go from making

their lives easier for a percent of

their patients. So we were serving with

our athals one device to 100% of their

patients uh and their whole chronic care

panel with you know the teleahalth tools

and the revenue cycle tools. And and I

think you know my looking back the

reason this time period was so valuable

is we would not have had you know a

valuegiving relationship with these

customers had it not been for the

device. They trusted us. They saw us as

more than just a device company. They

saw us as those kids that would show up

in their office, set up the device,

train them how to use it. And because of

that, they would tell us about all these

other problems they had in their

clinics. Like, hey, when I get the

result from the device, I have to upload

it into this portal manually, and then I

have to call up a pharmacist and tell

them like, hey, the fax is coming

through. We would just hear these, you

know, these complaints like, well, it

would be amazing if I could, you know,

upload this automatically or if I could

fill this claim form out automatically

to get the next set of medication

adjudicated. And that's when we started

building software and it honestly felt

like playing on easy mode after two

years of building building hardware and

dealing with the FDA.

>> You know, at some point presumably the

software starts to take off and become a

pretty significant part of your

business. At at some point do you you

what is that process like of leaning in

on that and perhaps you know abandoning

the thing that or or at least leaning

out of the thing that has been your

initial driver that got you there in the

first place that has this big pharma

ideal for example. What does that

process look like? It was interesting

because we had two paths in front of us.

Be the best aellis swan medical device

company in the world and just you know

build a great compounding 40 50%

annualized grower in that in that

segment. Alternatively it was bet on

yourself again and expand the TAM and do

more for these customers um based on the

learnings we were getting from the

market.

>> And also when was this by the way?

>> This was in 2020. Um

>> so you've been at this now for like

fourish years. Like this has been your

thing. You're a med device company.

you're growing this thing and now 2020

is coming. A lot of things are happening

in the world uh as I'm sure is going to

be relevant here. And then also you're

starting to see that there's an

interesting new opportunity in software.

>> Yep. And I remember it was I forget was

it might have been Robin Hood or it

might have been some company. We had

this interesting uh fireside that we

went to and they talked a lot about

share of wallet and this concept of you

want to do more and more for your

consumer and go from just being you know

with Robin Hood like their toy stock

couple hundred bucks playing around to

like no this is like a meaningful

financial product for your entire

portfolio. For us, the the analogy was

was apt because we went we we wanted to

go from treating 1% of their patients,

these refractory schizophrenia patients

that were treatment resistant to

treating all their patients. And I

remember Deep, my co-founder, she was in

one of these clinics. She spent a lot of

time in these clinics and she was like,

"It's insane. We are helping them for 1%

of their patients and they love us so

much. Imagine if we could build tools

for all of their patients, their full

thousand patient panel." And and I think

that's when we took the bet. We're like,

we can build a multi-billion dollar

software business, sensor business in

this segment by expanding what we do for

these customers and and grow a lot

faster than the 100,000 total patients

that had this one condition.

>> And then this was before the LLM era was

really taking off. You were obviously a

machine learning company and you had a

lot of experience with building

software. Um, but like what did that

initial software look like? Like what

types of problems could you solve then?

Presumably now you can solve many more.

Yeah, we started with the basics which

is one uh co hit and and as a result our

providers needed ways to monitor their

patients in their homes. So we built a

basic telealth portal and a set of

remote monitoring devices that connected

into it and allowed them to interact

with those patients many of whom also

used the FLS1 device in the point of

care or in their homes. The very next

thing we started playing around with was

workflow automation as it related to

claim submission and payments. So

everything from collecting dollars from

these patients to uh you know starting

to submit claims to insurance and first

we did it just for or you know this

subset of remote monitoring and cloopine

patients and then expanded organically

to more and more of the practice

>> and and these are problems you basically

discovered from being in their offices

like you know you and your co-founder

are in there and you're noticing and

hearing from them about whatever other

problems they're facing and you're kind

of seeing okay like as exciting as this

is we sort of have to solve these as we

go.

>> Exactly. There is no other way to

discover these problems than being in

your customers dayto-day and and and

just seeing them because there's so much

nuance to where they get stuck and it's

not something you can derive you know by

reading an article or you know watching

a video online. You you have to be in

the in the heat of things.

>> Yeah, totally. I mean I think we see

this with with companies every day today

especially in this sort of new

technological moment is that it's very

hard to conceive of great startup ideas

kind of in your home or in a lab. It's

much more likely to encounter them

somewhere in the field by being in

there. And so for you guys that sort of

was a was a game changer in putting you

in that place.

>> 100%.

>> Well, okay. Well, why don't we, you

know, kind of move forward now a little

bit to to today and and get there by

kind of talking about what the last few

years has looked like in healthcare

generally. I mean, there's been an

incredible amount of changes happening

in healthcare. To your point, there's

these massive mega conglomerate style

hospital businesses. There's

consolidation in payers and PBMs. for

you guys as you see this sort of

changing landscape where do you one

where do you fit into that landscape and

then also how does that changing

landscape affect your business and as

you think about growing it

>> I I think it's it's it's really

interesting is if you go back to the

'90s uh the life of a physician was

great they had a tight panel of patients

they knew them personally they treated

them and if there was you know chronic

diseases they were able to talk to the

patient very quickly idea of a concierge

doctor was was fairly common you know

anyone in upper middle middle class had

a concierge doctor of some kind. And as

the bloat that came from insurance and

regulation and and and just what we

turned healthcare into into this sort of

like admin state

>> is what really broke down the

productivity of a physician where they

turned into these these, you know, cogs

and this greater wheel and there's, you

know, patient productivity numbers they

have to hit. There's countless tasks

they have to do. regardless of how

efficient they are, they always take

documentation home and they're filling

out paperwork. And I see it as this

great misuse of of talent in that we

have some of the most intelligent,

well-trained, uh, uh, well-intentioned

people in the country spending their

time doing tasks that really software

should be doing for them. Uh, LLMs are

not. And what we've seen happen is a lot

of physicians give up on this dream of

the private practice or the

physician-owned practice in change for

joining a large health system that does

take on a lot of the overhead that you

now need to pay in order to just operate

a health system.

>> This is like negotiating with insurance

companies and with PBMs and having uh

farmer relationships and all that kind

of stuff,

>> paying for an EMR, dealing with

malpractice. And so all of this made it

harder to practice care and and and

turned it more into this again like this

admin business. And and I think now with

LLMs, there's this generational

opportunity to free the physician and

and really bring them back to what they

love doing the most, which is rendering

care. And I think our work is in service

of that. Uh take the work tax and just

nuke it. take every part of their

revenue cycle, their back office, their

documentation, um, and and use software

to automate, uh, large chunks of that.

>> And I mean, you guys sell to, you know,

both this mid-market and larger market

or or any kind of enterprise market.

It feels like what you're saying here is

that this midm market is what you see as

potentially best suited to take

advantage of of LLM and especially sort

of this broader technological adoption

because it frees them from needing to

consolidate into these bigger

enterprises and perhaps the advantage of

the bigger enterprises won't be as much

the case in the future. Is that kind of

how you see it or do you think not so

much?

>> I I I think it goes both ways. one I

think we will see it's sort of the

analogy we use is like Amazon and

Shopify where over the last 20 years or

10 years or whatever time period Amazon

has grown remarkably um as it's you know

taken a lot of share from traditional

retail and also expanded what we order

online but then Shopify has exploded as

well as more people are able to start

their own businesses and I think you

will have this aspect of the physician

who practices in a health system but

then also has their own practice um and

chooses their hours and because of the

fact that software is automating more

and more of their day, they can see more

patients and do it in in a way that is

sustainable and gives them energy as

opposed to drains them out by the end of

the day.

>> Yeah. So, let's talk about that a little

bit in terms of software automating more

and more of their day. There's been a

lot of software adoption in healthcare

over the last 10 years. Most of it

through electronic health records and,

you know, kind of payment portals with

insurance and whatnot. Um, but a lot of

it has been promised to doctors as

saving them time or helping patients in

various ways over the last decade. But

the critique has often been that well,

it's not really been about that. It's

been about getting paid for care. It's

been about billing mostly and not

actually about care. How do you see what

that technological adoption story looks

like over the last decade and then over

this next decade? Like what are things

you think are going to be fundamentally

different about this technological

moment? It's it's definitely true

because a lot of the digitization and

creation of EMRs came from a place of

compliance and a place of of billing and

institutionalizing a lot of regulatory

requirements as opposed to true you know

unleashing the doctor. You know do Dr.

classical who's one of our uh you know

he used to run Thomas Jefferson

University the health system um you know

he calls this the epidemic which is like

there's just this proliferation of of

work tax in every organ in the country

in IT departments in you know in the

physicians day-to-day and it didn't

solve the problem it set out to solve

because the software was built from the

perspective of the CIO's office and the

CFO's office

>> who are optimized around getting paid

basically.

>> Exactly. optimized around getting paid

and not getting sued and and those are

the you know the two the two drivers you

know the CLLO's office as well now I

think you're seeing a lot of

physician-driven adoption of these tools

one postco we really stress the system

out we we burnt we burnt out our

physicians or burnt out our nurses and

there's almost no option but for them to

use these tools uh uh and and and push

for the adoption of these tools in order

to just keep up with their with their

day-to-day patient volumes our loudest

advocates even in largest health systems

are physicians.

>> Well, this is really fascinating. That

feels like a huge difference, right? I

imagine in selling the first version of

EHRs five or six years ago, you're Yeah.

largely selling to central IT teams,

CIOS, that type of thing. Here, you're

saying today when you approach selling

your recent software, it's an individual

doctor who's your advocate.

>> 100%. We have we have a self-s serve

ambient documentation tool called

Scribe. And

>> yeah, how does that work actually? Can

you tell us a bit about that?

>> I would say it's one of the explosions

in software categories recently. You've

probably had in terms of like really

fast LLM adoption, I think you've had

coding tools like the winds surfs and

the cursors of the world and then you

you've had ambient documentation tools

um in the in the healthcare world. And

ambient documentation listens to the

conversation that a patient and a

physician have. It summarizes that and

then it generates all of the

documentation based on being trained on

previous approvals, previous pieces of

documentation and basically hands it off

to the revenue cycle teams ready to go.

>> It's like a perfect problem for LM. This

thing LM would be really good.

>> It's a perfect problem for LM. You have

transcription, you have summarization,

you have, you know, references and

citations back to clinical source

material all in this one, you know, one

set of models. Uh, and that tool, we

went from 2023 doing maybe 100,000

appointments through that tool to this

year, we'll probably do 20 to 25 million

appointments through that tool. There's

a self-s serve motion that has gone from

last year zero appointments to this year

5 million appointments. Um,

>> self-s serve motion as in you don't go

through the hospital IT team at It's a

physician finds it online. Uh we

advertise directly to them. They sign up

often paying with their own credit card.

Um

>> that seems new.

>> It seems crazy. I mean I was shocked

when when we saw this working. Uh it was

this is not behavior you've seen in

healthcare before.

>> And from that we you know we then go

upsell almost like Slack or Dropbox or

one of these traditional software tools

the whole enterprise where we pitch the

integrations into their EMR the whole

workflow tool. Um, and that's been a

massive, you know, source of of of

expansion for us.

>> If I'm a founder, you know, I might be

thinking, well, isn't it going to be

really annoying in terms of HIPPA and in

terms of all these rules and whatnot in

terms of actually

>> getting a doctor to sign up? Like, it's

not even intuitive to me that a doctor

is allowed to self-sign up for this type

of thing. How do you guys think about

that? Is that has something changed that

makes it easy or was it never actually

as hard as people thought?

>> It's a good question. And I think

there's definitely a lot of compliance

departments in hospitals that are very

much against self- adopted software, but

the productivity gains are so great that

it's happening one way or another. Um,

and the options that a compliance

department at a hospital has is either,

you know, sign up and make this part of

the institution or, you know, tell your

physicians to continue being burnt out.

And usually they're picking the one that

that ends up with happier physicians.

You know, there are totally ways for

there to be HIPPA compliant self-s

served tools. Now, that might defy the

individual policies of a of a private

institution like like a hospital itself,

but from a pure HIPO compliance

standpoint, right?

>> It's not actually a legal risk. It's

just the hospital's own compliance.

>> It's their own internal policies.

>> Yeah. And so, what does it actually look

like in terms of other technological

problems you face? Like what are some

>> maybe just diving into some technical

challenges? Yeah. Um I can imagine for

something like scribing at least to me

intuitively that seems relatively

straightforward as a technical challenge

but perhaps at that scale actually not

so much whether in that one or in some

of the other products you guys have what

are some of the hardest technical

problems you guys face in actually

making software here especially for a

company that is selling to healthcare as

opposed to any other enterprise SAS

company

>> in just the case of scribing simple at

first glance but when you scale you know

you're dealing with millions of hours of

audio that is often being uploaded over

shaky networks in the basement of doc in

you know a hospital in a lab you know

pharmacy that's a you know freestanding

part of a freestanding clinic in rural

America so you have to build great

offline mode and you know retention and

and and the ability to you know upload

these things silently in the background

and all these like traditional

challenges that consumer apps have

solved I think are now being solved in

healthcare as well. Number two, you

know, we process billions of dollars

worth of claims volume every year in our

revenue cycle business. And as part of a

revenue cycle is it's stripe for

healthcare, right? You're the interface

between the physician and the payer and

the payers at a 100 blocks that make it

hard to get paid. Everything from losing

the claim to, you know, not telling you

that it was denied for a couple weeks so

you don't hit your, you know, so you hit

your timely filing limit and then can't

get paid anymore. We've built models

that are constantly checking for all of

these threats. Uh I almost see it as

like a like a cyber security problem.

Every denial is is something broke or

this is the payer trying to attack the

physician. We have to debug what went

wrong and help the physician get paid

out. A good example is all these payers

have APIs now, right?

>> They just work 80% of the time. What

what that means is is that you might get

responses that are, you know, missing

some of the adjudication or there's a

there's a secondary insurance that needs

to pay out and then you need to go

separately hit that secondary insurance

or these two insurance companies aren't

talking to each other and they both got

a medical record but are claiming they

didn't and you have to show them that

audit trail. And so the technical

challenges are are very interesting

because you're doing this at scale and

you need to monitor these processes

constantly. Um but when you solve them,

physicians win, which is very exciting.

>> Yeah. And especially today if they are

your advocates. I mean I think you get

this even I imagine much faster sales

motion than probably in the early days

where if you can show them these

technical solutions and they can

actually advocate and like get your

thing incorporated into their hospital.

I mean that's huge for you guys.

>> I mean for our revenue cycle business

the word of mouth is literally my

revenue went up by 15%. And that is the

most potent word of mouth in in the

world and that's why we're seeing that

business scale so quickly. Um, I think

in our ambient business, the word of

mouth is I save two hours every day and

I did it with this free tool that I

found online and if I want to get it

integrated, it's, you know, like a

hundred bucks a month or whatever the

price is. So that that's to me why why

why I think these tools are spreading so

quickly now.

>> And how do you think about yourself in

terms of a the types of software you

guys have? You know, you call yourself

sort of a compound software company,

sort of a rippling for healthcare as

we've talked about in the past.

>> What's the basis for that strategy? you

know, how do you guys think about growth

and why grow this way versus doubling

down on one of these and trying to make

that be a massive business in itself? I

>> I think the complexities of healthcare

and the way that people get paid in

healthcare and really where the

operating cash flow lies really make it

like you can only really build a hundred

billion dollar business in healthcare as

a platform company. That being said, you

have to start as a point solution. We

started as a med device and then

eventually turned into this again a

point solution for a very specific

subset of claims and then over time

turned into this platform. Uh but now

that we're this platform, we also have

amazing distribution unlocks. We get to

work with General Catalyst and their

health assurance framework. They have 40

health systems that are all signed up.

HCA is on our board. It's hundred

billion dollar hospital empire. And so I

think in the same way that in the 2010s,

Palunteer solved distribution and

defense and a lot of call it like

Fortune 100 uh use cases and then just

aggregated smart people to go work on

crazy problems. We're doing that in

healthcare. The hardest problem in

healthcare having now spent eight nine

years in it is distribution and you can

reinvent the distribution wheel over

eight nine years for yourself or you can

go build you know effectively point

solutions as part of this platform um

and not worry about the revenue engine

because the backend revenue cycle has

that covered

>> right as long as you ultimately have

revenue cycle management at some part of

your platform. Exactly. Kind of works.

>> Exactly. I mean do you think that

changes now in this era of kind of

doctorled advocacy of acquiring software

you know if if for example it is much

easier for hospitals to acquire software

than ever before especially smaller

clinics does that change the calculus

there at all or do you think uh

ultimately the same dynamics lie

>> what you will see is a lot of these like

small couple millionaire ARR businesses

um that might be distributed in this

somewhat novel way now that physicians

realize the productivity gains far

outweigh you know getting wrap on your

knuckles from compliance. It It's an

approach. The issue is is that the

number of companies, there's like 40

different ambient scribing companies.

They all hit like 100 200k in revenue

and then plateau out and then the next

kind of great wall that they hit is, you

know, there's probably a dozen of them

that got to a couple million in revenue

and then they all plateaued out as well.

It's not a venture business in my

opinion. Like a single point solution

unless you rapidly expand distribution

or you rapidly expand your platform,

it's not a venture scale business. I can

imagine then how you think about

counterpositioning to what might be many

new upstart companies thinking about

this like do you think of it as kind of

as long as we own revenue cycle

management you know these folks can

basically help us innovate on companies

on on new areas and basically you guys

would go acquire them

>> yeah our our strategy has been you know

bundle as much as possible in terms of

we think long term

>> it's kind of the original Microsoft

business model too

>> it is 100% the original Microsoft

business model and a great example is um

in the 80s there was a software that

came I'll call Grammatic and it was an a

word correcting software or like a

grammar checking software and you know

it ripped like it it really took off and

and it looked amazing. I think whatever

the equivalent of VCs were in the back,

you know, back then like were super

excited about the business like I'm sure

they raised a lot of money from a lot of

banks or whoever was funding them and

then Microsoft came out with autocorrect

as part of Word and this business just

died and and this will happen

unfortunately to a lot of these LLM

rapper businesses today and particularly

in healthcare. I think there's a whole

host of vanilla point solution scribing

companies that have developed and the

minute Epic turns on their own native

ambient scribe, those businesses will go

poof because they have no connection to

the, you know, revenue cycle, the

payment stack or anything more

meaningful than just that one layer.

>> And actually on on the note of Epic, I

mean, how do you think about

counterpositioning to them? I mean,

they're sort of the kings of

distribution in terms of healthcare, at

least for the last decade.

>> Presumably, you're trying to take their

crown. Um, but how do you think about

counterpositioning against their

existing distribution?

>> Everything I just said is exactly what

Epic would say about us.

>> Yeah, exactly. And so the the way that I

think about it is one of our core

company values is speed. And there are a

group of health systems in this country

who can't wait for Epic to build these

solutions for them over a multi-year

product roadmap. Uh I think the best

stat that I have here is there you know

the two best run health systems in the

country from a four you know from a

profitability standpoint are HCA which

is like 12 billion in free cash flow and

then tenant which is also billions in

free cash flow. Neither of these

companies are are on Epic. In fact they

avoid Epic like the plague. And the

reason is is that Epic often sucks out

all of the operating cash flow of the

academics and businesses that they work

with. You have these like multiundred

million dollar implementations. And the

the way that I see it is if you want to

be a fastmoving growing business, you

probably can't be on Epic because the

CIO's office becomes this captured asset

and they're just waiting for Epic to

release stuff or at least Epic can't be

central to your strategy. Maybe you use

them for your EMR, but you're you're

building the system of engagement and

other tools on top of it and around it.

Um, and and we aim to be that platform

for that system of engagement on top of

the EMR.

>> And it seems like you're innovating on a

very different go to market compared to

them, too. To your point, they're

probably going to the CIO's office.

That's right. You guys, it sounds like

experimenting with ways to go straight

to doctors and then kind of make your

way to the CIO's office from doctor

demand as opposed to from some large

enterprise deal that's forced on

everyone.

>> And ultimately, you have to earn the

trust of the CIO. There's no doubt about

that. But you can build a lot of

momentum and add a lot of value uh

through this physicianled adoption. You

know, even at our largest health system

partners, we have forward deployed

engineering teams that will sit and, you

know, just work with the medical

directors and the leading physicians in

that facility to try to make their day a

little easier. And that's it. That's all

their job is.

>> I mean, much like you and your

co-founder in the early days.

>> Exactly. We try to replicate what worked

in the in the early days into these

forward deployed pods and so far it's

yielded like we a lot of our net new

products and a lot of our net new

initiatives have come from that. Maybe

changing gears a little bit at the end

here is I'd love to just talk a little

bit about how you see the future of

healthcare broadly. I mean you obviously

have had a front row seat to both

participating in and shaping what this

field has looked like for the last

several years. Now on on the flip side

from from my perspective as a patient um

it feels like not a lot has changed from

my experience as a patient over the last

let's say 20 years. So, you know, if I

were to go to a doctor in 2005 versus

2015 versus today,

>> sure, more stuff is on a computer, but

kind of fundamentally I'm kind of

waiting for the same amount of time and

talking to more or less the same people

and maybe there's a couple new things,

but for the most part, it feels very

similar. One, do you think that's an

accurate characterization? And then

also, how do you think of the next 10

years? Like, do you think it's going to

look kind of similar for patients this

upcoming decade?

>> I I think for the most part, I I agree,

which is that the core of the experience

hasn't changed. I think we're starting

to see innovation now on the fringes and

that there's a lot of things that you

had to go in for that you can now do via

teleahalth visit.

>> Co sort of accelerated a lot of this

>> definitely accelerated a lot of that and

and that's great. I think

>> a system that is inherently supply

constraint the more that you can take

out of you know having to actually show

up at this facility and and and you know

wait for someone to see you the better

it is both for the system as well as for

the patient. Where I think we're

starting to see now uh rapid adoption is

the physician's life is starting to

actually become easier in that these

LLMs are saving them time provably, you

know, a couple hours every day. And

because of that, they're either going

to, you know, their panels will open up,

they'll be able to see more patients. So

getting an appointment with a specialist

hopefully starts feeling a little

easier, you know, moving forward. But

then number two, I think on the whole

these systems are starting to adopt AI

based solutions for things like their

front desk to assist with you know call

automation um the you know patient

engagement. We have a system that for

colonoscopy prep it's called engage. It

will send you a reminder and converse

with you based on the source material

and recommendation from the physician

and it boosts it and improves no-show

rates by 50%. um because the the prep

work is done and you know you don't end

up having to cancel the appointment last

minute because you messed something up

and and so those metrics are all moving

in the right direction. I think where

what healthcare should become with LLMs

in the next couple years is, you know,

you should get your most critical care

in a hospital or in a, you know, in

front of a physician where there really

is something like like, you know, phys

physiootherapy, like a P like a PT

actually taking care of you. Um, or

surgeon operating,

>> a surgeon or a deep complex diagnosis

that requires multiple tests in one day.

Those things should happen in in an

office. But with LLMs, more and more

care, I think can transition out of the

home with sensors that monitor the

patient, virtual care. Um, and and I

think that world is finally becoming

real.

>> And what do you see the transition in

that world looking like? I mean, there's

all these pretty deeply baked incentive

systems that incentivize bringing people

into the hospital even for relatively

small things cuz you have to bill for

it. To your point, a lot of the revenue

ultimately comes from RCM. Do you see

that the those sort of entrenched

financial motives actually preventing

that future from happening or do you see

that it's going to happen inevitably

just from like kind of this abundant

intelligence age we're we're entering?

>> I think it will happen inevitably from

you know this the abundance of

intelligence that's emerging. I think a

lot of your your fringe care will will

again just be things that patients can

do entirely on their own. And as a

result, health systems will be forced to

really specialize in in doing those high

margin, high dollar procedures that

they're uniquely good at. And because of

that specialization, I think the quality

of that care will improve as well. And

you know, you see things like intuitive

surgical and like robotic care actually

the fact that it's a thing in in in the

states is is is pretty remarkable. So I

think at the at the highest end of our

spectrum technology has in fact improved

quality of care meaningfully. There's

like new imunotherapies cleared every

year etc. And then at the very lowest

end of acuity in terms of basic

tellahalth that's improved as well. And

it's this kind of middle that's now

going to massive middle

>> by both by both sides. And hopefully

that the rate at which you know that

that middle gets compressed will be

fast. I mean one of the areas you see

the kind of big research labs most

excited about is you know like Microsoft

for example put out this thing called

medical super intelligence on the path

to medical super intelligence recently

and there's this sort of excitement

around the possibility of actually um in

many ways kind of replacing or or at

least very heavily augmenting but

ultimately fundamentally in the long

term replacing the kind of fundamental

kind of diagnostic process that a lot of

doctors do with

>> super intelligent AI systems you know

ones that could order tests and then

respond to feedback back and iteratively

iteratively work on those. I mean how do

you think about those? Like do you do do

you see that being as part of the

future? And also if that were you know

what do we need to do now to make that

future happen? Like what's kind of

missing in the current world to make

that happen?

>> When it comes to ordering or when it

comes to assessing a test and then

recommending a plan of care these models

will probably surpass humans and if they

have if not like already have have

surpassed humans. the the idea is that

just the way that our regulatory system

is set up from a malpractice standpoint,

litigation standpoint, a billing

standpoint, you probably still want a

responsible party, I think these health

systems will be the mega adopters of a

lot of these tools of intelligence

because it will probably help them from

a malpractice standpoint. If you can

point to a model has synthesized

millions of data points and come to this

conclusion and we are, you know,

basically going through its

recommendations, that's a way better

world than doctor kind of stuck his

finger in the air and guessed what was

happening. And so I I see physicians

embracing what we call these super

intelligent co-pilots that synthesize

information and and take action. You

will probably still for complex things

need to see the physician, get that

final assessment and then the speed and

accuracy at which you go from not

knowing what you have to then something

happening. I think that will get

compressed which is net good. The

reimbursement modalities, I'm I'm not

the world's biggest believer in this

like value based care world. And you

know, we've been talking about it for

like 15 years. I I just think there's

something about simple capitalism that

works well in America which is here's a

service here's a cost and you know we

will improve the accuracy with which we

render XYZ services and maybe that you

know that requires better auditing or

that requires these super intelligence

systems not recommending unnecessary or

incorrect medical procedures but this

procedure driven healthcare world seems

to work like America does have good

healthcare outcomes relative to the rest

of the world.

>> Yeah. And and perhaps, you know, if

there's significantly less unnecessary

care happening where people are able to

triage themselves. Yes. In conversation

with some sort of super intelligent

>> chat system or whatever that system

might be in the future with video and

with sensors and whatnot.

>> Perhaps that opens up the opportunity

for medical professionals to focus on

the cases where you can actually deliver

really great outcomes versus clogging up

ERS with people who don't need to be

there.

>> What 100%.

>> What kind of future do you think that

actually looks like 10 years from now

then? like do you see yourself do you

think we're going to go to the doctor

every year? Do you think we're going to

um only go when like we actually have a

catastrophic issue only or something in

between?

>> I I think in 10 years you're going to

have a lot you're going to have sensors

kind of built into the home. You're

going to have sensors that are you know

wearables on patients. The Apple Watch

and AirPods are becoming medical medical

devices very quickly. And that's a

that's a good thing because that passive

care where you know we can detect things

like aphib and you know detect spikes in

certain biomarkers is really good

because that's that's how you treat them

very very early. the concept of health

insurance and the concept of how care is

rendered is going to look a lot more

like how our cars are are handled in the

sense of like auto insurance

>> catastrophic insurance

>> catastrophic and and so when something

bad you break your arm you have cancer

or you have a you know very complex

disease there will be an insurance

mechanism and really your insurance will

be focused on that kind of care and then

I think the rest of the care it's

actually cheaper for it to be out of

pocket and for it to be you know outside

of the loop and complexity of the

insurance system where you can just get

a direct prescription or you you know

see a provider on teleahalth or even in

person for basic therapy um and because

the system will have shrunk and we have

deleted parts it will become a more

efficient system

>> and do you also see it reverting kind of

back to that version you said in the 80s

and 90s it was great to be a doctor so

>> do you see in many ways the sort of

consolidation we see happening in this

field will at least slow down and

perhaps even revert back to seeing more

people open private practices

>> I I really think so I I I would love a

world where the market cap of United

Health is you know a fifth

>> but every doctor is a millionaire Like

that is a good world for America because

patients are getting treated better. The

people that are actually rendering care

are where value is acrewing. It is

insane to me that if you think about it,

the most valuable healthcare company in

the country, what they effectively do is

like send you a plastic card in the mail

once a year.

>> Yes. There's, you know, it's the straw

man. There's more that they do, but

really at its core, that is your

interaction with Totally.

>> And and then being denied claims perhaps

>> and then denying claims and and so it is

it's an insane system. It needs to

change. Way too much value is accurate

that layer. And I think it needs to come

back to the physician.

>> D, maybe just to wrap things up, you

know, now you've you've been at this for

for quite a while. There's many new

founders excited about thinking about

this current technological moment in

healthcare and AI. You know, what advice

would you give them in terms of how to

compete with someone like you guys? Um,

but also how to build a business that

can endure and kind of capture this

current technological moment. I think

our biggest learning is get in front of

the customer and then just solve

problems for them and and start with one

specific problem even if regardless of

how small it looks. Our first you know

our initial TAM was laughable. It was a

100,000 patients in the entire country

but people and you know our investors

and our team and our you know early you

know founding engineers they took a bet

on us because they believed that we

could take the momentum from that to

build something even bigger. uh and and

so I would just say don't be afraid of

starting very small and very specific

because that is more valuable in the

long term than you know trying to build

a platform from right out the gates. The

other thing that I will say is some of

the best companies being founded right

now are folks that came from the palente

you know did a tour of duty six seven

years at a palunteer solve didn't have

to worry so much about distribution and

just solve problems for customers and

then saw that scale and learned about

the distribution um and so you know my

selfish pitch is like if you're building

in healthcare one of the fastest ways to

both become very wealthy is compound at

a company like Kamir and then also learn

these skills that you would otherwise

have to kind of learn denovo you know,

might take years in in order to amass.

>> Yeah, I totally agree. I mean, I think

one of the best ways to think about it

being becoming a great startup founder

is to work at another great startup, one

that's growing and one where you can get

exposure to customers. I think that kind

of forward deployed role you described

feels a lot like the experience of an

early founder, maybe with perhaps less

total independence, but a lot of the

same skill set that you might gain.

>> 100%. Yeah.

>> And ultimately focusing on making things

people want. Makes sense.

>> Yep. Make things people want.

>> Awesome. Thanks so much. Thanks for

joining us. Great.

>> Appreciate it.

[Music]

Key Vocabulary

Vocabulary	Meanings
prototype /ˈprəʊtətaɪp/ B2	noun - an early sample or model of a product used to test a concept
revenue /ˈrevənjuː/ B2	noun - income generated from normal business activities
automation /ˌɔːtəˈmeɪʃən/ C1	noun - use of technology to perform tasks without human intervention
clinical /ˈklɪnɪkəl/ B2	adjective - relating to the observation and treatment of actual patients
diagnosis /ˌdaɪəɡˈnoʊsɪs/ B2	noun - identification of the nature of an illness or problem
technology /tekˈnɒlədʒi/ B2	noun - the application of scientific knowledge for practical purposes
model /ˈmɒdəl/ B2	noun - a simplified representation used to explain or predict something
claims /kleɪmz/ B2	noun - requests for payment submitted to an insurer
billing /ˈbɪlɪŋ/ B2	noun - the process of preparing and sending invoices
workflow /ˈwɜːrkfloʊ/ C1	noun - a sequence of steps needed to complete a task
engagement /ɪnˈɡeɪdʒmənt/ C1	noun - active participation or involvement, especially of users or customers
platform /ˈplætˌfɔːrm/ B2	noun - a base of technologies that allows other applications to run
innovation /ˌɪnəˈveɪʃən/ C1	noun - the introduction of new ideas, methods or products
scaling /ˈskeɪlɪŋ/ C1	noun - the process of increasing a system’s capacity to handle more work
distribution /ˌdɪstrɪˈbjuːʃən/ C1	noun - the act of delivering products or services to users
adoption /əˈdɒpʃən/ C1	noun - the act of starting to use a new technology or practice
iteration /ˌɪtərˈeɪʃən/ C1	noun - repeating a process in order to improve a product or result
deployment /dɪˈplɔɪmənt/ C1	noun - the act of putting a system or software into active use
analytics /ˌænəˈlɪtɪks/ C1	noun - the systematic analysis of data or statistics
compliance /kəmˈplaɪəns/ C1	noun - conforming to rules, standards, or laws

Key Grammar Structures

• I actually applied immediately. I got rejected because I was in high school.

  ➔ Past Simple (Passive Voice)

  ➔ Use of the past simple passive voice to indicate that the subject (I) was the recipient of the action (rejected). "I **got rejected**" shows that something was done *to* the subject.

• The market is a great I mean it's just like punch in the face. Like it teaches you a lot like very quickly.

  ➔ Simile

  ➔ Use of the simile "**like a punch in the face**" to vividly describe the harsh and impactful nature of the market's lessons.

• I think our biggest learning is get in front of the customer and then just solve problems for them and and start with one specific problem regardless of how small it looks.

  ➔ Imperative Mood

  ➔ Use of the imperative mood to give direct advice or instructions. "**Get** in front of the customer," "**solve** problems," "**start** with one specific problem" are all direct commands.

• We went from having no revenue to having 3 or 4 million in revenue in like a year.

  ➔ Gerunds

  ➔ Use of gerunds ("**having** no revenue" and "**having** 3 or 4 million") as objects of the preposition "from" and "to" respectively. This indicates a transition between two states.

• It's basically Stripe for healthcare and doctors.

  ➔ Metaphor

  ➔ Use of the metaphor "**Stripe for healthcare**" to provide a concise and understandable comparison, indicating that the service is analogous to Stripe in its function within the healthcare industry.

• They saw us as more than just a device company.

  ➔ Comparative Adjective

  ➔ Use of the comparative adjective "**more** than" to show a higher degree of value or perception compared to being "just a device company."