1
00:00:00,719 --> 00:00:04,000
Exciting things are happening at Becker's Healthcare. Stay

2
00:00:04,000 --> 00:00:06,240
ahead of industry trends with the new Becker's

3
00:00:06,240 --> 00:00:08,500
CFO plus Revenue Cycle podcast,

4
00:00:08,800 --> 00:00:11,119
your go to source for insights from top

5
00:00:11,119 --> 00:00:14,224
healthcare finance leaders. Tune in wherever you get

6
00:00:14,224 --> 00:00:14,964
your podcasts.

7
00:00:15,904 --> 00:00:18,464
And don't miss the tenth annual Health IT

8
00:00:18,464 --> 00:00:22,155
plus Digital Health plus RCM conference, happening September

9
00:00:22,155 --> 00:00:24,464
30 to 10/03/2025

10
00:00:24,464 --> 00:00:25,204
in Chicago.

11
00:00:25,664 --> 00:00:27,125
Join thousands of executives,

12
00:00:27,469 --> 00:00:30,429
engage with industry leaders, and explore the future

13
00:00:30,429 --> 00:00:31,570
of health care innovation.

14
00:00:32,270 --> 00:00:36,750
Learn more about our upcoming events at beckershospitalreview.com.

15
00:00:36,750 --> 00:00:37,570
See you there.

16
00:00:38,590 --> 00:00:40,829
This is Laura Dierda with the Becker's Healthcare

17
00:00:40,829 --> 00:00:42,875
Podcast. I'm thrilled today to be joined by

18
00:00:42,875 --> 00:00:46,314
Michael Hasselberg, chief digital health officer at UR

19
00:00:46,314 --> 00:00:49,114
Medicine and professor of psychiatry, clinical nursing, and

20
00:00:49,114 --> 00:00:51,434
data science at the University of Rochester. Michael,

21
00:00:51,434 --> 00:00:52,475
it's a pleasure to have you on the

22
00:00:52,475 --> 00:00:53,375
podcast today.

23
00:00:54,000 --> 00:00:55,539
Thank you for having me, Laura.

24
00:00:55,920 --> 00:00:57,379
Looking forward to our discussion.

25
00:00:58,000 --> 00:00:59,759
Absolutely. I am too. I I know it's,

26
00:00:59,759 --> 00:01:01,519
you know, got a lot going on there

27
00:01:01,519 --> 00:01:04,159
at UR Medicine, and and certainly something that,

28
00:01:04,399 --> 00:01:06,375
we'd love to hear more about how you're

29
00:01:06,454 --> 00:01:08,055
thinking about the future as well. But before

30
00:01:08,055 --> 00:01:09,575
we dive in, can you take a minute

31
00:01:09,575 --> 00:01:10,935
and tell us a little bit about UR

32
00:01:10,935 --> 00:01:12,474
Medicine and what makes it unique?

33
00:01:13,174 --> 00:01:15,495
Yeah. I'd love to. So those of you

34
00:01:15,495 --> 00:01:17,894
who are not familiar with UR Medicine, we

35
00:01:17,894 --> 00:01:18,394
are

36
00:01:19,640 --> 00:01:23,739
a large academic health system in Western

37
00:01:24,359 --> 00:01:26,780
Part of New York State. We are

38
00:01:27,640 --> 00:01:30,359
the largest health system outside of New York

39
00:01:30,359 --> 00:01:32,119
City in the in the state of New

40
00:01:32,119 --> 00:01:34,219
York. So we have a very large

41
00:01:35,284 --> 00:01:38,004
geographic reach. Essentially the whole Western half of

42
00:01:38,004 --> 00:01:39,064
the state is,

43
00:01:40,724 --> 00:01:41,625
our market

44
00:01:42,004 --> 00:01:44,265
that we serve. So very diverse

45
00:01:45,364 --> 00:01:49,049
patient population. Everything from the patients you would

46
00:01:49,130 --> 00:01:51,210
see in in an inner city of a

47
00:01:51,210 --> 00:01:52,670
mid sized city to

48
00:01:53,049 --> 00:01:55,370
you go 30 miles outside of the city

49
00:01:55,370 --> 00:01:57,549
limits and it's fairly rural.

50
00:01:58,730 --> 00:02:00,189
What's really special

51
00:02:00,490 --> 00:02:00,990
about,

52
00:02:01,689 --> 00:02:03,390
UR Medicine is

53
00:02:04,034 --> 00:02:05,974
our relationship with our university

54
00:02:06,354 --> 00:02:07,094
which is

55
00:02:08,114 --> 00:02:09,574
very very unique

56
00:02:10,034 --> 00:02:10,534
for

57
00:02:11,074 --> 00:02:13,094
academic health systems today.

58
00:02:13,474 --> 00:02:14,694
We are one of the

59
00:02:15,235 --> 00:02:18,139
few academic health systems in the country that

60
00:02:18,139 --> 00:02:21,419
is fully integrated into our parent university. Our

61
00:02:21,419 --> 00:02:24,219
parent university essentially owns the health system and

62
00:02:24,219 --> 00:02:25,439
our CEO

63
00:02:26,620 --> 00:02:28,939
and president of the health system reports right

64
00:02:28,939 --> 00:02:31,039
up to the president of the university.

65
00:02:31,580 --> 00:02:32,080
Our

66
00:02:33,155 --> 00:02:36,455
quaternary hospital sits right on the university's campus.

67
00:02:37,555 --> 00:02:40,435
And you know what makes that special for

68
00:02:40,435 --> 00:02:42,675
us and we actually look at as a

69
00:02:42,675 --> 00:02:43,175
differentiator

70
00:02:43,555 --> 00:02:45,840
is we have access to some of the

71
00:02:46,000 --> 00:02:49,539
most brilliant minds on our academic campus across

72
00:02:50,479 --> 00:02:50,979
engineering,

73
00:02:51,759 --> 00:02:52,259
across

74
00:02:52,800 --> 00:02:55,060
computer science and data science,

75
00:02:56,000 --> 00:02:58,580
even a world renowned school of music,

76
00:02:59,655 --> 00:03:02,395
and business school and education school.

77
00:03:03,175 --> 00:03:06,055
And we get to leverage that expertise and

78
00:03:06,055 --> 00:03:08,215
apply it to healthcare. And so we have

79
00:03:08,215 --> 00:03:08,875
a true

80
00:03:09,495 --> 00:03:12,855
innovation team at the University of Rochester supported

81
00:03:12,855 --> 00:03:13,835
by the university

82
00:03:15,210 --> 00:03:18,330
to help move the strategic priorities of the

83
00:03:18,330 --> 00:03:20,590
health system forward. Not a traditional

84
00:03:21,370 --> 00:03:24,030
research shop where you know our faculty

85
00:03:24,730 --> 00:03:25,389
the expectations

86
00:03:25,770 --> 00:03:28,030
aren't to get as much NIH funding

87
00:03:28,490 --> 00:03:31,705
as possible and publish papers. It's truly to

88
00:03:31,705 --> 00:03:32,205
innovate

89
00:03:33,064 --> 00:03:35,544
to support health system priorities. So under one

90
00:03:35,544 --> 00:03:36,764
roof we have faculty

91
00:03:37,064 --> 00:03:38,525
from all of our schools.

92
00:03:39,705 --> 00:03:42,264
Again engineering and computer science and data science

93
00:03:42,264 --> 00:03:44,104
and even the the School of Music under

94
00:03:44,104 --> 00:03:46,159
the same roof as faculty of the medical

95
00:03:46,159 --> 00:03:49,760
school, dental school, and nursing school. So having

96
00:03:49,760 --> 00:03:52,080
that really eclectic group of,

97
00:03:52,560 --> 00:03:54,819
minds allows us to come up with some

98
00:03:55,040 --> 00:03:57,760
really exciting solutions to some of our pressing

99
00:03:57,760 --> 00:03:59,540
problems in the health system side.

100
00:04:01,925 --> 00:04:03,525
That's great to hear. And what a unique

101
00:04:03,525 --> 00:04:05,764
opportunity to truly, truly be able to focus

102
00:04:05,764 --> 00:04:08,004
on the patient care side and and being

103
00:04:08,004 --> 00:04:10,104
innovative within health care delivery.

104
00:04:10,405 --> 00:04:12,164
I'm curious. Looking in the last year or

105
00:04:12,164 --> 00:04:14,584
so, what accomplishments are you most proud of?

106
00:04:14,724 --> 00:04:17,430
Yeah. So I I the the accomplishment that

107
00:04:17,649 --> 00:04:20,149
I'm most proud of, for our university is

108
00:04:20,209 --> 00:04:21,430
maybe not the sexiest,

109
00:04:22,209 --> 00:04:22,709
accomplishment,

110
00:04:23,250 --> 00:04:25,110
especially, you know, when you think of,

111
00:04:25,649 --> 00:04:26,949
all the great opportunities

112
00:04:27,250 --> 00:04:29,589
and innovation in health care right now.

113
00:04:30,055 --> 00:04:31,834
You know, our biggest accomplishment

114
00:04:32,694 --> 00:04:33,995
was the setting up

115
00:04:34,375 --> 00:04:36,795
of an artificial intelligence governance

116
00:04:37,334 --> 00:04:38,954
across the health system

117
00:04:39,735 --> 00:04:40,475
and actually

118
00:04:41,095 --> 00:04:44,579
university wide. You know, we we felt that

119
00:04:44,899 --> 00:04:46,919
having a governance that,

120
00:04:47,699 --> 00:04:50,099
could help us think through, you know, how

121
00:04:50,099 --> 00:04:51,159
do you deliver,

122
00:04:51,939 --> 00:04:55,139
this really exciting technology in in a safe

123
00:04:55,139 --> 00:04:57,800
and responsible way across all of our missions

124
00:04:58,355 --> 00:04:59,254
at the university,

125
00:04:59,714 --> 00:05:01,235
but also set it up in a way

126
00:05:01,235 --> 00:05:03,814
that to allow us to be nimble and

127
00:05:04,035 --> 00:05:04,535
to,

128
00:05:05,475 --> 00:05:05,975
accelerate,

129
00:05:06,514 --> 00:05:08,455
the advances using this technology

130
00:05:09,074 --> 00:05:11,235
was not a small task to do.

131
00:05:11,634 --> 00:05:14,939
And it took us, you know, easily six

132
00:05:15,160 --> 00:05:17,399
six months to a year to really get

133
00:05:17,399 --> 00:05:18,699
the governance structure

134
00:05:19,080 --> 00:05:19,580
right.

135
00:05:20,439 --> 00:05:22,860
But we were able to accomplish that. And,

136
00:05:23,080 --> 00:05:25,180
you know, that that I think has really

137
00:05:25,414 --> 00:05:28,055
set the foundation for us to continue to

138
00:05:28,055 --> 00:05:28,875
be leaders

139
00:05:29,254 --> 00:05:31,675
around the use of technology and specifically

140
00:05:32,375 --> 00:05:33,435
artificial intelligence,

141
00:05:34,055 --> 00:05:36,555
not only in health care, but in education

142
00:05:36,854 --> 00:05:39,675
and and and our research missions also.

143
00:05:42,370 --> 00:05:44,290
That's great to hear. And, you know, I

144
00:05:44,290 --> 00:05:46,769
I know looking at the governance process is

145
00:05:46,769 --> 00:05:48,689
not an easy one or a simple one

146
00:05:48,689 --> 00:05:50,769
at all. So many health systems, right now

147
00:05:50,769 --> 00:05:52,129
are still trying to figure it out and,

148
00:05:53,404 --> 00:05:55,345
create the the right policies

149
00:05:55,645 --> 00:05:57,165
and bringing the right people in order to

150
00:05:57,165 --> 00:05:59,564
make those decisions. Could you go a little

151
00:05:59,564 --> 00:06:01,665
bit deeper in there in terms of,

152
00:06:02,365 --> 00:06:04,285
how do you did you develop the governance

153
00:06:04,285 --> 00:06:06,145
process, especially around AI?

154
00:06:06,925 --> 00:06:09,410
Who should be involved, and and how do

155
00:06:09,430 --> 00:06:12,029
you, really then cascade the message once you've

156
00:06:12,029 --> 00:06:12,529
got,

157
00:06:13,149 --> 00:06:14,449
policies ready to go?

158
00:06:15,629 --> 00:06:16,589
Yeah. So,

159
00:06:17,069 --> 00:06:17,889
our governance

160
00:06:18,350 --> 00:06:20,029
starts right up at the top of the

161
00:06:20,029 --> 00:06:22,670
university. And so we have what we call

162
00:06:22,670 --> 00:06:24,125
an an artificial intelligence

163
00:06:24,664 --> 00:06:25,164
council

164
00:06:25,464 --> 00:06:27,785
that reports right up to the president of

165
00:06:27,785 --> 00:06:30,685
the university and the in the president's cabinets.

166
00:06:30,745 --> 00:06:31,245
And

167
00:06:31,704 --> 00:06:33,884
the artificial intelligence council

168
00:06:34,504 --> 00:06:36,125
has lots of different

169
00:06:36,584 --> 00:06:38,284
stakeholders on it including

170
00:06:39,759 --> 00:06:40,980
chairs and representatives

171
00:06:41,439 --> 00:06:41,939
from

172
00:06:42,399 --> 00:06:45,220
the AI committees, which I'll talk about next.

173
00:06:45,680 --> 00:06:47,459
But it also has stakeholders

174
00:06:47,759 --> 00:06:49,860
from privacy and legal,

175
00:06:50,720 --> 00:06:51,459
and compliance

176
00:06:51,839 --> 00:06:52,339
from

177
00:06:52,720 --> 00:06:53,220
IT

178
00:06:53,600 --> 00:06:54,420
to ethics,

179
00:06:55,694 --> 00:06:56,194
to,

180
00:06:56,654 --> 00:06:59,134
some of our our our content experts, our

181
00:06:59,134 --> 00:07:00,594
deep technical experts,

182
00:07:01,214 --> 00:07:02,115
a and AI.

183
00:07:02,735 --> 00:07:05,954
But we also have experts around thinking about

184
00:07:06,094 --> 00:07:08,414
AI use in our community and and how

185
00:07:08,414 --> 00:07:10,629
we can make it equitable

186
00:07:11,410 --> 00:07:15,169
across our our our community. Below the AI

187
00:07:15,169 --> 00:07:15,669
Council,

188
00:07:16,209 --> 00:07:18,930
we have what we call domain committees and

189
00:07:18,930 --> 00:07:20,870
we have five domain committees.

190
00:07:21,785 --> 00:07:23,884
We have a clinical domain committee,

191
00:07:24,665 --> 00:07:28,345
a research domain committee and education domain committee

192
00:07:28,345 --> 00:07:29,245
and administrative

193
00:07:29,625 --> 00:07:32,045
domain committee and a marketing and communications

194
00:07:32,985 --> 00:07:35,485
domain committee. And those committees have

195
00:07:37,209 --> 00:07:39,289
a diverse group of stakeholders that sit on

196
00:07:39,289 --> 00:07:41,689
those committees and the purpose of the AI

197
00:07:41,689 --> 00:07:43,629
Council is to really enable

198
00:07:44,409 --> 00:07:45,229
the committees

199
00:07:45,769 --> 00:07:47,629
at the domain level to

200
00:07:48,889 --> 00:07:49,629
be autonomous

201
00:07:50,735 --> 00:07:53,615
to to be able to identify what are

202
00:07:53,615 --> 00:07:55,794
the specific artificial intelligence

203
00:07:56,254 --> 00:07:58,654
drivers that align with the values of that

204
00:07:58,654 --> 00:07:59,154
domain,

205
00:08:00,975 --> 00:08:04,410
support those committees to develop their own guidelines,

206
00:08:04,629 --> 00:08:07,110
procedures, and in some cases, like in our

207
00:08:07,110 --> 00:08:09,610
clinical domain committee, actual policies,

208
00:08:10,949 --> 00:08:12,250
to support

209
00:08:12,550 --> 00:08:13,050
regulatory

210
00:08:13,589 --> 00:08:15,689
and compliance and safety

211
00:08:16,069 --> 00:08:16,569
concerns.

212
00:08:17,735 --> 00:08:21,414
But also, you know, allow those domain committees

213
00:08:21,414 --> 00:08:24,055
to inform the council about where we should

214
00:08:24,055 --> 00:08:25,914
be making strategic investments,

215
00:08:26,375 --> 00:08:27,834
at the university level,

216
00:08:28,375 --> 00:08:31,095
within and within AI. And so at the

217
00:08:31,095 --> 00:08:31,995
council level,

218
00:08:32,409 --> 00:08:34,509
you know, it's it's it's that

219
00:08:34,970 --> 00:08:37,610
group's job is to try to standardize where

220
00:08:37,610 --> 00:08:40,590
we can across all of the domains on

221
00:08:40,809 --> 00:08:41,950
our guiding principles,

222
00:08:42,809 --> 00:08:45,629
and again guidance and and policies

223
00:08:46,330 --> 00:08:48,029
to make sure that we have

224
00:08:48,464 --> 00:08:48,964
safe,

225
00:08:49,985 --> 00:08:50,804
and internally

226
00:08:51,264 --> 00:08:51,764
improved,

227
00:08:52,865 --> 00:08:55,745
AI tools that are equitable across our students,

228
00:08:55,745 --> 00:08:57,605
staff, and and and faculty.

229
00:08:58,384 --> 00:09:00,084
And, really serve

230
00:09:00,704 --> 00:09:01,524
as a group,

231
00:09:01,904 --> 00:09:05,009
sounding board for the the AI committees if

232
00:09:05,009 --> 00:09:05,990
they run into,

233
00:09:06,850 --> 00:09:09,830
issues that they need further strategic

234
00:09:10,690 --> 00:09:13,730
guidance on. And then within that governance structure,

235
00:09:13,730 --> 00:09:14,710
we have workflows,

236
00:09:15,970 --> 00:09:16,470
identified

237
00:09:16,929 --> 00:09:17,429
for

238
00:09:17,754 --> 00:09:20,314
the submission of new AI tools that,

239
00:09:20,794 --> 00:09:24,074
individuals within our university and health system would

240
00:09:24,074 --> 00:09:24,814
like to

241
00:09:25,674 --> 00:09:26,735
try or use.

242
00:09:27,115 --> 00:09:28,875
And so we've we've built a kind of

243
00:09:28,875 --> 00:09:31,694
a nimble workflow to get those tools vetted.

244
00:09:32,419 --> 00:09:34,519
And then we've also, you know, developed,

245
00:09:35,980 --> 00:09:39,220
a new contract addendum guidance for some of

246
00:09:39,220 --> 00:09:42,419
our long standing vendors who are starting to

247
00:09:42,419 --> 00:09:42,919
introduce

248
00:09:43,860 --> 00:09:47,004
AI functionality into their platforms to make sure

249
00:09:47,004 --> 00:09:49,504
that we have that partnership with our vendors

250
00:09:49,565 --> 00:09:50,384
and they are,

251
00:09:51,085 --> 00:09:53,105
you know, protecting our data,

252
00:09:53,565 --> 00:09:55,184
to the standards that we,

253
00:09:55,725 --> 00:09:58,524
hold our ourselves to at the university and

254
00:09:58,524 --> 00:10:00,924
that they are also thinking about what are

255
00:10:00,924 --> 00:10:01,664
the ethical

256
00:10:03,459 --> 00:10:03,959
bias,

257
00:10:04,819 --> 00:10:06,039
accuracy, reliability

258
00:10:06,579 --> 00:10:07,079
implications

259
00:10:07,699 --> 00:10:10,259
of these tools, especially on the generative AI

260
00:10:10,259 --> 00:10:10,759
side.

261
00:10:13,139 --> 00:10:15,620
That's helpful for understanding. Thank you, Michael, for

262
00:10:15,620 --> 00:10:17,785
going through that with us and especially, like

263
00:10:17,785 --> 00:10:19,625
you mentioned, getting into generative AI. I know

264
00:10:19,625 --> 00:10:22,105
that's a whole new, can of worms opening

265
00:10:22,105 --> 00:10:23,245
in some ways. So,

266
00:10:23,945 --> 00:10:27,225
very much appreciated. Now I'm curious looking ahead,

267
00:10:27,225 --> 00:10:28,345
where do you see some of the big

268
00:10:28,345 --> 00:10:30,504
growth opportunities over the next twelve months or

269
00:10:30,504 --> 00:10:31,004
so?

270
00:10:31,679 --> 00:10:34,740
Yeah. Well, let's, Laura stay on generative

271
00:10:35,120 --> 00:10:36,100
AI. You know,

272
00:10:36,399 --> 00:10:38,879
we we as a as a health system,

273
00:10:38,879 --> 00:10:40,500
you know, we were one of the

274
00:10:40,879 --> 00:10:41,379
early,

275
00:10:42,000 --> 00:10:42,500
users

276
00:10:42,879 --> 00:10:43,360
of,

277
00:10:44,054 --> 00:10:47,914
secured private instance of, GPT four, that specific

278
00:10:48,054 --> 00:10:49,195
foundation model.

279
00:10:49,815 --> 00:10:53,034
You know, we we also have other secure

280
00:10:53,095 --> 00:10:53,995
private instances

281
00:10:54,455 --> 00:10:57,174
of, public foundation models, and and we also

282
00:10:57,174 --> 00:10:59,759
have a supercomputer where we have

283
00:11:00,860 --> 00:11:01,519
a smaller,

284
00:11:02,299 --> 00:11:03,199
open source,

285
00:11:04,059 --> 00:11:05,919
model sitting on. And

286
00:11:06,299 --> 00:11:07,199
for a university,

287
00:11:08,299 --> 00:11:10,399
academic health system like ours,

288
00:11:11,019 --> 00:11:14,220
it has never been easier for us to

289
00:11:14,220 --> 00:11:16,284
develop our our own AI tools

290
00:11:16,904 --> 00:11:17,404
rapidly

291
00:11:17,784 --> 00:11:20,764
to solve our own problems by essentially,

292
00:11:21,625 --> 00:11:23,004
doing some prompt engineering,

293
00:11:24,264 --> 00:11:26,585
of these foundation models with our own data

294
00:11:26,585 --> 00:11:29,384
or fine tuning these foundation models on our

295
00:11:29,384 --> 00:11:30,044
own data.

296
00:11:30,389 --> 00:11:32,090
And so, like, we're really

297
00:11:33,350 --> 00:11:36,410
very narrowly focused in on where are those,

298
00:11:37,750 --> 00:11:38,250
manual

299
00:11:38,950 --> 00:11:39,450
administrative

300
00:11:40,790 --> 00:11:42,889
tasks across the system

301
00:11:43,715 --> 00:11:46,695
that if we were able to automate those,

302
00:11:47,315 --> 00:11:47,815
using,

303
00:11:49,634 --> 00:11:51,254
generative AI tools,

304
00:11:51,795 --> 00:11:53,955
where what what would have the biggest, not

305
00:11:53,955 --> 00:11:55,415
only return on investments,

306
00:11:55,795 --> 00:11:56,190
but,

307
00:11:57,629 --> 00:11:58,289
you know,

308
00:11:58,669 --> 00:11:59,169
decreasing,

309
00:12:00,269 --> 00:12:03,570
mitigation risks and and have the biggest impact

310
00:12:04,190 --> 00:12:05,570
really on our employees,

311
00:12:06,029 --> 00:12:07,250
our our clinicians,

312
00:12:08,029 --> 00:12:09,809
and our staff? That's where we're

313
00:12:10,304 --> 00:12:10,804
focusing

314
00:12:11,184 --> 00:12:13,264
our efforts on and we're making a lot

315
00:12:13,264 --> 00:12:16,965
of headway. We are cranking out tools in,

316
00:12:17,184 --> 00:12:20,304
you know, days to weeks. Tools that typically

317
00:12:20,304 --> 00:12:21,924
would have taken our engineering

318
00:12:22,304 --> 00:12:23,764
and data science teams

319
00:12:24,220 --> 00:12:26,539
six months to a year to build. So

320
00:12:26,539 --> 00:12:30,080
it's it's been really exciting and quite powerful,

321
00:12:31,820 --> 00:12:35,759
you know, our ability to to develop internally

322
00:12:36,299 --> 00:12:36,960
by having

323
00:12:37,455 --> 00:12:40,434
access to these these massive foundation models.

324
00:12:43,295 --> 00:12:45,295
That's great to hear. And, you know, really

325
00:12:45,295 --> 00:12:47,615
cool to have that opportunity to speed things

326
00:12:47,615 --> 00:12:49,634
up and, like you mentioned, just,

327
00:12:50,654 --> 00:12:53,070
do more and more as time goes on,

328
00:12:53,309 --> 00:12:55,309
with degenerative AI. I know a lot of

329
00:12:55,309 --> 00:12:57,809
opportunity there, but I can imagine also challenges.

330
00:12:58,029 --> 00:12:59,789
What are some of the big, challenges that

331
00:12:59,789 --> 00:13:00,529
you're anticipating

332
00:13:01,070 --> 00:13:02,769
heading into the future as well?

333
00:13:03,389 --> 00:13:04,929
Yeah. You know, I think,

334
00:13:05,309 --> 00:13:06,769
one of the biggest challenges

335
00:13:07,149 --> 00:13:07,889
is understanding

336
00:13:08,945 --> 00:13:10,245
what the regulatory

337
00:13:11,264 --> 00:13:15,445
landscape may be around the use of of

338
00:13:15,745 --> 00:13:18,544
of specifically generative AI on on the health

339
00:13:18,544 --> 00:13:20,625
care side. And, you know, we know that

340
00:13:20,625 --> 00:13:23,345
there is a lot of changes happening at

341
00:13:23,345 --> 00:13:24,404
the federal level,

342
00:13:25,100 --> 00:13:27,339
in this space and trying to understand that.

343
00:13:27,339 --> 00:13:29,519
But when push comes to shove,

344
00:13:29,899 --> 00:13:32,059
you know, as a as a health care

345
00:13:32,059 --> 00:13:32,559
institution,

346
00:13:33,019 --> 00:13:33,740
you know, we

347
00:13:34,459 --> 00:13:37,980
our primary business is providing high quality care

348
00:13:37,980 --> 00:13:38,879
to our patients.

349
00:13:39,434 --> 00:13:42,394
And so, like, that is that's the the

350
00:13:42,394 --> 00:13:44,235
essence of what we do. And so it

351
00:13:44,235 --> 00:13:46,095
doesn't really matter, you know,

352
00:13:46,634 --> 00:13:47,134
what

353
00:13:47,595 --> 00:13:50,735
regulations come out or maybe lack of regulations

354
00:13:50,955 --> 00:13:53,195
that may come out. We still have to

355
00:13:53,195 --> 00:13:54,830
hold ourselves accountable

356
00:13:55,529 --> 00:13:56,029
of

357
00:13:56,410 --> 00:13:59,149
of making sure that these technologies

358
00:14:00,250 --> 00:14:00,990
are safe

359
00:14:01,370 --> 00:14:02,190
and trustworthy,

360
00:14:03,370 --> 00:14:03,870
when,

361
00:14:04,330 --> 00:14:07,210
delivered within clinical care settings. And we need

362
00:14:07,210 --> 00:14:08,190
to make sure

363
00:14:08,495 --> 00:14:10,514
that we can create assurances

364
00:14:10,975 --> 00:14:13,855
for our clinicians using these tools that they

365
00:14:13,855 --> 00:14:14,654
can trust,

366
00:14:15,774 --> 00:14:17,154
the the the output,

367
00:14:17,615 --> 00:14:18,914
that that we're seeing

368
00:14:19,375 --> 00:14:22,414
from from the AI. And so, you know,

369
00:14:22,414 --> 00:14:23,074
I think,

370
00:14:24,029 --> 00:14:26,290
you know, really trying to

371
00:14:26,670 --> 00:14:29,389
get a handle on that and and really

372
00:14:29,389 --> 00:14:30,290
trying to

373
00:14:30,910 --> 00:14:32,769
figure out how do we create,

374
00:14:34,269 --> 00:14:36,450
pre deployment validation opportunities

375
00:14:36,990 --> 00:14:39,250
of of these generative tools and

376
00:14:39,565 --> 00:14:40,384
post deployment

377
00:14:40,845 --> 00:14:42,144
auditing and monitoring,

378
00:14:43,004 --> 00:14:43,504
systems,

379
00:14:44,125 --> 00:14:46,384
to watch what these tools are doing.

380
00:14:46,764 --> 00:14:48,925
You know, that that is a a space

381
00:14:48,925 --> 00:14:51,345
that I I see a lot of opportunity,

382
00:14:51,485 --> 00:14:54,149
but again, a lot of challenge just given

383
00:14:54,149 --> 00:14:55,129
there's so much

384
00:14:56,710 --> 00:14:58,170
black box or

385
00:14:58,550 --> 00:15:01,990
things we don't know about the these these

386
00:15:01,990 --> 00:15:04,550
large foundation models. And so, you know, that

387
00:15:04,550 --> 00:15:05,450
has been the,

388
00:15:05,910 --> 00:15:06,570
I think,

389
00:15:07,105 --> 00:15:08,804
biggest thing that we've been

390
00:15:09,184 --> 00:15:11,825
running into and thinking about. Again, never been

391
00:15:11,825 --> 00:15:14,384
easier to build a point solution. Like, we

392
00:15:14,544 --> 00:15:16,544
we we can we can do that really,

393
00:15:16,544 --> 00:15:17,365
really quickly.

394
00:15:17,745 --> 00:15:20,404
But to be able to go from pilot

395
00:15:20,465 --> 00:15:20,965
to

396
00:15:21,620 --> 00:15:24,019
full scale across the health system in a

397
00:15:24,019 --> 00:15:25,720
safe and trustworthy way,

398
00:15:26,899 --> 00:15:29,700
that has been a challenge and difficult. And,

399
00:15:29,700 --> 00:15:30,519
again, it's

400
00:15:30,820 --> 00:15:31,879
because of

401
00:15:32,339 --> 00:15:33,639
not having that

402
00:15:34,985 --> 00:15:36,205
clarity or mechanisms,

403
00:15:37,144 --> 00:15:40,764
in place to to, again, monitor and audit

404
00:15:40,904 --> 00:15:42,284
these tools at scale.

405
00:15:44,745 --> 00:15:47,144
That's fascinating to hear. You know? And, certainly,

406
00:15:47,144 --> 00:15:47,965
I can imagine,

407
00:15:48,664 --> 00:15:50,639
a a challenge that others others in your

408
00:15:50,639 --> 00:15:52,799
position are are seeing in health systems as

409
00:15:52,799 --> 00:15:55,440
well. And I'm curious, you know, especially when

410
00:15:55,440 --> 00:15:58,600
you think about, the taking pilots or some

411
00:15:58,600 --> 00:16:00,240
of the small scale things that you're doing

412
00:16:00,240 --> 00:16:03,039
and then seeing successes, rolling them out, and

413
00:16:03,039 --> 00:16:03,519
and,

414
00:16:04,079 --> 00:16:05,299
defining challenges

415
00:16:05,764 --> 00:16:08,004
within that. I guess, could you share with

416
00:16:08,004 --> 00:16:10,424
us any examples of needing to

417
00:16:10,804 --> 00:16:12,565
kind of put the brakes on a project

418
00:16:12,565 --> 00:16:13,065
or,

419
00:16:14,004 --> 00:16:15,924
something that was like a a a near

420
00:16:15,924 --> 00:16:18,565
miss that really kinda highlighted what you're talking

421
00:16:18,565 --> 00:16:19,304
about here?

422
00:16:20,990 --> 00:16:23,090
You know, not not per se,

423
00:16:24,269 --> 00:16:25,409
a near miss,

424
00:16:25,950 --> 00:16:29,009
yet, at least, because we are very,

425
00:16:30,429 --> 00:16:33,490
thoughtful of leaving our tools that we develop

426
00:16:33,754 --> 00:16:36,714
turned on outside of production for, you know,

427
00:16:36,714 --> 00:16:40,075
some some cases, you know, six months after

428
00:16:40,075 --> 00:16:42,334
we've built the tool and kind of watching

429
00:16:42,794 --> 00:16:45,355
the tool outside of production and comparing to

430
00:16:45,355 --> 00:16:46,975
what we're seeing from,

431
00:16:47,889 --> 00:16:48,710
our humans,

432
00:16:49,490 --> 00:16:49,990
managing

433
00:16:50,610 --> 00:16:51,110
that

434
00:16:51,570 --> 00:16:52,070
task,

435
00:16:53,409 --> 00:16:54,549
within production.

436
00:16:55,250 --> 00:16:58,049
What we have seen, however, you know, with

437
00:16:58,049 --> 00:16:59,269
things of,

438
00:17:00,049 --> 00:17:00,789
you know, these

439
00:17:01,304 --> 00:17:04,825
these foundation models changing so quickly and getting

440
00:17:04,825 --> 00:17:05,325
updated,

441
00:17:05,865 --> 00:17:07,964
so quickly. We have seen

442
00:17:08,585 --> 00:17:09,724
things that,

443
00:17:10,505 --> 00:17:11,244
when we

444
00:17:11,625 --> 00:17:12,845
deployed a tool,

445
00:17:13,305 --> 00:17:16,184
again, in in outside of production, we've been

446
00:17:16,184 --> 00:17:18,690
watching it. It, you know, maybe did really,

447
00:17:18,690 --> 00:17:20,150
really well at,

448
00:17:22,769 --> 00:17:26,690
triaging a patient message to the the right

449
00:17:26,690 --> 00:17:27,190
endpoint,

450
00:17:27,809 --> 00:17:30,049
or the right clinician or staff member that

451
00:17:30,049 --> 00:17:31,589
that message should go to.

452
00:17:32,450 --> 00:17:35,384
But, you know, as there's updates on that,

453
00:17:35,625 --> 00:17:38,424
on the foundation model, we're finding that, you

454
00:17:38,424 --> 00:17:41,065
know, some messages are getting caught in the

455
00:17:41,065 --> 00:17:44,105
filters of the foundation model that, you know,

456
00:17:44,105 --> 00:17:47,005
initially weren't getting, caught. And so,

457
00:17:47,369 --> 00:17:48,970
you know, we're having to go in and

458
00:17:48,970 --> 00:17:50,730
try to understand why is it getting caught

459
00:17:50,730 --> 00:17:52,809
in the filter and changing some of our

460
00:17:52,809 --> 00:17:54,890
prompts to make sure it gets to the

461
00:17:55,130 --> 00:17:56,589
to to the right place.

462
00:17:57,049 --> 00:18:00,109
You know, we're really trying to be sensitive

463
00:18:00,809 --> 00:18:01,309
around

464
00:18:03,224 --> 00:18:05,244
what problems we're trying to solve

465
00:18:05,625 --> 00:18:06,125
that

466
00:18:06,664 --> 00:18:09,404
if if the if the model,

467
00:18:09,945 --> 00:18:11,805
or tool was to get it wrong,

468
00:18:12,585 --> 00:18:15,059
and and and, you know, that if there

469
00:18:15,059 --> 00:18:17,240
was a near miss or an actual miss,

470
00:18:17,700 --> 00:18:20,579
the the risk of harm is is gonna

471
00:18:20,579 --> 00:18:23,460
be much lower. And so most of the

472
00:18:23,460 --> 00:18:25,799
tools that we've developed thus far,

473
00:18:26,740 --> 00:18:27,640
there still

474
00:18:28,179 --> 00:18:28,679
absolutely

475
00:18:29,204 --> 00:18:30,724
has to be a human in the loop.

476
00:18:30,724 --> 00:18:31,845
And when I say a human in the

477
00:18:31,845 --> 00:18:33,544
loop, it's one of our clinicians

478
00:18:33,845 --> 00:18:36,744
or one of our staff members and employees,

479
00:18:38,244 --> 00:18:39,384
that validates,

480
00:18:40,164 --> 00:18:40,984
the appropriateness

481
00:18:41,605 --> 00:18:43,924
of of the output. And so having that

482
00:18:43,924 --> 00:18:44,825
extra layer

483
00:18:45,309 --> 00:18:46,690
in, I think, has,

484
00:18:48,109 --> 00:18:50,349
created that safety net so we haven't had

485
00:18:50,349 --> 00:18:52,830
that miss yet. I I I don't think,

486
00:18:52,830 --> 00:18:54,450
you know, we're ready,

487
00:18:55,470 --> 00:18:55,970
culturally,

488
00:18:56,750 --> 00:18:58,609
or from a societal standpoint

489
00:18:58,910 --> 00:19:01,025
yet. Or I'm not even sure if the

490
00:19:01,025 --> 00:19:02,964
technology is ready yet to

491
00:19:03,265 --> 00:19:03,924
be directly

492
00:19:04,305 --> 00:19:08,085
interfacing with with our patients without having that,

493
00:19:09,345 --> 00:19:11,045
clinician in the loop yet.

494
00:19:13,750 --> 00:19:15,910
Got it. Absolutely. I think that's such helpful

495
00:19:15,910 --> 00:19:18,009
perspective. It it really truly illustrates,

496
00:19:18,309 --> 00:19:19,830
some of the the things that we've been

497
00:19:19,830 --> 00:19:21,210
going through today. So,

498
00:19:21,590 --> 00:19:23,509
I I appreciate it. Before we wrap up

499
00:19:23,509 --> 00:19:25,830
here, I'm curious. What is the number one

500
00:19:25,830 --> 00:19:27,315
thing that you're doing right now to set

501
00:19:27,315 --> 00:19:29,654
set up your medicine for long term success?

502
00:19:30,835 --> 00:19:32,994
Yeah. That's an easy question. You know, for

503
00:19:32,994 --> 00:19:34,454
us, you know, it's

504
00:19:34,994 --> 00:19:35,734
your medicine

505
00:19:36,194 --> 00:19:38,274
is not gonna figure this out all by

506
00:19:38,274 --> 00:19:38,774
ourselves.

507
00:19:39,954 --> 00:19:42,274
Industry is not gonna figure this out all

508
00:19:42,274 --> 00:19:42,934
by themselves.

509
00:19:43,369 --> 00:19:46,330
The feds and states are not, from the

510
00:19:46,330 --> 00:19:48,650
government level, are not gonna figure them figure

511
00:19:48,650 --> 00:19:50,029
this out all by themselves.

512
00:19:50,490 --> 00:19:52,670
It's truly gonna be through partnerships.

513
00:19:52,970 --> 00:19:54,109
And so we have

514
00:19:54,570 --> 00:19:57,744
really thought out of of, you know, what

515
00:19:57,744 --> 00:19:58,884
are the the

516
00:19:59,265 --> 00:20:01,525
the the right groups of consortiums,

517
00:20:02,065 --> 00:20:02,964
of industry,

518
00:20:04,224 --> 00:20:04,724
private,

519
00:20:05,265 --> 00:20:05,765
public,

520
00:20:07,105 --> 00:20:09,184
groups that we can join and have a

521
00:20:09,184 --> 00:20:10,244
seat at the table

522
00:20:10,660 --> 00:20:12,920
to kinda talk through some of these

523
00:20:13,539 --> 00:20:17,220
difficult challenges and learn from others about how

524
00:20:17,220 --> 00:20:17,720
they're

525
00:20:18,259 --> 00:20:20,740
overcoming these challenges or the way that they're

526
00:20:20,740 --> 00:20:23,565
thinking about that. And so we've made a

527
00:20:23,565 --> 00:20:26,384
conscious effort and we have protected our time

528
00:20:26,845 --> 00:20:28,525
of a lot of our leaders in our

529
00:20:28,525 --> 00:20:29,025
organizations

530
00:20:29,884 --> 00:20:30,544
to participate

531
00:20:31,005 --> 00:20:31,505
in,

532
00:20:32,125 --> 00:20:33,664
again, you know, these

533
00:20:34,605 --> 00:20:37,164
these consortiums that are forming around AI and

534
00:20:37,164 --> 00:20:39,690
health care and you know, really trying to

535
00:20:39,690 --> 00:20:40,410
get out,

536
00:20:40,809 --> 00:20:42,670
and travel to other health systems

537
00:20:43,289 --> 00:20:45,630
and meet with our industry partners

538
00:20:46,009 --> 00:20:48,650
and really understand how they're trying to tackle

539
00:20:48,650 --> 00:20:50,670
these problems. Because I actually

540
00:20:51,015 --> 00:20:52,634
believe that the future,

541
00:20:53,015 --> 00:20:55,115
I mean, where we're really gonna be transformative

542
00:20:55,335 --> 00:20:56,955
in health care in this country,

543
00:20:57,414 --> 00:20:59,815
is is is gonna be more, you know,

544
00:20:59,815 --> 00:21:02,934
on this sharing of ideas and almost on,

545
00:21:02,934 --> 00:21:06,420
like, an open source concept of of us

546
00:21:06,660 --> 00:21:09,400
across industry, across providers,

547
00:21:09,779 --> 00:21:10,759
across payers,

548
00:21:11,380 --> 00:21:12,359
across government,

549
00:21:13,460 --> 00:21:16,580
you know, giving away some of our our

550
00:21:16,580 --> 00:21:18,820
learnings and sharing it with others so others

551
00:21:18,820 --> 00:21:20,900
can build on top of that. And that's

552
00:21:20,900 --> 00:21:23,144
where I think we're really going to be

553
00:21:23,144 --> 00:21:26,284
able to leverage technology like generative AI to

554
00:21:26,744 --> 00:21:28,984
transform health care for the better of our

555
00:21:28,984 --> 00:21:29,724
our patients

556
00:21:31,224 --> 00:21:34,264
and our clinicians. And so again for us,

557
00:21:34,264 --> 00:21:34,924
you know,

558
00:21:36,039 --> 00:21:38,539
long term success is gonna be through partnerships

559
00:21:39,079 --> 00:21:42,299
and making sure we find those those right

560
00:21:42,359 --> 00:21:45,400
groups that have a similar why and vision

561
00:21:45,400 --> 00:21:46,380
that we have.

562
00:21:49,445 --> 00:21:50,965
I love that. Michael, thank you so much

563
00:21:50,965 --> 00:21:52,565
for joining us on the podcast today. This

564
00:21:52,565 --> 00:21:54,404
has been just a really fun conversation, and

565
00:21:54,404 --> 00:21:56,005
I look forward to connecting with you again

566
00:21:56,005 --> 00:21:56,505
soon.

567
00:21:57,045 --> 00:21:58,904
Thank you, Laura, for having me.