8d6b3f72-c2f0-4422-8631-62277aa4ab86/7-0
00:00:03.567 --> 00:00:07.410
today with AVPs at Memorial Hermann
Health System.

8d6b3f72-c2f0-4422-8631-62277aa4ab86/7-1
00:00:07.410 --> 00:00:13.137
Corey Pace is the AVP of Business
Development and Patient Solutions for the

8d6b3f72-c2f0-4422-8631-62277aa4ab86/7-2
00:00:13.137 --> 00:00:17.432
Institute of Rehabilitation and Research,
known as TIER.

8d6b3f72-c2f0-4422-8631-62277aa4ab86/7-3
00:00:17.432 --> 00:00:23.309
And I'm here with Chaitanya Vampati,
the AVP of AI and Analytics for Memorial

8d6b3f72-c2f0-4422-8631-62277aa4ab86/7-4
00:00:23.309 --> 00:00:24.967
Hermann Health System.

8d6b3f72-c2f0-4422-8631-62277aa4ab86/8-0
00:00:25.047 --> 00:00:26.407
Thank you both for joining today.

8d6b3f72-c2f0-4422-8631-62277aa4ab86/9-0
00:00:26.727 --> 00:00:28.407
Thank you. Thank you.

8d6b3f72-c2f0-4422-8631-62277aa4ab86/6-0
00:00:26.967 --> 00:00:27.607
Thank you.

8d6b3f72-c2f0-4422-8631-62277aa4ab86/11-0
00:00:28.087 --> 00:00:34.337
For those who don't know, as background,
Memorial Hermann Health System is a 14

8d6b3f72-c2f0-4422-8631-62277aa4ab86/11-1
00:00:34.337 --> 00:00:38.399
hospital, 4,
300 bed health system headquartered in

8d6b3f72-c2f0-4422-8631-62277aa4ab86/11-2
00:00:38.399 --> 00:00:42.931
Houston, Texas.
And today I'd like to talk with Corey and

8d6b3f72-c2f0-4422-8631-62277aa4ab86/11-3
00:00:42.931 --> 00:00:48.087
Chaitanya about a Gen AI pilot that is
being implemented at TIER.

8d6b3f72-c2f0-4422-8631-62277aa4ab86/12-0
00:00:48.767 --> 00:00:53.994
in order to address a very cumbersome
problem. So I'd like to invite Corey,

8d6b3f72-c2f0-4422-8631-62277aa4ab86/12-1
00:00:53.994 --> 00:00:59.222
could you please elaborate upon the
problem and its origins before we delve

8d6b3f72-c2f0-4422-8631-62277aa4ab86/12-2
00:00:59.222 --> 00:01:01.767
into the solution and implementation?

8d6b3f72-c2f0-4422-8631-62277aa4ab86/13-0
00:01:02.007 --> 00:01:05.275
Yeah, absolutely. So I,
as part of my role,

8d6b3f72-c2f0-4422-8631-62277aa4ab86/13-1
00:01:05.275 --> 00:01:08.544
I am over our clinical admissions for
tier.

8d6b3f72-c2f0-4422-8631-62277aa4ab86/13-2
00:01:08.544 --> 00:01:13.150
And part of that process includes
evaluating patient records,

8d6b3f72-c2f0-4422-8631-62277aa4ab86/13-3
00:01:13.150 --> 00:01:18.795
looking at their entire hospital stay to
see if they are appropriate for an

8d6b3f72-c2f0-4422-8631-62277aa4ab86/13-4
00:01:18.795 --> 00:01:20.727
inpatient rehab admission.

8d6b3f72-c2f0-4422-8631-62277aa4ab86/15-0
00:01:21.527 --> 00:01:25.618
Especially at TIER,
we take a lot of really complex patients.

8d6b3f72-c2f0-4422-8631-62277aa4ab86/15-1
00:01:25.618 --> 00:01:30.303
And so their hospital stay might have
been month, even more sometimes.

8d6b3f72-c2f0-4422-8631-62277aa4ab86/15-2
00:01:30.303 --> 00:01:35.714
And so their records are quite extensive.
And we have nurses that are responsible

8d6b3f72-c2f0-4422-8631-62277aa4ab86/15-3
00:01:35.714 --> 00:01:40.267
for looking through those,
finding all of the things that would make

8d6b3f72-c2f0-4422-8631-62277aa4ab86/15-4
00:01:40.267 --> 00:01:42.247
that person rehab appropriate.

8d6b3f72-c2f0-4422-8631-62277aa4ab86/16-0
00:01:42.567 --> 00:01:46.808
or finding things that might be a barrier
to admission.

8d6b3f72-c2f0-4422-8631-62277aa4ab86/16-1
00:01:46.808 --> 00:01:51.959
We get external records sent to us that
are hundreds of pages long.

8d6b3f72-c2f0-4422-8631-62277aa4ab86/16-2
00:01:51.959 --> 00:01:56.655
We had an example the other day that was
over 800 pages long.

8d6b3f72-c2f0-4422-8631-62277aa4ab86/16-3
00:01:56.655 --> 00:02:01.502
And so as you can imagine,
it is a large task for our nurses to

8d6b3f72-c2f0-4422-8631-62277aa4ab86/16-4
00:02:01.502 --> 00:02:02.487
review those.

8d6b3f72-c2f0-4422-8631-62277aa4ab86/17-0
00:02:02.887 --> 00:02:08.352
and to successfully and efficiently find
all of the data that they need.

8d6b3f72-c2f0-4422-8631-62277aa4ab86/17-1
00:02:08.352 --> 00:02:13.892
So there is a large opportunity for error
when we get records that large.

8d6b3f72-c2f0-4422-8631-62277aa4ab86/17-2
00:02:13.892 --> 00:02:18.010
And it also takes us time.
And we all know acute care,

8d6b3f72-c2f0-4422-8631-62277aa4ab86/17-3
00:02:18.010 --> 00:02:22.727
a big focus of acute, like for discharges,
turnaround time for

8d6b3f72-c2f0-4422-8631-62277aa4ab86/18-0
00:02:22.967 --> 00:02:27.214
getting those patients to their next
level of care in a timely manner.

8d6b3f72-c2f0-4422-8631-62277aa4ab86/18-1
00:02:27.214 --> 00:02:30.204
And we don't ever want to be the barrier
to that.

8d6b3f72-c2f0-4422-8631-62277aa4ab86/18-2
00:02:30.204 --> 00:02:34.511
So we're always looking at how can we
increase that time, or I'm sorry,

8d6b3f72-c2f0-4422-8631-62277aa4ab86/18-3
00:02:34.511 --> 00:02:38.817
decrease the time that it takes from
referral to clearing that patient.

8d6b3f72-c2f0-4422-8631-62277aa4ab86/18-4
00:02:38.817 --> 00:02:42.167
And so how do we do that without adding
more people? We

8d6b3f72-c2f0-4422-8631-62277aa4ab86/19-0
00:02:42.207 --> 00:02:46.279
look to technology to help us.
And so I put together a proposal and

8d6b3f72-c2f0-4422-8631-62277aa4ab86/19-1
00:02:46.279 --> 00:02:49.093
essentially took it to our system saying,
hey,

8d6b3f72-c2f0-4422-8631-62277aa4ab86/19-2
00:02:49.093 --> 00:02:53.884
I really want us to look at something
that will help us dig through the charts,

8d6b3f72-c2f0-4422-8631-62277aa4ab86/19-3
00:02:53.884 --> 00:02:58.315
quickly put together a summary,
help us figure out what may or may not be

8d6b3f72-c2f0-4422-8631-62277aa4ab86/19-4
00:02:58.315 --> 00:03:02.567
missing so that we can quickly go back to
the referral source and say,

8d6b3f72-c2f0-4422-8631-62277aa4ab86/21-0
00:03:02.767 --> 00:03:08.902
we are missing yesterday's therapy notes.
And then also look at what are their labs,

8d6b3f72-c2f0-4422-8631-62277aa4ab86/20-0
00:03:06.727 --> 00:03:07.127
Yeah.

8d6b3f72-c2f0-4422-8631-62277aa4ab86/21-1
00:03:08.902 --> 00:03:15.110
what are their vitals, what are, you know,
any sort of imaging that we're waiting on,

8d6b3f72-c2f0-4422-8631-62277aa4ab86/21-2
00:03:15.110 --> 00:03:21.029
and then hopefully help us determine if
they are appropriate for that rehab level

8d6b3f72-c2f0-4422-8631-62277aa4ab86/21-3
00:03:21.029 --> 00:03:21.607
of care.

8d6b3f72-c2f0-4422-8631-62277aa4ab86/23-0
00:03:21.847 --> 00:03:24.745
So Corey,
you noticed the problem at Memorial

8d6b3f72-c2f0-4422-8631-62277aa4ab86/23-1
00:03:24.745 --> 00:03:28.965
Hermann that was facing your clinical
admissions team. Eventually,

8d6b3f72-c2f0-4422-8631-62277aa4ab86/23-2
00:03:28.965 --> 00:03:33.942
Gautic Chaitanya and he began formulating
a solution. But before we get there,

8d6b3f72-c2f0-4422-8631-62277aa4ab86/23-3
00:03:33.942 --> 00:03:38.289
I think our listeners may be interested
in understanding a bit more.

8d6b3f72-c2f0-4422-8631-62277aa4ab86/23-4
00:03:38.289 --> 00:03:41.880
Is there a system,
a process at Memorial Hermann whereby

8d6b3f72-c2f0-4422-8631-62277aa4ab86/23-5
00:03:41.880 --> 00:03:44.967
AVPs are empowered to come up with ideas
and ask

8d6b3f72-c2f0-4422-8631-62277aa4ab86/28-0
00:03:45.087 --> 00:03:48.258
for budget or solutions from the health
system leadership team?

8d6b3f72-c2f0-4422-8631-62277aa4ab86/22-0
00:03:47.367 --> 00:03:47.847
Yeah.

8d6b3f72-c2f0-4422-8631-62277aa4ab86/28-1
00:03:48.258 --> 00:03:51.775
And how does that process work?
Because I think a lot of our listeners

8d6b3f72-c2f0-4422-8631-62277aa4ab86/28-2
00:03:51.775 --> 00:03:54.055
are saying, yeah,
I am encountering problems,

8d6b3f72-c2f0-4422-8631-62277aa4ab86/28-3
00:03:54.055 --> 00:03:57.424
but I don't know what to do about it.
Maybe I don't have a personal

8d6b3f72-c2f0-4422-8631-62277aa4ab86/24-0
00:03:54.167 --> 00:03:54.487
Yeah.

8d6b3f72-c2f0-4422-8631-62277aa4ab86/28-4
00:03:57.424 --> 00:04:00.892
relationship with the system CTO or COO.
How did you go about saying,

8d6b3f72-c2f0-4422-8631-62277aa4ab86/25-0
00:03:58.767 --> 00:03:58.967
Yeah.

8d6b3f72-c2f0-4422-8631-62277aa4ab86/28-5
00:04:00.892 --> 00:04:04.955
I'm hearing from my team that there's a
problem and I want to escalate and see if

8d6b3f72-c2f0-4422-8631-62277aa4ab86/26-0
00:04:02.327 --> 00:04:02.447
Yeah.

8d6b3f72-c2f0-4422-8631-62277aa4ab86/28-6
00:04:04.955 --> 00:04:06.887
we can find a solution? Did you just...

8d6b3f72-c2f0-4422-8631-62277aa4ab86/27-0
00:04:06.807 --> 00:04:07.127
Yeah.

8d6b3f72-c2f0-4422-8631-62277aa4ab86/29-0
00:04:07.047 --> 00:04:09.607
hear about Gen AI,
how did that process work?

8d6b3f72-c2f0-4422-8631-62277aa4ab86/30-0
00:04:10.567 --> 00:04:16.167
So we actually have a really great
platform for anybody within our system to

8d6b3f72-c2f0-4422-8631-62277aa4ab86/30-1
00:04:16.167 --> 00:04:21.549
submit ideas. It's called iGenerate.
You can put in anybody from, I mean,

8d6b3f72-c2f0-4422-8631-62277aa4ab86/30-2
00:04:21.549 --> 00:04:25.549
any level of our organization can submit
an idea, hey,

8d6b3f72-c2f0-4422-8631-62277aa4ab86/30-3
00:04:25.549 --> 00:04:28.167
this seems like a really cool thing.

8d6b3f72-c2f0-4422-8631-62277aa4ab86/31-0
00:04:28.567 --> 00:04:31.535
submit it,
and then we have a fun little thing

8d6b3f72-c2f0-4422-8631-62277aa4ab86/31-1
00:04:31.535 --> 00:04:36.525
called Shark Tank that once, you know,
they all get vetted through our system.

8d6b3f72-c2f0-4422-8631-62277aa4ab86/31-2
00:04:36.525 --> 00:04:41.704
We have hundreds of ideas and then take
kind of the top ideas that look like they

8d6b3f72-c2f0-4422-8631-62277aa4ab86/31-3
00:04:41.704 --> 00:04:46.567
might be really successful for multiple
patient experience, safety, revenue.

8d6b3f72-c2f0-4422-8631-62277aa4ab86/32-0
00:04:47.847 --> 00:04:54.462
And then it and it goes to this Shark
Tank style thing where you present it to

8d6b3f72-c2f0-4422-8631-62277aa4ab86/32-1
00:04:54.462 --> 00:05:00.240
to our executive level team that has
representatives from, you know,

8d6b3f72-c2f0-4422-8631-62277aa4ab86/32-2
00:05:00.240 --> 00:05:04.845
our ISD finance,
kind of the spectrum of our C-suites.

8d6b3f72-c2f0-4422-8631-62277aa4ab86/32-3
00:05:04.845 --> 00:05:09.367
And that my idea got taken to that to
then be able to

8d6b3f72-c2f0-4422-8631-62277aa4ab86/35-0
00:05:09.607 --> 00:05:13.477
get the system level support needed.
I will also say each,

8d6b3f72-c2f0-4422-8631-62277aa4ab86/33-0
00:05:12.447 --> 00:05:12.927
Mhm.

8d6b3f72-c2f0-4422-8631-62277aa4ab86/35-1
00:05:13.477 --> 00:05:17.151
if you are not going through that process,
we all have,

8d6b3f72-c2f0-4422-8631-62277aa4ab86/35-2
00:05:17.151 --> 00:05:22.399
each campus has their own process where
you can take things up through your own

8d6b3f72-c2f0-4422-8631-62277aa4ab86/35-3
00:05:22.399 --> 00:05:26.204
leadership team.
And then we have individual campus level

8d6b3f72-c2f0-4422-8631-62277aa4ab86/35-4
00:05:26.204 --> 00:05:29.287
meetings with our ISD teams. Shatanya,
you can

8d6b3f72-c2f0-4422-8631-62277aa4ab86/36-0
00:05:29.407 --> 00:05:32.407
talk more to about other projects in our
AI council.

8d6b3f72-c2f0-4422-8631-62277aa4ab86/34-0
00:05:29.687 --> 00:05:30.167
Yeah.

8d6b3f72-c2f0-4422-8631-62277aa4ab86/37-0
00:05:33.047 --> 00:05:33.287
No.

8d6b3f72-c2f0-4422-8631-62277aa4ab86/38-0
00:05:33.367 --> 00:05:36.391
So, Chaitanya,
can you talk about the intake process?

8d6b3f72-c2f0-4422-8631-62277aa4ab86/38-1
00:05:36.391 --> 00:05:40.759
How did you become aware of this project?
Did you sit on the Shark Tank pitch

8d6b3f72-c2f0-4422-8631-62277aa4ab86/38-2
00:05:40.759 --> 00:05:44.007
session? And if not,
who did and how did they filter this

8d6b3f72-c2f0-4422-8631-62277aa4ab86/38-3
00:05:44.007 --> 00:05:45.127
project down to you?

8d6b3f72-c2f0-4422-8631-62277aa4ab86/40-0
00:05:45.847 --> 00:05:49.097
Yeah,
I think the team presented it to the

8d6b3f72-c2f0-4422-8631-62277aa4ab86/40-1
00:05:49.097 --> 00:05:54.541
Shark Tank. And we have, I mean,
we refer to our Chief Digital Officer.

8d6b3f72-c2f0-4422-8631-62277aa4ab86/40-2
00:05:54.541 --> 00:05:59.152
The Chief Digital Officer typically sits
on that Shark Tank.

8d6b3f72-c2f0-4422-8631-62277aa4ab86/40-3
00:05:59.152 --> 00:06:04.822
We also have people representatives,
project managers who sit on the Shark

8d6b3f72-c2f0-4422-8631-62277aa4ab86/40-4
00:06:04.822 --> 00:06:09.055
Tank and so on.
So this was one that was highlighted as

8d6b3f72-c2f0-4422-8631-62277aa4ab86/39-0
00:06:06.567 --> 00:06:07.047
Mhm.

8d6b3f72-c2f0-4422-8631-62277aa4ab86/40-5
00:06:09.055 --> 00:06:09.887
some things

8d6b3f72-c2f0-4422-8631-62277aa4ab86/43-0
00:06:09.967 --> 00:06:14.983
as we are building up our maturity in AI,
especially agentic solutions and using

8d6b3f72-c2f0-4422-8631-62277aa4ab86/41-0
00:06:14.207 --> 00:06:14.887
Mmh.

8d6b3f72-c2f0-4422-8631-62277aa4ab86/43-1
00:06:14.983 --> 00:06:19.628
large language models internally,
was something that was passed down to me

8d6b3f72-c2f0-4422-8631-62277aa4ab86/43-2
00:06:19.628 --> 00:06:23.716
from those teams who were sitting on
those platforms saying, hey,

8d6b3f72-c2f0-4422-8631-62277aa4ab86/43-3
00:06:23.716 --> 00:06:28.794
is this something that is even possible
or even feasible to build out internally?

8d6b3f72-c2f0-4422-8631-62277aa4ab86/43-4
00:06:28.794 --> 00:06:31.767
Because most health systems today,
when they...

8d6b3f72-c2f0-4422-8631-62277aa4ab86/44-0
00:06:31.887 --> 00:06:37.406
When they think about agentic solutions,
a lot of them are not thinking, "Hey,

8d6b3f72-c2f0-4422-8631-62277aa4ab86/44-1
00:06:37.406 --> 00:06:41.946
can we even build that first?
" Like a lot of them are utilizing

8d6b3f72-c2f0-4422-8631-62277aa4ab86/44-2
00:06:41.946 --> 00:06:45.439
vendors,
utilizing either startups or established

8d6b3f72-c2f0-4422-8631-62277aa4ab86/44-3
00:06:45.439 --> 00:06:50.748
EMR systems and tools within them to be
able to use them for these types of

8d6b3f72-c2f0-4422-8631-62277aa4ab86/44-4
00:06:50.748 --> 00:06:51.447
solutions.

8d6b3f72-c2f0-4422-8631-62277aa4ab86/45-0
00:06:51.767 --> 00:06:56.180
And this was something that was passed
down because of the novelty of the

8d6b3f72-c2f0-4422-8631-62277aa4ab86/45-1
00:06:56.180 --> 00:07:00.354
particular solution and how useful it was
going to be for the system.

8d6b3f72-c2f0-4422-8631-62277aa4ab86/45-2
00:07:00.354 --> 00:07:04.887
So that's how the project came about to
at least our group to be evaluated.

8d6b3f72-c2f0-4422-8631-62277aa4ab86/46-0
00:07:05.407 --> 00:07:09.687
So you decided,
how did you decide to build this solution

8d6b3f72-c2f0-4422-8631-62277aa4ab86/46-1
00:07:09.687 --> 00:07:14.778
instead of purchase it from a vendor?
And second, what Gen AI model,

8d6b3f72-c2f0-4422-8631-62277aa4ab86/46-2
00:07:14.778 --> 00:07:20.312
are you leveraging an open source Gen AI
model to build your own solution?

8d6b3f72-c2f0-4422-8631-62277aa4ab86/46-3
00:07:20.312 --> 00:07:24.887
Or are you leveraging a commercial clot
or Open AI or Gemini?

8d6b3f72-c2f0-4422-8631-62277aa4ab86/47-0
00:07:25.447 --> 00:07:25.927
Product.

8d6b3f72-c2f0-4422-8631-62277aa4ab86/49-0
00:07:26.327 --> 00:07:28.760
Yeah,
let me talk to the first part of that

8d6b3f72-c2f0-4422-8631-62277aa4ab86/49-1
00:07:28.760 --> 00:07:32.244
particular question.
So I think constantly Corey and I go back

8d6b3f72-c2f0-4422-8631-62277aa4ab86/49-2
00:07:32.244 --> 00:07:35.838
and forth as soon as we see anything in
the industry, like, hey,

8d6b3f72-c2f0-4422-8631-62277aa4ab86/49-3
00:07:35.838 --> 00:07:40.151
is this something that can replace,
you know, our argument what we are doing,

8d6b3f72-c2f0-4422-8631-62277aa4ab86/49-4
00:07:40.151 --> 00:07:44.520
right? Like, so that is a constant search.
Corey has done tremendous research.

8d6b3f72-c2f0-4422-8631-62277aa4ab86/49-5
00:07:44.520 --> 00:07:47.727
I've done research to see if there is
anything out there.

8d6b3f72-c2f0-4422-8631-62277aa4ab86/50-0
00:07:48.007 --> 00:07:53.146
or even off-the-shelf AI systems like
Claude or Copilot or anything like that,

8d6b3f72-c2f0-4422-8631-62277aa4ab86/48-0
00:07:48.207 --> 00:07:48.687
Mhm.

8d6b3f72-c2f0-4422-8631-62277aa4ab86/50-1
00:07:53.146 --> 00:07:58.416
can they do the same thing that we are
trying to do through a series of prompts,

8d6b3f72-c2f0-4422-8631-62277aa4ab86/50-2
00:07:58.416 --> 00:08:03.165
right? So after that evaluation,
it became fairly evident that this is a

8d6b3f72-c2f0-4422-8631-62277aa4ab86/50-3
00:08:03.165 --> 00:08:07.328
very specific use case.
Tier is a very specialized hospital and

8d6b3f72-c2f0-4422-8631-62277aa4ab86/50-4
00:08:07.328 --> 00:08:11.687
the workflow that they're looking at for
the rehab intake process.

8d6b3f72-c2f0-4422-8631-62277aa4ab86/52-0
00:08:12.327 --> 00:08:17.689
is a very specific workflow that requires
some thought process to be built in.

8d6b3f72-c2f0-4422-8631-62277aa4ab86/51-0
00:08:16.647 --> 00:08:17.127
Mhm.

8d6b3f72-c2f0-4422-8631-62277aa4ab86/52-1
00:08:17.689 --> 00:08:21.558
Once we've decided it,
and because of the novelty of the

8d6b3f72-c2f0-4422-8631-62277aa4ab86/52-2
00:08:21.558 --> 00:08:25.020
solution,
we chose to think of it like an internal

8d6b3f72-c2f0-4422-8631-62277aa4ab86/52-3
00:08:25.020 --> 00:08:30.450
lean startup, right? Like an agile team.
And one of the first things we did was

8d6b3f72-c2f0-4422-8631-62277aa4ab86/52-4
00:08:30.450 --> 00:08:32.487
that, and Cody remembers this,

8d6b3f72-c2f0-4422-8631-62277aa4ab86/54-0
00:08:33.127 --> 00:08:35.818
We said, hey, Corey,
we can't take on this yet.

8d6b3f72-c2f0-4422-8631-62277aa4ab86/54-1
00:08:35.818 --> 00:08:39.631
Let us do some discovery internally and
think this process through.

8d6b3f72-c2f0-4422-8631-62277aa4ab86/53-0
00:08:39.607 --> 00:08:39.767
Mm.

8d6b3f72-c2f0-4422-8631-62277aa4ab86/54-2
00:08:39.631 --> 00:08:42.603
Let's look at the architecture and think
it through.

8d6b3f72-c2f0-4422-8631-62277aa4ab86/54-3
00:08:42.603 --> 00:08:47.033
Out of that came out the answers to your
question, Jordan, which is like, hey,

8d6b3f72-c2f0-4422-8631-62277aa4ab86/54-4
00:08:47.033 --> 00:08:50.453
what model should we use?
Should it be an open source model?

8d6b3f72-c2f0-4422-8631-62277aa4ab86/54-5
00:08:50.453 --> 00:08:51.687
How can we pitch this?

8d6b3f72-c2f0-4422-8631-62277aa4ab86/55-0
00:08:52.087 --> 00:08:56.542
technologies to view.
So to put things in perspective,

8d6b3f72-c2f0-4422-8631-62277aa4ab86/55-1
00:08:56.542 --> 00:09:00.754
like so we use AWS Bedrock as one of the,
you know,

8d6b3f72-c2f0-4422-8631-62277aa4ab86/55-2
00:09:00.754 --> 00:09:07.072
secure LLM orchestration platforms,
along with internal AIML tools that we've

8d6b3f72-c2f0-4422-8631-62277aa4ab86/55-3
00:09:07.072 --> 00:09:11.041
built up.
So we use for this particular intake a

8d6b3f72-c2f0-4422-8631-62277aa4ab86/55-4
00:09:11.041 --> 00:09:11.527
mix of

8d6b3f72-c2f0-4422-8631-62277aa4ab86/56-0
00:09:11.847 --> 00:09:15.184
open source models,
some things from OpenAI,

8d6b3f72-c2f0-4422-8631-62277aa4ab86/56-1
00:09:15.184 --> 00:09:21.043
and primarily the big chunk of heavy
lifting is done by Anthropic and Claude's

8d6b3f72-c2f0-4422-8631-62277aa4ab86/56-2
00:09:21.043 --> 00:09:26.976
model coming out of AWS Bedrock in a
secure instance hosted by Memorial Herman.

8d6b3f72-c2f0-4422-8631-62277aa4ab86/56-3
00:09:26.976 --> 00:09:32.167
So those are the types of things that we
use to orchestrate this. And

8d6b3f72-c2f0-4422-8631-62277aa4ab86/59-0
00:09:32.487 --> 00:09:37.203
A lot of work we do,
we did here is not sort of like 1 prompt,

8d6b3f72-c2f0-4422-8631-62277aa4ab86/59-1
00:09:37.203 --> 00:09:41.620
like take this,
run it through one prompt and do it. Yeah,

8d6b3f72-c2f0-4422-8631-62277aa4ab86/57-0
00:09:40.927 --> 00:09:41.407
Mhm.

8d6b3f72-c2f0-4422-8631-62277aa4ab86/59-2
00:09:41.620 --> 00:09:46.336
when we first looked at it, you know,
the volume of the pages,

8d6b3f72-c2f0-4422-8631-62277aa4ab86/59-3
00:09:46.336 --> 00:09:51.127
the types of faxes we are getting in for
these were so complex.

8d6b3f72-c2f0-4422-8631-62277aa4ab86/58-0
00:09:51.367 --> 00:09:51.527
Mm.

8d6b3f72-c2f0-4422-8631-62277aa4ab86/60-0
00:10:03.327 --> 00:10:03.847
Mhm.

8d6b3f72-c2f0-4422-8631-62277aa4ab86/61-0
00:10:03.887 --> 00:10:07.927
several iterations to be able to even
architecture it. And then we decided like,

8d6b3f72-c2f0-4422-8631-62277aa4ab86/61-1
00:10:07.927 --> 00:10:10.571
hey, let's bring on the nurses.
Let's work together.

8d6b3f72-c2f0-4422-8631-62277aa4ab86/61-2
00:10:10.571 --> 00:10:12.167
We think we can accomplish this.

8d6b3f72-c2f0-4422-8631-62277aa4ab86/63-0
00:10:12.487 --> 00:10:17.628
So I'd like to delve now,
and I think our listeners would like to

8d6b3f72-c2f0-4422-8631-62277aa4ab86/63-1
00:10:17.628 --> 00:10:21.056
hear you delve now into a dual
explanation.

8d6b3f72-c2f0-4422-8631-62277aa4ab86/63-2
00:10:21.056 --> 00:10:25.107
How did you collaborate between both of
your teams?

8d6b3f72-c2f0-4422-8631-62277aa4ab86/63-3
00:10:25.107 --> 00:10:31.028
I think there's a lot of times solutions
break down when technology and the

8d6b3f72-c2f0-4422-8631-62277aa4ab86/63-4
00:10:31.028 --> 00:10:33.287
business owners are in silos.

8d6b3f72-c2f0-4422-8631-62277aa4ab86/68-0
00:10:33.487 --> 00:10:37.331
and the business owner makes a request,
and technology runs off and builds a

8d6b3f72-c2f0-4422-8631-62277aa4ab86/62-0
00:10:33.687 --> 00:10:34.007
Yeah.

8d6b3f72-c2f0-4422-8631-62277aa4ab86/68-1
00:10:37.331 --> 00:10:39.977
solution,
and then a year later comes back and says,

8d6b3f72-c2f0-4422-8631-62277aa4ab86/68-2
00:10:39.977 --> 00:10:43.221
voila, here we are. And they're like, oh,
that's a little wrong,

8d6b3f72-c2f0-4422-8631-62277aa4ab86/65-0
00:10:41.127 --> 00:10:41.607
Mhm.

8d6b3f72-c2f0-4422-8631-62277aa4ab86/64-0
00:10:41.207 --> 00:10:41.447
The.

8d6b3f72-c2f0-4422-8631-62277aa4ab86/68-3
00:10:43.221 --> 00:10:45.917
or that's not a thing,
and you didn't get the buy-in,

8d6b3f72-c2f0-4422-8631-62277aa4ab86/68-4
00:10:45.917 --> 00:10:49.162
and people aren't sure about it,
and then it just kind of fails.

8d6b3f72-c2f0-4422-8631-62277aa4ab86/68-5
00:10:49.162 --> 00:10:52.407
But how could you put,
I think our listeners would like to hear,

8d6b3f72-c2f0-4422-8631-62277aa4ab86/67-0
00:10:51.927 --> 00:10:52.167
Yeah.

8d6b3f72-c2f0-4422-8631-62277aa4ab86/70-0
00:10:53.127 --> 00:10:57.768
Exactly how, once, Chaitanya,
once your team decided to go ahead and

8d6b3f72-c2f0-4422-8631-62277aa4ab86/70-1
00:10:57.768 --> 00:11:01.535
build something,
how did you collaborate with Corey and

8d6b3f72-c2f0-4422-8631-62277aa4ab86/70-2
00:11:01.535 --> 00:11:06.647
her team to ensure that there is a
sufficient amount of clinical buy-in and

8d6b3f72-c2f0-4422-8631-62277aa4ab86/70-3
00:11:06.647 --> 00:11:11.827
that you were addressing the actual needs
of how do we make a patient visit,

8d6b3f72-c2f0-4422-8631-62277aa4ab86/70-4
00:11:11.827 --> 00:11:14.047
I think the metrics was turnover.

8d6b3f72-c2f0-4422-8631-62277aa4ab86/71-0
00:11:14.127 --> 00:11:19.287
from admission to next stage of clinical
care more efficient and more successful.

8d6b3f72-c2f0-4422-8631-62277aa4ab86/69-0
00:11:18.567 --> 00:11:19.047
No.

8d6b3f72-c2f0-4422-8631-62277aa4ab86/73-0
00:11:21.447 --> 00:11:27.736
I can start and then hand it over to
Corey and you can basically fill in the

8d6b3f72-c2f0-4422-8631-62277aa4ab86/72-0
00:11:26.247 --> 00:11:26.487
Yeah.

8d6b3f72-c2f0-4422-8631-62277aa4ab86/73-1
00:11:27.736 --> 00:11:34.597
missing pieces if I may. So it was clear,
like this problem was fairly challenging,

8d6b3f72-c2f0-4422-8631-62277aa4ab86/73-2
00:11:34.597 --> 00:11:38.110
right?
And it was also clear that the data

8d6b3f72-c2f0-4422-8631-62277aa4ab86/73-3
00:11:38.110 --> 00:11:40.887
science and AI group does not know

8d6b3f72-c2f0-4422-8631-62277aa4ab86/75-0
00:11:40.967 --> 00:11:44.715
the clinical nuances that Corey and her
team at TIER know, right?

8d6b3f72-c2f0-4422-8631-62277aa4ab86/75-1
00:11:44.715 --> 00:11:48.804
So we decided like this is not going to
be something that's going to be

8d6b3f72-c2f0-4422-8631-62277aa4ab86/75-2
00:11:48.804 --> 00:11:53.176
successful if we want to follow it.
If we can go take this as a requirements

8d6b3f72-c2f0-4422-8631-62277aa4ab86/75-3
00:11:53.176 --> 00:11:56.470
document and go inside and build
something and come back,

8d6b3f72-c2f0-4422-8631-62277aa4ab86/75-4
00:11:56.470 --> 00:12:00.332
it was never going to work.
So what we said was that we're going to

8d6b3f72-c2f0-4422-8631-62277aa4ab86/75-5
00:12:00.332 --> 00:12:01.127
meet biweekly.

8d6b3f72-c2f0-4422-8631-62277aa4ab86/76-0
00:12:01.607 --> 00:12:05.752
and weekly sometimes.
And we need people like nurses,

8d6b3f72-c2f0-4422-8631-62277aa4ab86/76-1
00:12:05.752 --> 00:12:10.972
actual users on the call weekly.
So we said we'll do it in an agile

8d6b3f72-c2f0-4422-8631-62277aa4ab86/76-2
00:12:10.972 --> 00:12:14.887
fashion.
The objective would be to bring something

8d6b3f72-c2f0-4422-8631-62277aa4ab86/76-3
00:12:14.887 --> 00:12:20.337
of useful nature to each of these
meetings. So let's work on for some,

8d6b3f72-c2f0-4422-8631-62277aa4ab86/76-4
00:12:20.337 --> 00:12:22.487
identify 3, 4 things to fix.

8d6b3f72-c2f0-4422-8631-62277aa4ab86/77-0
00:12:22.527 --> 00:12:27.833
small things that should only take a week
or two to fix, identify those, fix them,

8d6b3f72-c2f0-4422-8631-62277aa4ab86/77-1
00:12:27.833 --> 00:12:31.541
and show up with the working prototype at
these meetings.

8d6b3f72-c2f0-4422-8631-62277aa4ab86/77-2
00:12:31.541 --> 00:12:36.272
And the key was for us to start with the
working prototype right from the

8d6b3f72-c2f0-4422-8631-62277aa4ab86/77-3
00:12:36.272 --> 00:12:39.341
beginning.
We had something rudimentary working

8d6b3f72-c2f0-4422-8631-62277aa4ab86/77-4
00:12:39.341 --> 00:12:42.345
almost weeks into the project.
And since then,

8d6b3f72-c2f0-4422-8631-62277aa4ab86/77-5
00:12:42.345 --> 00:12:44.647
it was all about constant iterating.

8d6b3f72-c2f0-4422-8631-62277aa4ab86/78-0
00:12:45.527 --> 00:12:50.167
And it is imperative and I think I want
to reiterate what you said.

8d6b3f72-c2f0-4422-8631-62277aa4ab86/80-0
00:12:51.127 --> 00:12:55.703
Solutions such as this will not be
successful if they are built in silos.

8d6b3f72-c2f0-4422-8631-62277aa4ab86/80-1
00:12:55.703 --> 00:13:00.279
And it is actually to be talked through
that the business is building the

8d6b3f72-c2f0-4422-8631-62277aa4ab86/80-2
00:13:00.279 --> 00:13:02.938
solution,
utilizing the technology and the

8d6b3f72-c2f0-4422-8631-62277aa4ab86/80-3
00:13:02.938 --> 00:13:07.204
engineers behind the scenes.
They are just listening to the constant

8d6b3f72-c2f0-4422-8631-62277aa4ab86/79-0
00:13:06.407 --> 00:13:06.847
No.

8d6b3f72-c2f0-4422-8631-62277aa4ab86/80-4
00:13:07.204 --> 00:13:11.224
feedback and iterating.
And that's the only way to be successful

8d6b3f72-c2f0-4422-8631-62277aa4ab86/80-5
00:13:11.224 --> 00:13:14.687
in this domain. Corey,
you might want to add something.

8d6b3f72-c2f0-4422-8631-62277aa4ab86/81-0
00:13:14.007 --> 00:13:18.298
Yeah, I think that was great summary.
We definitely have the weekly,

8d6b3f72-c2f0-4422-8631-62277aa4ab86/81-1
00:13:18.298 --> 00:13:21.781
bi-weekly meeting,
bi-weekly and then sometimes weekly.

8d6b3f72-c2f0-4422-8631-62277aa4ab86/81-2
00:13:21.781 --> 00:13:26.756
We have a Zoom channel that the nurses
are also included on where every single,

8d6b3f72-c2f0-4422-8631-62277aa4ab86/81-3
00:13:26.756 --> 00:13:30.799
like I get things popping up all the time.
Hey, this file is in,

8d6b3f72-c2f0-4422-8631-62277aa4ab86/81-4
00:13:30.799 --> 00:13:32.727
I'm not seeing it yet. And then

8d6b3f72-c2f0-4422-8631-62277aa4ab86/83-0
00:13:32.887 --> 00:13:35.676
the engineer, oh, yep,
here's what's going on.

8d6b3f72-c2f0-4422-8631-62277aa4ab86/83-1
00:13:35.676 --> 00:13:38.763
And so they'll fix it real time.
In these meetings,

8d6b3f72-c2f0-4422-8631-62277aa4ab86/83-2
00:13:38.763 --> 00:13:43.511
it's been really helpful to just pull up
a few summaries and the nurses will be

8d6b3f72-c2f0-4422-8631-62277aa4ab86/83-3
00:13:43.511 --> 00:13:47.310
like, this is amazing.
What would be even more amazing is if we

8d6b3f72-c2f0-4422-8631-62277aa4ab86/83-4
00:13:47.310 --> 00:13:50.337
had X, Y,
and Z and the team will either say, yes,

8d6b3f72-c2f0-4422-8631-62277aa4ab86/83-5
00:13:50.337 --> 00:13:53.127
we can do that immediately or that might
be an

8d6b3f72-c2f0-4422-8631-62277aa4ab86/84-0
00:13:53.167 --> 00:13:58.916
a more of an optimization phase.
They have worked very closely with our

8d6b3f72-c2f0-4422-8631-62277aa4ab86/84-1
00:13:58.916 --> 00:14:03.548
nursing team to really kind of fine tune
the information.

8d6b3f72-c2f0-4422-8631-62277aa4ab86/84-2
00:14:03.548 --> 00:14:09.777
The nurses have really spent a lot of
time validating the information that is

8d6b3f72-c2f0-4422-8631-62277aa4ab86/84-3
00:14:09.777 --> 00:14:15.127
being pulled together to make sure that
it's not missing things or

8d6b3f72-c2f0-4422-8631-62277aa4ab86/85-0
00:14:15.447 --> 00:14:20.328
or giving incorrect information.
And it's been really from the beginning,

8d6b3f72-c2f0-4422-8631-62277aa4ab86/85-1
00:14:20.328 --> 00:14:22.967
pretty accurate from the very beginning.

8d6b3f72-c2f0-4422-8631-62277aa4ab86/86-0
00:14:23.367 --> 00:14:26.472
Corey, what fears did your nurses,
physicians,

8d6b3f72-c2f0-4422-8631-62277aa4ab86/86-1
00:14:26.472 --> 00:14:31.691
and other clinical staff have at the
beginning of this process when they first

8d6b3f72-c2f0-4422-8631-62277aa4ab86/86-2
00:14:31.691 --> 00:14:33.607
started talking to Chaitanya?

8d6b3f72-c2f0-4422-8631-62277aa4ab86/87-0
00:14:34.487 --> 00:14:40.727
I think I probably had the most fears of
we take a lot of really specialized

8d6b3f72-c2f0-4422-8631-62277aa4ab86/87-1
00:14:40.727 --> 00:14:46.400
patients at this hospital that a lot of
other rehab facilities don't.

8d6b3f72-c2f0-4422-8631-62277aa4ab86/87-2
00:14:46.400 --> 00:14:53.207
And so what I really wanted to make sure
of is that as we were building this model,

8d6b3f72-c2f0-4422-8631-62277aa4ab86/88-0
00:14:53.607 --> 00:14:57.299
that we weren't, and we don't to this day,
have it say, yes,

8d6b3f72-c2f0-4422-8631-62277aa4ab86/88-1
00:14:57.299 --> 00:15:01.656
they're appropriate for rehab or no,
they're not appropriate for rehab.

8d6b3f72-c2f0-4422-8631-62277aa4ab86/88-2
00:15:01.656 --> 00:15:06.135
Because what I don't want is for our
clinical teams to lose that decision

8d6b3f72-c2f0-4422-8631-62277aa4ab86/88-3
00:15:06.135 --> 00:15:08.919
making and that clinical judgment
themselves.

8d6b3f72-c2f0-4422-8631-62277aa4ab86/88-4
00:15:08.919 --> 00:15:12.127
I want it to pull the information
together for them.

8d6b3f72-c2f0-4422-8631-62277aa4ab86/89-0
00:15:12.567 --> 00:15:15.809
and then then be able to make that
decision of, yes,

8d6b3f72-c2f0-4422-8631-62277aa4ab86/89-1
00:15:15.809 --> 00:15:20.214
this is an appropriate patient.
It may not be a typical patient that as

8d6b3f72-c2f0-4422-8631-62277aa4ab86/89-2
00:15:20.214 --> 00:15:23.089
of right now,
our tool might say would be more

8d6b3f72-c2f0-4422-8631-62277aa4ab86/89-3
00:15:23.089 --> 00:15:27.861
appropriate for another level of care
because we do take those patients here.

8d6b3f72-c2f0-4422-8631-62277aa4ab86/89-4
00:15:27.861 --> 00:15:32.633
But that may be down the road as it
learns more and as we are feeding it more

8d6b3f72-c2f0-4422-8631-62277aa4ab86/89-5
00:15:32.633 --> 00:15:33.367
information,

8d6b3f72-c2f0-4422-8631-62277aa4ab86/90-0
00:15:33.607 --> 00:15:39.061
it does start picking up those nuances
for those patients that are appropriate

8d6b3f72-c2f0-4422-8631-62277aa4ab86/90-1
00:15:39.061 --> 00:15:43.687
but not typical for rehab.
So I think that was kind of my fear is.

8d6b3f72-c2f0-4422-8631-62277aa4ab86/91-0
00:15:45.367 --> 00:15:51.311
Does our team lose that clinical decision?
Do they rely on it too much? Do they,

8d6b3f72-c2f0-4422-8631-62277aa4ab86/91-1
00:15:51.311 --> 00:15:56.815
as we have new people come in and they
are relying solely on this, I mean,

8d6b3f72-c2f0-4422-8631-62277aa4ab86/91-2
00:15:56.815 --> 00:16:01.878
I think this is true for all of AI,
myself included in how I use it,

8d6b3f72-c2f0-4422-8631-62277aa4ab86/91-3
00:16:01.878 --> 00:16:04.887
but do we lose our own clinical decision?

8d6b3f72-c2f0-4422-8631-62277aa4ab86/92-0
00:16:05.127 --> 00:16:08.247
judgment because we're relying so much on
it.

8d6b3f72-c2f0-4422-8631-62277aa4ab86/93-0
00:16:08.967 --> 00:16:11.960
Could you,
we've talked a lot about getting to the

8d6b3f72-c2f0-4422-8631-62277aa4ab86/93-1
00:16:11.960 --> 00:16:14.836
solution.
Just for the benefit of our listeners,

8d6b3f72-c2f0-4422-8631-62277aa4ab86/93-2
00:16:14.836 --> 00:16:19.062
can you provide an overview of what the
solution actually does from the

8d6b3f72-c2f0-4422-8631-62277aa4ab86/93-3
00:16:19.062 --> 00:16:20.647
perspective of an end user?

8d6b3f72-c2f0-4422-8631-62277aa4ab86/94-0
00:16:21.047 --> 00:16:25.632
Yeah, it's really cool. Essentially,
it gives us the very first page is a

8d6b3f72-c2f0-4422-8631-62277aa4ab86/94-1
00:16:25.632 --> 00:16:30.650
checklist of all of the things that we
have said is required for an admission or

8d6b3f72-c2f0-4422-8631-62277aa4ab86/94-2
00:16:30.650 --> 00:16:34.058
for a referral.
And it will tell us immediately out of

8d6b3f72-c2f0-4422-8631-62277aa4ab86/94-3
00:16:34.058 --> 00:16:37.651
these 800 pages,
all of these elements are here and these

8d6b3f72-c2f0-4422-8631-62277aa4ab86/94-4
00:16:37.651 --> 00:16:41.616
three things are missing.
And so our intake team can quickly go

8d6b3f72-c2f0-4422-8631-62277aa4ab86/94-5
00:16:41.616 --> 00:16:43.847
back to the referral source and say,

8d6b3f72-c2f0-4422-8631-62277aa4ab86/96-0
00:16:44.167 --> 00:16:47.011
hey, we've got your referral,
we're processing it,

8d6b3f72-c2f0-4422-8631-62277aa4ab86/96-1
00:16:47.011 --> 00:16:50.971
but we also need these three data
elements so that we can finalize the

8d6b3f72-c2f0-4422-8631-62277aa4ab86/95-0
00:16:49.007 --> 00:16:49.527
Mhm.

8d6b3f72-c2f0-4422-8631-62277aa4ab86/96-2
00:16:50.971 --> 00:16:55.656
process. And then, and then it just goes,
we have given it our pre-admission screen

8d6b3f72-c2f0-4422-8631-62277aa4ab86/96-3
00:16:55.656 --> 00:16:59.282
data, what we're looking for,
and it goes head to toe. Here are,

8d6b3f72-c2f0-4422-8631-62277aa4ab86/96-4
00:16:59.282 --> 00:17:02.907
here's the head to toe summary of the
patient present, you know,

8d6b3f72-c2f0-4422-8631-62277aa4ab86/96-5
00:17:02.907 --> 00:17:03.967
summary of present.

8d6b3f72-c2f0-4422-8631-62277aa4ab86/99-0
00:17:04.087 --> 00:17:11.147
hospitalization, any lines, tubes, drains,
labs, imaging, vitals,

8d6b3f72-c2f0-4422-8631-62277aa4ab86/99-1
00:17:11.147 --> 00:17:19.384
kind of just the whole summary in about
3:00 to 5:00 pages, maybe, Shatanya,

8d6b3f72-c2f0-4422-8631-62277aa4ab86/97-0
00:17:18.647 --> 00:17:19.207
Yes.

8d6b3f72-c2f0-4422-8631-62277aa4ab86/99-2
00:17:19.384 --> 00:17:22.487
instead of hundreds of pages.

8d6b3f72-c2f0-4422-8631-62277aa4ab86/100-0
00:17:23.607 --> 00:17:28.165
So then our team is able to look through
that, quickly see it,

8d6b3f72-c2f0-4422-8631-62277aa4ab86/98-0
00:17:24.007 --> 00:17:24.407
Um...

8d6b3f72-c2f0-4422-8631-62277aa4ab86/100-1
00:17:28.165 --> 00:17:34.097
and then it is not integrated into Epic
at this time. And so they have a summary,

8d6b3f72-c2f0-4422-8631-62277aa4ab86/100-2
00:17:34.097 --> 00:17:39.885
and then they're using that summary to
help complete their documentation within

8d6b3f72-c2f0-4422-8631-62277aa4ab86/100-3
00:17:39.885 --> 00:17:40.247
Epic.

8d6b3f72-c2f0-4422-8631-62277aa4ab86/101-0
00:17:40.567 --> 00:17:47.973
How does the model's output compared to
the manual summary of that same 800 page

8d6b3f72-c2f0-4422-8631-62277aa4ab86/101-1
00:17:47.973 --> 00:17:54.190
PDF that you used to do?
Is there like an 80% overlap or are there,

8d6b3f72-c2f0-4422-8631-62277aa4ab86/101-2
00:17:54.190 --> 00:17:59.127
are there, you know,
is there any hallucination? What

8d6b3f72-c2f0-4422-8631-62277aa4ab86/102-0
00:17:59.207 --> 00:18:01.927
Are you seeing,
how does it compare to the manual process?

8d6b3f72-c2f0-4422-8631-62277aa4ab86/104-0
00:18:03.447 --> 00:18:07.519
I think for the most part,
it is much more comprehensive.

8d6b3f72-c2f0-4422-8631-62277aa4ab86/104-1
00:18:07.519 --> 00:18:12.996
What we've heard feedback from the team
is it has caught things that they may

8d6b3f72-c2f0-4422-8631-62277aa4ab86/104-2
00:18:12.996 --> 00:18:18.262
have missed because it was buried in that.
I have not heard, and Shatanya,

8d6b3f72-c2f0-4422-8631-62277aa4ab86/104-3
00:18:18.262 --> 00:18:22.967
I don't know if you have of really any
hallucinations. It has been

8d6b3f72-c2f0-4422-8631-62277aa4ab86/103-0
00:18:20.247 --> 00:18:20.727
Billions.

8d6b3f72-c2f0-4422-8631-62277aa4ab86/105-0
00:18:23.087 --> 00:18:24.647
pretty accurate from the get-go.

8d6b3f72-c2f0-4422-8631-62277aa4ab86/106-0
00:18:25.927 --> 00:18:29.822
Yeah, I mean,
I think a lot of feedback that we've got,

8d6b3f72-c2f0-4422-8631-62277aa4ab86/106-1
00:18:29.822 --> 00:18:33.022
I mean,
and we have a spreadsheet that tracks

8d6b3f72-c2f0-4422-8631-62277aa4ab86/106-2
00:18:33.022 --> 00:18:38.657
every feedback and our manager over it is
very **** about tracking it because we

8d6b3f72-c2f0-4422-8631-62277aa4ab86/106-3
00:18:38.657 --> 00:18:42.831
want to make sure that we don't miss
anything. So it's not,

8d6b3f72-c2f0-4422-8631-62277aa4ab86/106-4
00:18:42.831 --> 00:18:45.127
there's a lot of things that are.

8d6b3f72-c2f0-4422-8631-62277aa4ab86/108-0
00:18:45.287 --> 00:18:48.687
sort of buried in this, you know,
800 page faxes.

8d6b3f72-c2f0-4422-8631-62277aa4ab86/108-1
00:18:48.687 --> 00:18:54.128
It's not always in a standard format or
it's not always even in the same order,

8d6b3f72-c2f0-4422-8631-62277aa4ab86/108-2
00:18:54.128 --> 00:18:57.393
right?
So people tend to miss things and it has

8d6b3f72-c2f0-4422-8631-62277aa4ab86/107-0
00:18:55.807 --> 00:18:56.287
Mhm.

8d6b3f72-c2f0-4422-8631-62277aa4ab86/108-3
00:18:57.393 --> 00:19:01.950
caught things in some page in the last
few pages of the whole fax.

8d6b3f72-c2f0-4422-8631-62277aa4ab86/108-4
00:19:01.950 --> 00:19:07.390
So we get routinely feedback like, hey,
how amazingly it has caught things that

8d6b3f72-c2f0-4422-8631-62277aa4ab86/108-5
00:19:07.390 --> 00:19:09.567
they would have probably missed.

8d6b3f72-c2f0-4422-8631-62277aa4ab86/109-0
00:19:09.927 --> 00:19:12.476
Now, you asked a very,
very specific question,

8d6b3f72-c2f0-4422-8631-62277aa4ab86/109-1
00:19:12.476 --> 00:19:16.760
and an interesting one on hallucination,
and you asked about the fears of what

8d6b3f72-c2f0-4422-8631-62277aa4ab86/109-2
00:19:16.760 --> 00:19:19.309
this model might do.
And joining both of them,

8d6b3f72-c2f0-4422-8631-62277aa4ab86/109-3
00:19:19.309 --> 00:19:23.538
when we first started on the project,
the AI team was tremendously fearful of

8d6b3f72-c2f0-4422-8631-62277aa4ab86/109-4
00:19:23.538 --> 00:19:26.521
hallucination,
because that was a big thing back then,

8d6b3f72-c2f0-4422-8631-62277aa4ab86/109-5
00:19:26.521 --> 00:19:29.287
right? Like,
because it's a large context of data,

8d6b3f72-c2f0-4422-8631-62277aa4ab86/110-0
00:19:29.527 --> 00:19:33.894
and we are asking specific information,
is it hallucinating from this large?

8d6b3f72-c2f0-4422-8631-62277aa4ab86/110-1
00:19:33.894 --> 00:19:36.616
Like all of the data cannot fit in one
context.

8d6b3f72-c2f0-4422-8631-62277aa4ab86/110-2
00:19:36.616 --> 00:19:40.926
It's not like you could upload this
document new cloud or Chad GPD and say,

8d6b3f72-c2f0-4422-8631-62277aa4ab86/110-3
00:19:40.926 --> 00:19:44.158
summarize this, like it's too big.
So one of the things,

8d6b3f72-c2f0-4422-8631-62277aa4ab86/110-4
00:19:44.158 --> 00:19:48.865
the techniques we used was just chunking
and breaking it into pieces and trying to

8d6b3f72-c2f0-4422-8631-62277aa4ab86/110-5
00:19:48.865 --> 00:19:50.567
make sure that each piece only

8d6b3f72-c2f0-4422-8631-62277aa4ab86/111-0
00:19:51.527 --> 00:19:55.958
extracts information about its piece
rather than thinking through the big

8d6b3f72-c2f0-4422-8631-62277aa4ab86/111-1
00:19:55.958 --> 00:19:59.969
picture and all of that,
and then combining all of that into a big

8d6b3f72-c2f0-4422-8631-62277aa4ab86/111-2
00:19:59.969 --> 00:20:02.903
picture.
So we run through various layers of AI,

8d6b3f72-c2f0-4422-8631-62277aa4ab86/111-3
00:20:02.903 --> 00:20:05.777
not just one AI,
and that has almost eliminated

8d6b3f72-c2f0-4422-8631-62277aa4ab86/111-4
00:20:05.777 --> 00:20:09.190
hallucination.
The biggest concern and the trick that we

8d6b3f72-c2f0-4422-8631-62277aa4ab86/111-5
00:20:09.190 --> 00:20:10.807
had to build out is that...

8d6b3f72-c2f0-4422-8631-62277aa4ab86/113-0
00:20:11.447 --> 00:20:14.634
it missing things.
Like we were very careful about,

8d6b3f72-c2f0-4422-8631-62277aa4ab86/113-1
00:20:14.634 --> 00:20:19.721
like we should not have a situation where
it is present in the document and the AI

8d6b3f72-c2f0-4422-8631-62277aa4ab86/112-0
00:20:18.047 --> 00:20:18.167
Yeah.

8d6b3f72-c2f0-4422-8631-62277aa4ab86/113-2
00:20:19.721 --> 00:20:23.582
thinks it's not there.
We initially had a few cases where that

8d6b3f72-c2f0-4422-8631-62277aa4ab86/113-3
00:20:23.582 --> 00:20:28.240
was happening and we quickly addressed it
by looking at various techniques,

8d6b3f72-c2f0-4422-8631-62277aa4ab86/113-4
00:20:28.240 --> 00:20:29.527
changing models, etc.

8d6b3f72-c2f0-4422-8631-62277aa4ab86/115-0
00:20:29.727 --> 00:20:32.483
You know,
doing the AI techniques that we are

8d6b3f72-c2f0-4422-8631-62277aa4ab86/115-1
00:20:32.483 --> 00:20:35.718
routinely used to,
but we never we were worried about

8d6b3f72-c2f0-4422-8631-62277aa4ab86/115-2
00:20:35.718 --> 00:20:39.014
hallucination,
and I think we engineered a solution to

8d6b3f72-c2f0-4422-8631-62277aa4ab86/115-3
00:20:39.014 --> 00:20:43.687
engineer out the hallucination and make
sure it is accurate most of the time.

8d6b3f72-c2f0-4422-8631-62277aa4ab86/119-0
00:20:44.247 --> 00:20:49.107
I think our listeners might be interested
in hearing, just be asking themselves,

8d6b3f72-c2f0-4422-8631-62277aa4ab86/119-1
00:20:49.107 --> 00:20:53.787
is the primary data source, this PDF,
included in the chart as a reference in

8d6b3f72-c2f0-4422-8631-62277aa4ab86/119-2
00:20:53.787 --> 00:20:57.687
case someone does have a question and
wants to look at page 621?

8d6b3f72-c2f0-4422-8631-62277aa4ab86/116-0
00:20:56.007 --> 00:20:56.327
Yeah.

8d6b3f72-c2f0-4422-8631-62277aa4ab86/117-0
00:20:57.767 --> 00:20:58.327
Yes.

8d6b3f72-c2f0-4422-8631-62277aa4ab86/118-0
00:20:58.727 --> 00:20:58.847
Yeah.

8d6b3f72-c2f0-4422-8631-62277aa4ab86/121-0
00:20:58.887 --> 00:21:05.313
Okay, and okay, and then I want to ask,
and you say it's not integrated into Epic.

8d6b3f72-c2f0-4422-8631-62277aa4ab86/121-1
00:21:05.313 --> 00:21:11.506
How is this information being integrated
into the EHR? How is that information,

8d6b3f72-c2f0-4422-8631-62277aa4ab86/121-2
00:21:11.506 --> 00:21:15.687
like what is the data source from Epic's
perspective?

8d6b3f72-c2f0-4422-8631-62277aa4ab86/123-0
00:21:16.087 --> 00:21:20.964
So our nurse liaisons complete a
pre-admission screen that has fields that

8d6b3f72-c2f0-4422-8631-62277aa4ab86/123-1
00:21:20.964 --> 00:21:25.842
they have to fill in and complete.
And then so they are using this summary

8d6b3f72-c2f0-4422-8631-62277aa4ab86/122-0
00:21:23.087 --> 00:21:23.567
Mhm.

8d6b3f72-c2f0-4422-8631-62277aa4ab86/123-2
00:21:25.842 --> 00:21:31.110
and the like they're validating that the
summary is accurate from the references

8d6b3f72-c2f0-4422-8631-62277aa4ab86/123-3
00:21:31.110 --> 00:21:35.207
that it gives for the for the actual
record. And then they are

8d6b3f72-c2f0-4422-8631-62277aa4ab86/124-0
00:21:35.927 --> 00:21:40.167
basically transposing some of that
information into our pre-admission screen.

8d6b3f72-c2f0-4422-8631-62277aa4ab86/126-0
00:21:40.487 --> 00:21:45.149
And what's the degree of,
is there copying and pasting or is it

8d6b3f72-c2f0-4422-8631-62277aa4ab86/126-1
00:21:45.149 --> 00:21:50.759
really just retyping and transposing
information and to what extent has that

8d6b3f72-c2f0-4422-8631-62277aa4ab86/126-2
00:21:50.759 --> 00:21:53.527
led to errors of any kind, even typos?

8d6b3f72-c2f0-4422-8631-62277aa4ab86/125-0
00:21:53.927 --> 00:21:54.247
Yeah.

8d6b3f72-c2f0-4422-8631-62277aa4ab86/127-0
00:21:55.767 --> 00:21:56.647
Not a big problem.

8d6b3f72-c2f0-4422-8631-62277aa4ab86/130-0
00:21:55.927 --> 00:21:58.614
Um, no,
it hasn't really been a problem that

8d6b3f72-c2f0-4422-8631-62277aa4ab86/130-1
00:21:58.614 --> 00:22:00.167
we've that we've run into.

8d6b3f72-c2f0-4422-8631-62277aa4ab86/129-0
00:21:59.607 --> 00:22:00.007
Yeah.

8d6b3f72-c2f0-4422-8631-62277aa4ab86/132-0
00:22:00.567 --> 00:22:04.286
A quick question I think we can easily
address.

8d6b3f72-c2f0-4422-8631-62277aa4ab86/132-1
00:22:04.286 --> 00:22:10.407
Any difficulty in applying the LLM to a
PDF as opposed to another data format?

8d6b3f72-c2f0-4422-8631-62277aa4ab86/131-0
00:22:08.487 --> 00:22:08.967
Ohh.

8d6b3f72-c2f0-4422-8631-62277aa4ab86/133-0
00:22:10.967 --> 00:22:14.442
Yeah, I think, I mean,
I would say like there is a there is a

8d6b3f72-c2f0-4422-8631-62277aa4ab86/133-1
00:22:14.442 --> 00:22:18.645
difficulty here that you mentioned.
I'm glad you are asking this question.

8d6b3f72-c2f0-4422-8631-62277aa4ab86/133-2
00:22:18.645 --> 00:22:23.073
It is a technical challenge that we spend,
you know, nights and weekends over.

8d6b3f72-c2f0-4422-8631-62277aa4ab86/133-3
00:22:23.073 --> 00:22:25.595
It's not a question about the format of
PDF.

8d6b3f72-c2f0-4422-8631-62277aa4ab86/133-4
00:22:25.595 --> 00:22:29.687
It's just the fact that the PDF are
essentially, you know, faxes like...

8d6b3f72-c2f0-4422-8631-62277aa4ab86/135-0
00:22:29.727 --> 00:22:34.912
files converted into PDFs,
and they come in all kinds of different

8d6b3f72-c2f0-4422-8631-62277aa4ab86/134-0
00:22:31.767 --> 00:22:32.087
Yeah.

8d6b3f72-c2f0-4422-8631-62277aa4ab86/135-1
00:22:34.912 --> 00:22:38.626
formats.
Sometimes they have images embedded as

8d6b3f72-c2f0-4422-8631-62277aa4ab86/135-2
00:22:38.626 --> 00:22:44.585
text, sometimes that is text and images,
sometimes that is only text, right?

8d6b3f72-c2f0-4422-8631-62277aa4ab86/135-3
00:22:44.585 --> 00:22:49.847
So it's the mixture of formats that is
handwritten notes sometimes.

8d6b3f72-c2f0-4422-8631-62277aa4ab86/137-0
00:22:50.087 --> 00:22:55.624
There is all these different permutations
and combinations, imaging results,

8d6b3f72-c2f0-4422-8631-62277aa4ab86/137-1
00:22:55.624 --> 00:22:59.651
and so on. The challenge here,
the technical challenge,

8d6b3f72-c2f0-4422-8631-62277aa4ab86/136-0
00:22:57.207 --> 00:22:57.687
Mhm.

8d6b3f72-c2f0-4422-8631-62277aa4ab86/137-2
00:22:59.651 --> 00:23:05.404
is to be able to have an LLM configured
in such a way that it's able to look at

8d6b3f72-c2f0-4422-8631-62277aa4ab86/137-3
00:23:05.404 --> 00:23:09.287
all of these contiguous things as one
context, right?

8d6b3f72-c2f0-4422-8631-62277aa4ab86/140-0
00:23:09.687 --> 00:23:13.903
You need to take in a bunch of text,
an image in the middle,

8d6b3f72-c2f0-4422-8631-62277aa4ab86/140-1
00:23:13.903 --> 00:23:19.571
and a bunch of text all into one context.
And the image may not be like an X-ray.

8d6b3f72-c2f0-4422-8631-62277aa4ab86/140-2
00:23:19.571 --> 00:23:23.649
The image may be still text scanned in as
an image, right?

8d6b3f72-c2f0-4422-8631-62277aa4ab86/138-0
00:23:21.687 --> 00:23:22.287
Mm-hmm.

8d6b3f72-c2f0-4422-8631-62277aa4ab86/140-3
00:23:23.649 --> 00:23:28.487
So that was the big technical challenge,
not particularly PDF per se.

8d6b3f72-c2f0-4422-8631-62277aa4ab86/139-0
00:23:28.967 --> 00:23:29.127
Mm.

8d6b3f72-c2f0-4422-8631-62277aa4ab86/142-0
00:23:29.967 --> 00:23:34.902
It's all of these various elements
embedded and mixed into one grouping.

8d6b3f72-c2f0-4422-8631-62277aa4ab86/142-1
00:23:34.902 --> 00:23:37.607
That was the biggest challenge to solve.

8d6b3f72-c2f0-4422-8631-62277aa4ab86/141-0
00:23:37.927 --> 00:23:38.087
Ohh.

8d6b3f72-c2f0-4422-8631-62277aa4ab86/143-0
00:23:37.967 --> 00:23:43.473
So it sounds like your LLM has been able
to accommodate a mix of different types,

8d6b3f72-c2f0-4422-8631-62277aa4ab86/143-1
00:23:43.473 --> 00:23:46.361
sizes,
and formats of data sources without

8d6b3f72-c2f0-4422-8631-62277aa4ab86/143-2
00:23:46.361 --> 00:23:51.464
needing to normalize, aggregate,
or deduplicate. Is there any, would there,

8d6b3f72-c2f0-4422-8631-62277aa4ab86/143-3
00:23:51.464 --> 00:23:54.822
is there any,
are you leveraging your integration

8d6b3f72-c2f0-4422-8631-62277aa4ab86/143-4
00:23:54.822 --> 00:24:00.127
engine at all to do that sort of process
before feeding the data into the LLM?

8d6b3f72-c2f0-4422-8631-62277aa4ab86/145-0
00:24:01.007 --> 00:24:06.777
No, I mean, we are basically,
this particular bot essentially looks at

8d6b3f72-c2f0-4422-8631-62277aa4ab86/145-1
00:24:06.777 --> 00:24:13.279
one fax as its own entity. So, you know,
when we get multiple faxes on the same

8d6b3f72-c2f0-4422-8631-62277aa4ab86/144-0
00:24:10.127 --> 00:24:10.727
Mm-hmm.

8d6b3f72-c2f0-4422-8631-62277aa4ab86/145-2
00:24:13.279 --> 00:24:17.180
patient,
we typically they're all combined into

8d6b3f72-c2f0-4422-8631-62277aa4ab86/145-3
00:24:17.180 --> 00:24:18.887
one sort of document.

8d6b3f72-c2f0-4422-8631-62277aa4ab86/147-0
00:24:19.687 --> 00:24:24.509
And we somewhat purposefully designed it
this way as we are building this out,

8d6b3f72-c2f0-4422-8631-62277aa4ab86/147-1
00:24:24.509 --> 00:24:27.867
because, I mean,
we didn't want to build out like huge

8d6b3f72-c2f0-4422-8631-62277aa4ab86/147-2
00:24:27.867 --> 00:24:32.140
integrations and, you know,
we want to see that the confidence in the

8d6b3f72-c2f0-4422-8631-62277aa4ab86/147-3
00:24:32.140 --> 00:24:36.780
output it produces is there before we go
integrate into a bunch of systems.

8d6b3f72-c2f0-4422-8631-62277aa4ab86/147-4
00:24:36.780 --> 00:24:39.527
And routinely,
these are external referrals.

8d6b3f72-c2f0-4422-8631-62277aa4ab86/149-0
00:24:39.847 --> 00:24:44.095
coming into the organization.
So we may not have all of the patient

8d6b3f72-c2f0-4422-8631-62277aa4ab86/146-0
00:24:39.927 --> 00:24:40.407
Mhm.

8d6b3f72-c2f0-4422-8631-62277aa4ab86/149-1
00:24:44.095 --> 00:24:46.407
data internally to normalize anyways.

8d6b3f72-c2f0-4422-8631-62277aa4ab86/148-0
00:24:44.967 --> 00:24:45.127
Yeah.

8d6b3f72-c2f0-4422-8631-62277aa4ab86/150-0
00:24:47.007 --> 00:24:49.750
So we are approaching the end of this
podcast episode,

8d6b3f72-c2f0-4422-8631-62277aa4ab86/150-1
00:24:49.750 --> 00:24:52.393
and I think this is a very interesting
conversation,

8d6b3f72-c2f0-4422-8631-62277aa4ab86/150-2
00:24:52.393 --> 00:24:56.184
as I'm sure our listeners do as well.
I have a few more questions I want to

8d6b3f72-c2f0-4422-8631-62277aa4ab86/150-3
00:24:56.184 --> 00:24:58.877
throw at you,
but I'll throw them at once and let you

8d6b3f72-c2f0-4422-8631-62277aa4ab86/150-4
00:24:58.877 --> 00:25:02.618
kind of pick and choose which ones you'd
like to respond to since we don't

8d6b3f72-c2f0-4422-8631-62277aa4ab86/150-5
00:25:02.618 --> 00:25:04.912
necessarily need to address all of them.
One,

8d6b3f72-c2f0-4422-8631-62277aa4ab86/150-6
00:25:04.912 --> 00:25:07.207
I think some of our listeners may be just.
..

8d6b3f72-c2f0-4422-8631-62277aa4ab86/152-0
00:25:07.287 --> 00:25:11.191
thinking very generally,
how has this impacted clinical workflow

8d6b3f72-c2f0-4422-8631-62277aa4ab86/152-1
00:25:11.191 --> 00:25:14.434
and revenue?
And the second thing I think some of our

8d6b3f72-c2f0-4422-8631-62277aa4ab86/152-2
00:25:14.434 --> 00:25:18.999
listeners may be thinking is there are
tight margins at healthcare delivery

8d6b3f72-c2f0-4422-8631-62277aa4ab86/152-3
00:25:18.999 --> 00:25:23.083
systems across the country,
and they're only becoming tighter since

8d6b3f72-c2f0-4422-8631-62277aa4ab86/152-4
00:25:23.083 --> 00:25:26.327
recent changes at the federal level with
legislation.

8d6b3f72-c2f0-4422-8631-62277aa4ab86/151-0
00:25:25.327 --> 00:25:25.847
Mhm.

8d6b3f72-c2f0-4422-8631-62277aa4ab86/153-0
00:25:27.127 --> 00:25:31.124
I'm wondering to what extent,
many organizations are wondering,

8d6b3f72-c2f0-4422-8631-62277aa4ab86/153-1
00:25:31.124 --> 00:25:34.683
if we build a new software solution at
our organization,

8d6b3f72-c2f0-4422-8631-62277aa4ab86/153-2
00:25:34.683 --> 00:25:38.993
to what extent can we package it and
resell it as a product to other

8d6b3f72-c2f0-4422-8631-62277aa4ab86/153-3
00:25:38.993 --> 00:25:43.676
organizations with the same problem,
since software obviously has a higher

8d6b3f72-c2f0-4422-8631-62277aa4ab86/153-4
00:25:43.676 --> 00:25:46.487
margin than delivery of healthcare
services,

8d6b3f72-c2f0-4422-8631-62277aa4ab86/154-0
00:25:46.727 --> 00:25:51.025
And to what extent can we transform
Memorial Hermann not only to a healthcare

8d6b3f72-c2f0-4422-8631-62277aa4ab86/154-1
00:25:51.025 --> 00:25:54.167
delivery organization,
but to a software product vendor?

8d6b3f72-c2f0-4422-8631-62277aa4ab86/156-0
00:25:56.167 --> 00:25:58.486
Yeah,
I think I'll let Corey at least address

8d6b3f72-c2f0-4422-8631-62277aa4ab86/156-1
00:25:58.486 --> 00:26:00.855
the first part of the question,
which is like,

8d6b3f72-c2f0-4422-8631-62277aa4ab86/155-0
00:25:58.567 --> 00:25:58.887
Yeah.

8d6b3f72-c2f0-4422-8631-62277aa4ab86/156-2
00:26:00.855 --> 00:26:04.686
how is it impacting clinical workflows?
And I think you did mention patient

8d6b3f72-c2f0-4422-8631-62277aa4ab86/156-3
00:26:04.686 --> 00:26:06.955
experience.
We might want to address patient

8d6b3f72-c2f0-4422-8631-62277aa4ab86/156-4
00:26:06.955 --> 00:26:10.887
experience because that's why we did this
all of things to begin with, Corey.

8d6b3f72-c2f0-4422-8631-62277aa4ab86/157-0
00:26:11.327 --> 00:26:14.424
Yeah,
I think patient experience was definitely

8d6b3f72-c2f0-4422-8631-62277aa4ab86/157-1
00:26:14.424 --> 00:26:18.618
one of the top drivers.
And one of the things that we frequently

8d6b3f72-c2f0-4422-8631-62277aa4ab86/157-2
00:26:18.618 --> 00:26:22.037
hear is that,
especially for these complex patients,

8d6b3f72-c2f0-4422-8631-62277aa4ab86/157-3
00:26:22.037 --> 00:26:25.263
it just takes us a while to get to a yes
or a no.

8d6b3f72-c2f0-4422-8631-62277aa4ab86/157-4
00:26:25.263 --> 00:26:30.167
And so we are frequently looking at how
do we speed that up? How do we get?

8d6b3f72-c2f0-4422-8631-62277aa4ab86/158-0
00:26:30.567 --> 00:26:34.355
give them an answer one way or the other
either way.

8d6b3f72-c2f0-4422-8631-62277aa4ab86/158-1
00:26:34.355 --> 00:26:39.287
And so that is definitely that piece.
The other component is that...

8d6b3f72-c2f0-4422-8631-62277aa4ab86/159-0
00:26:40.967 --> 00:26:45.074
We are not able to just add stuff like
our referrals are going up,

8d6b3f72-c2f0-4422-8631-62277aa4ab86/159-1
00:26:45.074 --> 00:26:48.569
which is great,
but we are not really able to just throw

8d6b3f72-c2f0-4422-8631-62277aa4ab86/159-2
00:26:48.569 --> 00:26:51.880
more and more and more staff at them
based, you know,

8d6b3f72-c2f0-4422-8631-62277aa4ab86/159-3
00:26:51.880 --> 00:26:55.313
exactly on what you just said with
margins being tight.

8d6b3f72-c2f0-4422-8631-62277aa4ab86/159-4
00:26:55.313 --> 00:27:00.034
And so it's how do we how do we look at
other alternatives and technology is

8d6b3f72-c2f0-4422-8631-62277aa4ab86/159-5
00:27:00.034 --> 00:27:00.647
there and.

8d6b3f72-c2f0-4422-8631-62277aa4ab86/160-0
00:27:00.967 --> 00:27:06.142
Why wouldn't we use that to help
supplement, not necessarily replace?

8d6b3f72-c2f0-4422-8631-62277aa4ab86/160-1
00:27:06.142 --> 00:27:11.909
I think we have so much more that our
team can do if they are not bogged down

8d6b3f72-c2f0-4422-8631-62277aa4ab86/160-2
00:27:11.909 --> 00:27:17.232
into digging through 800 pages.
Now they can spend more time talking to

8d6b3f72-c2f0-4422-8631-62277aa4ab86/160-3
00:27:17.232 --> 00:27:22.407
the patients, talking to the families,
being in front of the patient,

8d6b3f72-c2f0-4422-8631-62277aa4ab86/161-0
00:27:23.687 --> 00:27:28.563
doing an actual visual assessment of the
patient sometimes when that has been

8d6b3f72-c2f0-4422-8631-62277aa4ab86/161-1
00:27:28.563 --> 00:27:31.501
harder.
So I think taking some of this tedious

8d6b3f72-c2f0-4422-8631-62277aa4ab86/161-2
00:27:31.501 --> 00:27:35.940
work off of them will, and again,
we're at the very beginning of this,

8d6b3f72-c2f0-4422-8631-62277aa4ab86/161-3
00:27:35.940 --> 00:27:39.754
like our full team really just got access
to this last week.

8d6b3f72-c2f0-4422-8631-62277aa4ab86/161-4
00:27:39.754 --> 00:27:42.567
We went from just partial team to full
team.

8d6b3f72-c2f0-4422-8631-62277aa4ab86/162-0
00:27:43.047 --> 00:27:46.683
having access.
And so I think the full impact is yet to

8d6b3f72-c2f0-4422-8631-62277aa4ab86/162-1
00:27:46.683 --> 00:27:51.943
be seen as they really learn it and as we
continue to modify it and optimize it.

8d6b3f72-c2f0-4422-8631-62277aa4ab86/162-2
00:27:51.943 --> 00:27:57.397
But that is my vision and my hope is that
taking this tedious work off of them will

8d6b3f72-c2f0-4422-8631-62277aa4ab86/162-3
00:27:57.397 --> 00:28:02.462
actually help them work at the top of
their license more and be able to be in

8d6b3f72-c2f0-4422-8631-62277aa4ab86/162-4
00:28:02.462 --> 00:28:07.007
front of the patient and having more
conversations with the families.

8d6b3f72-c2f0-4422-8631-62277aa4ab86/163-0
00:28:08.767 --> 00:28:11.267
Yeah, I will add to your second question.
I mean,

8d6b3f72-c2f0-4422-8631-62277aa4ab86/163-1
00:28:11.267 --> 00:28:15.167
can we commercialize this and should we,
and I think there is a model for all

8d6b3f72-c2f0-4422-8631-62277aa4ab86/163-2
00:28:15.167 --> 00:28:18.917
healthcare systems to sort of collaborate
in tools such as this and have a

8d6b3f72-c2f0-4422-8631-62277aa4ab86/163-3
00:28:18.917 --> 00:28:22.567
licensing model where, you know,
you are compensated for all of the work

8d6b3f72-c2f0-4422-8631-62277aa4ab86/163-4
00:28:22.567 --> 00:28:25.717
that you've kind of put in,
but also they are able to reap the

8d6b3f72-c2f0-4422-8631-62277aa4ab86/163-5
00:28:25.717 --> 00:28:28.167
benefits as opposed to us starting from
scratch.

8d6b3f72-c2f0-4422-8631-62277aa4ab86/164-0
00:28:28.727 --> 00:28:32.353
And in general, I mean,
I'd like to sort of appeal to the

8d6b3f72-c2f0-4422-8631-62277aa4ab86/164-1
00:28:32.353 --> 00:28:37.104
audience, whoever are listening, I mean,
this is a day and time where AI is

8d6b3f72-c2f0-4422-8631-62277aa4ab86/164-2
00:28:37.104 --> 00:28:41.917
accelerating so, so fast that, you know,
I think they would love to, I mean,

8d6b3f72-c2f0-4422-8631-62277aa4ab86/164-3
00:28:41.917 --> 00:28:47.106
they would want to invest in building out
tools for their organization because not

8d6b3f72-c2f0-4422-8631-62277aa4ab86/164-4
00:28:47.106 --> 00:28:49.607
everything is solved by the vendors. And

8d6b3f72-c2f0-4422-8631-62277aa4ab86/165-0
00:28:49.767 --> 00:28:54.477
Some of these are like very custom to
your organization. Things are,

8d6b3f72-c2f0-4422-8631-62277aa4ab86/165-1
00:28:54.477 --> 00:28:59.461
tools are evolving so that building some
of these things may be, I mean,

8d6b3f72-c2f0-4422-8631-62277aa4ab86/165-2
00:28:59.461 --> 00:29:04.991
a commodity in two years or so, right?
So I think just an appeal to everybody to

8d6b3f72-c2f0-4422-8631-62277aa4ab86/165-3
00:29:04.991 --> 00:29:10.316
move fast and just start adding value
because a lot of these technologies is,

8d6b3f72-c2f0-4422-8631-62277aa4ab86/165-4
00:29:10.316 --> 00:29:14.207
and is going to become a commodity in two
to five years.

8d6b3f72-c2f0-4422-8631-62277aa4ab86/166-0
00:29:14.207 --> 00:29:14.727
I know so.

8d6b3f72-c2f0-4422-8631-62277aa4ab86/167-0
00:29:15.767 --> 00:29:20.678
I'd like to thank you both for walking us
through this very interesting Gen AI use

8d6b3f72-c2f0-4422-8631-62277aa4ab86/167-1
00:29:20.678 --> 00:29:24.997
case and how it's expecting patient
safety, patient experience, revenue,

8d6b3f72-c2f0-4422-8631-62277aa4ab86/167-2
00:29:24.997 --> 00:29:29.967
and allowing clinicians to operate at the
type of their license. For our listeners,

8d6b3f72-c2f0-4422-8631-62277aa4ab86/167-3
00:29:29.967 --> 00:29:34.287
again, this has been Corey Pace,
the AVP of BD and Patient Solutions for

8d6b3f72-c2f0-4422-8631-62277aa4ab86/168-0
00:29:34.367 --> 00:29:37.035
at the Institute for Rehabilitation and
Research. Corey,

8d6b3f72-c2f0-4422-8631-62277aa4ab86/168-1
00:29:37.035 --> 00:29:38.487
thank you for joining us today.

8d6b3f72-c2f0-4422-8631-62277aa4ab86/169-0
00:29:38.807 --> 00:29:39.687
Thank you.

8d6b3f72-c2f0-4422-8631-62277aa4ab86/170-0
00:29:39.927 --> 00:29:43.192
And Chaitanya Vampati,
the AVP of AI and Analytics for the

8d6b3f72-c2f0-4422-8631-62277aa4ab86/170-1
00:29:43.192 --> 00:29:47.287
Memorial Health Health System. Chaitanya,
thank you for joining us today.

8d6b3f72-c2f0-4422-8631-62277aa4ab86/171-0
00:29:47.607 --> 00:29:48.487
Thanks, a letter.