1
00:00:00,010 -->
00:00:01,120
Krishna Gade: Welcome and thank you
2
00:00:01,120 -->
00:00:03,430
for joining us on today's AI explained.
3
00:00:04,240 -->
00:00:06,100
I'm Krishna Gade, the Founder
4
00:00:06,100 -->
00:00:07,360
and CEO of Fiddler AI.
5
00:00:07,450 -->
00:00:08,560
I'll be your host today.
6
00:00:09,280 -->
00:00:10,750
You can put your questions in the q
7
00:00:10,750 -->
00:00:12,850
and a box anytime in the fireside chat.
8
00:00:12,969 -->
00:00:14,230
Uh, today's session will also
9
00:00:14,230 -->
00:00:16,660
be recorded, uh, and send to all
10
00:00:16,660 -->
00:00:17,950
the attendees after the session.
11
00:00:18,070 -->
00:00:18,400
Um.
12
00:00:19,555 -->
00:00:21,445
We have a very special guest today on
13
00:00:21,445 -->
00:00:24,474
AI, and that is Dr. Stacy Johnston,
14
00:00:24,814 -->
00:00:27,615
uh, CIO and digital executive
15
00:00:27,805 -->
00:00:30,145
officer at, uh, Beacon Health System.
16
00:00:30,925 -->
00:00:32,455
And today we are exploring with
17
00:00:32,455 -->
00:00:35,785
her on AI in healthcare and, you
18
00:00:35,785 -->
00:00:39,145
know, ROI, governance, um, controls,
19
00:00:39,235 -->
00:00:41,065
observability, all of those fun things.
20
00:00:41,754 -->
00:00:42,865
Welcome Dr. Johnston.
21
00:00:44,770 -->
00:00:45,129
Dr. Stacey Johnston: All right.
22
00:00:45,129 -->
00:00:46,419
Thank you so much.
23
00:00:47,800 -->
00:00:48,490
Krishna Gade: Dr. Johnston
24
00:00:48,490 -->
00:00:50,590
joined Beacon in October, 2024.
25
00:00:50,590 -->
00:00:52,780
She most recently served as the vice
26
00:00:52,780 -->
00:00:54,430
President, chief Applications Officer for
27
00:00:54,430 -->
00:00:57,419
Baptist Health in Jacksonville, Florida.
28
00:00:57,840 -->
00:00:59,170
Prior to that, she was the Chief
29
00:00:59,170 -->
00:01:00,580
Medical Information Officer.
30
00:01:01,285 -->
00:01:03,685
Uh, at, you know, Beaufort Memorial
31
00:01:03,685 -->
00:01:05,485
Hospital in South Carolina.
32
00:01:06,055 -->
00:01:07,885
During her time in these roles, uh,
33
00:01:07,885 -->
00:01:09,385
she was al she'll always maintain
34
00:01:09,385 -->
00:01:10,915
an active medical staff privileges,
35
00:01:10,945 -->
00:01:13,315
worked as a hospitalist, and she plans
36
00:01:13,315 -->
00:01:15,115
to continue this as, as she feels it,
37
00:01:15,235 -->
00:01:16,975
as, gives her a unique insight into
38
00:01:16,975 -->
00:01:18,715
how technology and care in intersect.
39
00:01:19,235 -->
00:01:20,745
And prior to this, she received her
40
00:01:20,755 -->
00:01:21,985
bachelor's degree from Washington
41
00:01:21,985 -->
00:01:24,805
College and Doctor of Medicine at Emory.
42
00:01:25,015 -->
00:01:26,965
So great to, uh.
43
00:01:27,835 -->
00:01:30,265
You know, great background and it's, it's,
44
00:01:30,505 -->
00:01:32,965
it's great to see, you know, technologists
45
00:01:32,965 -->
00:01:34,105
coming from, you know, different
46
00:01:34,105 -->
00:01:36,205
backgrounds and, and, and, uh, very
47
00:01:36,205 -->
00:01:38,210
excited to host you here, Dr. Johnston.
48
00:01:38,275 -->
00:01:38,605
Um,
49
00:01:39,055 -->
00:01:39,265
Dr. Stacey Johnston: yeah,
50
00:01:39,505 -->
00:01:40,615
again, thanks for having me.
51
00:01:40,615 -->
00:01:41,425
And now my camera's
52
00:01:41,425 -->
00:01:42,895
working, so we're good now.
53
00:01:42,895 -->
00:01:43,315
Krishna Gade: Awesome.
54
00:01:43,945 -->
00:01:44,635
So, so I guess let's
55
00:01:44,635 -->
00:01:45,565
start with that, right?
56
00:01:45,565 -->
00:01:47,095
Like you started as a clinician.
57
00:01:47,245 -->
00:01:49,315
Uh, what moment made you realize
58
00:01:49,315 -->
00:01:50,635
that healthcare needed systems
59
00:01:50,635 -->
00:01:51,895
level change and, you know, not
60
00:01:51,895 -->
00:01:53,695
just, uh, you know, bedside care?
61
00:01:54,805 -->
00:01:55,855
Dr. Stacey Johnston: So, honestly,
62
00:01:55,855 -->
00:01:56,995
a lot of this actually started
63
00:01:56,995 -->
00:01:58,465
straight out of residency.
64
00:01:58,585 -->
00:02:01,884
Um, so, you know, back in the day, uh,
65
00:02:01,914 -->
00:02:03,685
we were still on paper health, uh, paper
66
00:02:03,685 -->
00:02:07,524
records and, um, and then, uh, you know,
67
00:02:07,524 -->
00:02:09,415
we were switching over to electronic
68
00:02:09,415 -->
00:02:11,695
health records in residency and have,
69
00:02:11,695 -->
00:02:13,345
of course, at that time, many attendings
70
00:02:13,345 -->
00:02:14,725
were not super excited about it.
71
00:02:15,085 -->
00:02:16,195
Of course they gave it to the
72
00:02:16,195 -->
00:02:17,785
residents to try to figure it out.
73
00:02:18,265 -->
00:02:20,005
And even the residents were raising
74
00:02:20,005 -->
00:02:22,435
concerns and saying, you know, one
75
00:02:22,435 -->
00:02:24,325
thing I've heard a lot over my tenure
76
00:02:24,325 -->
00:02:27,145
in this role is, um, you know, we are
77
00:02:27,145 -->
00:02:28,855
the highest paid secretaries out there.
78
00:02:28,855 -->
00:02:31,405
And so this idea of putting in our
79
00:02:31,405 -->
00:02:34,375
orders, typing them in, um, you know,
80
00:02:34,375 -->
00:02:36,475
which just went against everything.
81
00:02:36,475 -->
00:02:38,245
A lot of people were trained, you
82
00:02:38,245 -->
00:02:40,495
know, to do, um, at that time.
83
00:02:40,735 -->
00:02:42,325
And so, um.
84
00:02:43,060 -->
00:02:44,800
And so what I realized was
85
00:02:44,800 -->
00:02:47,290
either I could buck the system.
86
00:02:47,649 -->
00:02:49,180
I mean, this was coming whether we
87
00:02:49,180 -->
00:02:50,829
wanted it to or not at this point
88
00:02:50,829 -->
00:02:52,720
in time as a regulatory requirement.
89
00:02:53,260 -->
00:02:54,910
And so, you know, I could
90
00:02:54,910 -->
00:02:56,260
either buck it or embrace it.
91
00:02:56,320 -->
00:02:58,240
And so what I realized was if I
92
00:02:58,240 -->
00:03:00,760
embraced it, it, it made my life easier.
93
00:03:01,089 -->
00:03:02,290
You know, the more I embraced it,
94
00:03:02,290 -->
00:03:04,150
the more time I spent in developing.
95
00:03:04,960 -->
00:03:06,760
My documentation templates or my
96
00:03:06,760 -->
00:03:09,070
order sets just kind of made my
97
00:03:09,070 -->
00:03:11,050
workflow easier and therefore made
98
00:03:11,050 -->
00:03:12,820
my colleagues' workflow easier.
99
00:03:12,820 -->
00:03:14,230
So at that time I was in family
100
00:03:14,230 -->
00:03:16,090
medicine and then became an
101
00:03:16,090 -->
00:03:17,440
attending as hospitalist.
102
00:03:17,440 -->
00:03:18,610
And interestingly enough,
103
00:03:18,610 -->
00:03:19,540
I went back to paper.
104
00:03:20,350 -->
00:03:23,260
And so then, uh, what led a second
105
00:03:23,260 -->
00:03:25,870
round of implementation of paper to an
106
00:03:25,870 -->
00:03:27,790
electronic health record by that time.
107
00:03:28,300 -->
00:03:29,740
I was the only person in that health
108
00:03:29,740 -->
00:03:32,110
system at Beaufort, uh, that had
109
00:03:32,110 -->
00:03:35,170
been on, um, any sort of CPOE.
110
00:03:35,170 -->
00:03:37,270
So next thing I know, I became the
111
00:03:37,270 -->
00:03:40,000
Chief Medical Information Officer and,
112
00:03:40,060 -->
00:03:42,010
um, what it, you know, that it was that
113
00:03:42,010 -->
00:03:44,140
mindset of, okay, if we spend time in
114
00:03:44,140 -->
00:03:46,360
the beginning really evaluating our
115
00:03:46,360 -->
00:03:48,940
workflows, um, trying to make them better.
116
00:03:49,690 -->
00:03:51,519
You know, by using the systems,
117
00:03:51,820 -->
00:03:53,380
how can it, you know, how can
118
00:03:53,380 -->
00:03:54,760
it save us time in the long run?
119
00:03:54,820 -->
00:03:57,700
And so versus if we don't put what I
120
00:03:57,700 -->
00:03:59,800
call care and feeding into the EHR,
121
00:04:00,190 -->
00:04:01,720
then it's just gonna, you know, all
122
00:04:01,720 -->
00:04:04,180
that pajama time, all of that, you know,
123
00:04:04,420 -->
00:04:06,910
that was happening in, in real life.
124
00:04:07,480 -->
00:04:10,660
And so I really, um, spent a lot of
125
00:04:10,660 -->
00:04:12,219
time on the hospitalist workflows
126
00:04:12,219 -->
00:04:13,750
and on their order sets and, um.
127
00:04:14,695 -->
00:04:17,125
You know, uh, automation and using
128
00:04:17,125 -->
00:04:19,615
Dragon Medical one and the templates,
129
00:04:19,615 -->
00:04:21,295
and so really building up our templates.
130
00:04:22,360 -->
00:04:24,040
Really building up auto text.
131
00:04:24,370 -->
00:04:26,620
And so the more we could, and that, that
132
00:04:26,620 -->
00:04:28,360
was the version of automating back then.
133
00:04:28,360 -->
00:04:30,190
Mm. You know, it was templates and,
134
00:04:30,250 -->
00:04:32,800
um, and even the Dragon Medical one,
135
00:04:32,800 -->
00:04:35,350
and then we went to the DMO mobile app.
136
00:04:35,350 -->
00:04:36,190
I mean, we're like, oh
137
00:04:36,190 -->
00:04:37,690
my gosh, this is amazing.
138
00:04:37,690 -->
00:04:39,790
And like AI, AI wasn't even,
139
00:04:40,300 -->
00:04:43,240
uh, and you know what thought?
140
00:04:43,420 -->
00:04:43,780
Yeah.
141
00:04:43,780 -->
00:04:45,610
I mean, we, no one even thought
142
00:04:45,610 -->
00:04:47,050
that that was even possible.
143
00:04:47,470 -->
00:04:49,210
Um, what, what I realized was.
144
00:04:49,540 -->
00:04:52,060
You know, taking the time to make it
145
00:04:52,060 -->
00:04:55,270
better for the physicians ultimately
146
00:04:55,300 -->
00:04:57,220
ended up making it better for patients.
147
00:04:57,250 -->
00:05:00,400
You know, um, that decreasing time in the
148
00:05:00,400 -->
00:05:03,520
EHR obviously is, you know, that, that
149
00:05:03,550 -->
00:05:05,980
EHR can contribute to burnout and burnout
150
00:05:05,980 -->
00:05:08,530
obviously leads to a lesser quality care.
151
00:05:09,040 -->
00:05:10,720
And so we really, you know,
152
00:05:10,720 -->
00:05:11,975
initially that triple aim.
153
00:05:13,105 -->
00:05:14,455
Uh, that you might have heard
154
00:05:14,455 -->
00:05:16,165
about is, you know, good quality
155
00:05:16,165 -->
00:05:18,115
care, reduced cost and you know,
156
00:05:18,115 -->
00:05:20,425
patient, um, more about the patient.
157
00:05:20,425 -->
00:05:22,075
And then adding in the quadruple
158
00:05:22,075 -->
00:05:24,265
aim was really talking about
159
00:05:24,265 -->
00:05:25,615
enhancing the caregiver experience.
160
00:05:25,615 -->
00:05:27,145
So I really took that to heart
161
00:05:27,235 -->
00:05:28,615
and, you know, making the
162
00:05:28,615 -->
00:05:30,115
caregiver experience better.
163
00:05:30,565 -->
00:05:31,945
And then now they're talking about the
164
00:05:31,945 -->
00:05:34,135
quintuple aim, which is equity in care.
165
00:05:34,135 -->
00:05:35,665
So, you know, I still think
166
00:05:35,665 -->
00:05:37,045
we have a lot of room to go.
167
00:05:37,540 -->
00:05:39,460
Especially in those last two components,
168
00:05:39,520 -->
00:05:40,810
you know, enhancing the caregiver
169
00:05:40,810 -->
00:05:43,120
experience and then in the equity of care.
170
00:05:43,690 -->
00:05:45,610
But I really think that, you know, now.
171
00:05:46,435 -->
00:05:49,615
In, in this AI era we are going, we're
172
00:05:49,615 -->
00:05:51,925
just getting the glimpse of it, like
173
00:05:51,925 -->
00:05:54,175
just a little taste of it, you know?
174
00:05:54,205 -->
00:05:55,765
Um, we're doing a lot, you know,
175
00:05:55,825 -->
00:05:58,435
radiology's embracing AI pretty well.
176
00:05:58,675 -->
00:06:01,255
Um, there are some cool note summarizer
177
00:06:01,255 -->
00:06:03,085
type of things, but, um, but we're
178
00:06:03,085 -->
00:06:06,325
not full AI in the clinical space yet.
179
00:06:06,685 -->
00:06:09,235
And I think once we get there, you
180
00:06:09,235 -->
00:06:10,495
know, of course through, you know,
181
00:06:10,495 -->
00:06:12,595
change management and appropriate data
182
00:06:12,595 -->
00:06:15,645
and trusting of the systems, um you
183
00:06:15,645 -->
00:06:17,470
know we're still, I think, you know,
184
00:06:17,740 -->
00:06:19,540
maybe a year or two out from that.
185
00:06:19,540 -->
00:06:21,040
But once we get there, I
186
00:06:21,040 -->
00:06:22,150
think the sky's the limit.
187
00:06:22,150 -->
00:06:23,860
And I think the docs are really gonna
188
00:06:23,860 -->
00:06:27,250
see time spent in the EHR is decreasing.
189
00:06:27,280 -->
00:06:29,140
They're going home on time, you know?
190
00:06:29,445 -->
00:06:30,790
They're, they're starting to find joy
191
00:06:30,790 -->
00:06:33,130
in medicine, you know, really embracing
192
00:06:33,550 -->
00:06:36,100
ambient, but more than just documentation,
193
00:06:36,160 -->
00:06:37,540
you know, from the ambient listening,
194
00:06:37,540 -->
00:06:39,550
but really doing your full workflow
195
00:06:39,670 -->
00:06:42,010
with an ambient solution, I think is,
196
00:06:42,370 -->
00:06:43,780
is ultimately where we need to go.
197
00:06:44,275 -->
00:06:45,265
Krishna Gade: No, it's, it's great to
198
00:06:45,265 -->
00:06:47,965
see technology helping, uh, practitioners
199
00:06:47,965 -->
00:06:51,055
that way right now it comes to, no, no.
200
00:06:51,055 -->
00:06:52,435
So it's an interesting, you know,
201
00:06:52,435 -->
00:06:53,785
because you had this two, uh,
202
00:06:53,905 -->
00:06:55,224
kind of split level experience,
203
00:06:55,224 -->
00:06:56,305
in some ways they're connected.
204
00:06:56,305 -->
00:06:58,525
But what do you see, you know,
205
00:06:58,555 -->
00:06:59,724
like technology, pure play
206
00:06:59,724 -->
00:07:01,974
technologies fundamentally
207
00:07:02,004 -->
00:07:03,265
misunderstanding about clinical?
208
00:07:03,436 -->
00:07:03,726
Dr. Stacey Johnston: Yeah,
209
00:07:03,781 -->
00:07:06,541
I, I think ultimately, um,
210
00:07:06,546 -->
00:07:09,811
spearheading technology is.
211
00:07:09,856 -->
00:07:13,066
It's not a technology issue,
212
00:07:13,396 -->
00:07:14,926
it is a workflow issue.
213
00:07:14,926 -->
00:07:16,516
It is a change management issue.
214
00:07:16,516 -->
00:07:17,746
It's a training issue.
215
00:07:18,166 -->
00:07:21,196
And so if you do not understand and
216
00:07:21,196 -->
00:07:23,566
embrace the workflows, um, if you
217
00:07:23,566 -->
00:07:25,126
don't understand the needs for change
218
00:07:25,126 -->
00:07:27,676
management or if you don't really have a
219
00:07:27,676 -->
00:07:30,736
good training program, um, around these
220
00:07:30,736 -->
00:07:33,556
AI solutions, then you're just gonna put
221
00:07:33,556 -->
00:07:35,266
'em in and they're gonna fail, you know?
222
00:07:35,266 -->
00:07:36,466
And so there are gonna
223
00:07:36,466 -->
00:07:38,086
be a lot of times, um.
224
00:07:38,491 -->
00:07:41,101
Where we, you know, where we implement
225
00:07:41,131 -->
00:07:42,421
and it's not meeting our needs.
226
00:07:42,481 -->
00:07:43,861
Well, why isn't it meeting our needs?
227
00:07:43,861 -->
00:07:45,511
Well, chances are we put in something
228
00:07:45,511 -->
00:07:47,821
because, you know, maybe the operators
229
00:07:47,821 -->
00:07:49,591
thought it would help solve a problem.
230
00:07:49,951 -->
00:07:51,421
But ultimately, you know, one of the
231
00:07:51,421 -->
00:07:53,701
things as being in the informatics is, is
232
00:07:53,701 -->
00:07:55,351
you learn, you know, to ask a question,
233
00:07:55,381 -->
00:07:56,851
what problems are you trying to solve?
234
00:07:56,941 -->
00:07:59,941
And maybe the problems of the physicians
235
00:07:59,941 -->
00:08:01,771
aren't necessarily aligning with your.
236
00:08:02,476 -->
00:08:03,886
Hospital operators.
237
00:08:04,246 -->
00:08:05,956
And so you need to make sure that
238
00:08:05,956 -->
00:08:08,416
we're in alignment from a system, what
239
00:08:08,416 -->
00:08:10,426
problems we truly are trying to solve.
240
00:08:10,846 -->
00:08:12,496
And that, you know, if you can get
241
00:08:12,496 -->
00:08:14,116
your, you know, your service line
242
00:08:14,116 -->
00:08:16,246
leaders, your physicians all to be
243
00:08:16,246 -->
00:08:18,316
in alignment and, you know, maybe
244
00:08:18,316 -->
00:08:20,356
it is about access for patients.
245
00:08:20,686 -->
00:08:22,396
Um, and maybe it's about, you know,
246
00:08:22,426 -->
00:08:24,136
physician burnout, but whatever
247
00:08:24,136 -->
00:08:25,426
problem you're trying to solve, we
248
00:08:25,426 -->
00:08:26,776
need to make sure that you're aligned
249
00:08:26,776 -->
00:08:29,266
from a system and you have a singular
250
00:08:29,266 -->
00:08:31,396
system approach to that problem.
251
00:08:31,776 -->
00:08:34,326
Before trying to deploy the solutions.
252
00:08:35,136 -->
00:08:35,646
Krishna Gade: Interesting.
253
00:08:36,156 -->
00:08:37,896
And now we are in the biggest technology
254
00:08:37,896 -->
00:08:39,996
shift, uh, now, now era Right.
255
00:08:39,996 -->
00:08:41,496
You know, everyone is saying that AI
256
00:08:41,496 -->
00:08:44,316
will transform healthcare, especially
257
00:08:44,316 -->
00:08:46,386
in, and, and from your perspective, you
258
00:08:46,386 -->
00:08:48,156
know, what's actually working today?
259
00:08:48,216 -->
00:08:50,076
You know, is it still kind of PowerPoint
260
00:08:50,076 -->
00:08:52,386
decks or is it, are there any interesting
261
00:08:52,446 -->
00:08:54,156
applications that you've seen work?
262
00:08:54,186 -->
00:08:55,626
Uh, you know, and giving ROI.
263
00:08:57,336 -->
00:08:58,046
Dr. Stacey Johnston: Yeah, so, you know,
264
00:08:58,046 -->
00:09:00,406
I, I think most health systems are pretty,
265
00:09:00,496 -->
00:09:02,536
are advancing themselves in the backend
266
00:09:02,536 -->
00:09:05,686
pieces, you know, um, automatic coding,
267
00:09:05,896 -->
00:09:08,686
um, you know, we are pursuing fully
268
00:09:08,686 -->
00:09:10,516
autonomous coding, which I think would
269
00:09:10,516 -->
00:09:12,736
be, you know, a game changer for us.
270
00:09:13,156 -->
00:09:14,836
Um, right now it's still mostly
271
00:09:14,836 -->
00:09:16,756
an augmented coding aspect.
272
00:09:17,146 -->
00:09:19,246
Um, we are moving forward.
273
00:09:19,346 -->
00:09:21,626
Towards, um, fully, um, autonomous
274
00:09:21,626 -->
00:09:24,116
verification of benefits, um, where
275
00:09:24,116 -->
00:09:26,516
an agentic AI solution is gonna do all
276
00:09:26,516 -->
00:09:28,376
of the, um, verification of benefits.
277
00:09:28,376 -->
00:09:30,276
We're also looking at, um, a
278
00:09:30,316 -->
00:09:33,736
agentic, um have call return.
279
00:09:34,016 -->
00:09:36,151
And so agents actually participating
280
00:09:36,151 -->
00:09:38,311
in, you know, voice call returns and,
281
00:09:38,371 -->
00:09:40,051
you know, interacting with the patients.
282
00:09:40,441 -->
00:09:41,671
Uh, so there are some really cool
283
00:09:41,671 -->
00:09:43,441
solutions out there doing that.
284
00:09:43,891 -->
00:09:45,901
Um, other areas where we've actually,
285
00:09:45,961 -->
00:09:48,181
um, expanded into agent use, um,
286
00:09:48,211 -->
00:09:50,101
fully autonomous agent use is, um.
287
00:09:51,266 -->
00:09:53,606
When we converted, we actually just did
288
00:09:53,606 -->
00:09:56,306
a four hospital acquisition, uh, recently
289
00:09:56,606 -->
00:09:58,676
and we put them on our tech stack in
290
00:09:58,736 -->
00:10:00,656
seven months, which was pretty aggressive,
291
00:10:01,076 -->
00:10:03,446
but you know, you only get this system
292
00:10:03,446 -->
00:10:05,066
available to back load appointments.
293
00:10:05,456 -->
00:10:07,616
We got it about two weeks of, uh, before
294
00:10:07,616 -->
00:10:10,166
we went live and two weeks to back
295
00:10:10,166 -->
00:10:11,516
load a hundred thousand appointments
296
00:10:11,576 -->
00:10:13,286
is not necessarily enough time.
297
00:10:13,286 -->
00:10:14,666
So we had to think differently.
298
00:10:14,666 -->
00:10:16,946
We could either hire 40 people to do this
299
00:10:17,366 -->
00:10:19,616
or we could use, um, some agentic AI.
300
00:10:20,656 -->
00:10:24,376
And so we were able, um, to basically
301
00:10:24,496 -->
00:10:26,446
do all of the ambulatory back
302
00:10:26,446 -->
00:10:28,186
load appointments with an agent.
303
00:10:28,186 -->
00:10:30,046
It took us about three weeks to build the
304
00:10:30,046 -->
00:10:33,256
agent, and then two weeks, um, from that
305
00:10:33,256 -->
00:10:35,116
time that the system became available
306
00:10:35,146 -->
00:10:36,946
to back load all of the appointments.
307
00:10:37,276 -->
00:10:39,316
Um, so that was, that was game
308
00:10:39,316 -->
00:10:41,106
changing and that was fully agent,
309
00:10:41,126 -->
00:10:45,106
a AI and, um, doing all of our, all
310
00:10:45,106 -->
00:10:46,516
of our scheduling for our patients.
311
00:10:46,936 -->
00:10:48,226
So I think that's pretty cool.
312
00:10:48,226 -->
00:10:49,156
We're looking at some.
313
00:10:49,816 -->
00:10:53,116
Um, digital FTEs and, um, you know,
314
00:10:53,176 -->
00:10:55,246
any, all these repeatable processes.
315
00:10:55,756 -->
00:10:58,126
We have a pretty good, robust, robust, um,
316
00:10:58,156 -->
00:11:00,646
robotic, um, process automation program.
317
00:11:00,646 -->
00:11:02,536
We have a good intake process and,
318
00:11:02,926 -->
00:11:04,906
and a governance around the RPA.
319
00:11:04,936 -->
00:11:07,756
But imagine laying, um, an age intake
320
00:11:07,816 -->
00:11:09,976
AI solution on top of your RPA.
321
00:11:10,636 -->
00:11:11,776
Um, we're not there yet, but I
322
00:11:11,776 -->
00:11:13,456
think that's just, you know, that's
323
00:11:13,456 -->
00:11:14,746
right around the corner as well too.
324
00:11:14,746 -->
00:11:16,036
So those are kind of all
325
00:11:16,036 -->
00:11:17,596
of your backend processes.
326
00:11:17,986 -->
00:11:18,886
And then when you think about.
327
00:11:19,516 -->
00:11:20,626
Yeah, when you think about
328
00:11:20,626 -->
00:11:22,576
the clinical aspects, um.
329
00:11:23,686 -->
00:11:25,726
We actually have a white paper that
330
00:11:25,756 -->
00:11:29,046
has been written about, um, uh, agentic
331
00:11:29,086 -->
00:11:31,666
AI and colon colon cancer screening.
332
00:11:31,666 -->
00:11:33,466
So really embracing the use of
333
00:11:33,496 -->
00:11:36,196
Cologuard and, uh, and as an
334
00:11:36,196 -->
00:11:37,456
initial screening instead of
335
00:11:37,456 -->
00:11:39,646
automatically going to a colonoscopy.
336
00:11:40,096 -->
00:11:42,226
Um, and a lot of this has to do with
337
00:11:42,226 -->
00:11:44,686
access a colonoscopies and Cologuard
338
00:11:44,686 -->
00:11:46,606
is now being viewed as, you know,
339
00:11:46,606 -->
00:11:48,826
a, a good first, um, solution.
340
00:11:49,401 -->
00:11:52,401
And so we, um, built an agent
341
00:11:52,461 -->
00:11:54,351
that automatically ordered
342
00:11:54,351 -->
00:11:56,391
Cologuard for the, this group
343
00:11:56,391 -->
00:11:58,581
of, um, patients that qualified.
344
00:11:58,581 -->
00:12:00,651
So we ordered 7,000.
345
00:12:00,711 -->
00:12:00,891
Um.
346
00:12:01,786 -->
00:12:03,166
Cologuard screenings.
347
00:12:03,226 -->
00:12:05,926
Um, and, you know, ended up getting,
348
00:12:06,226 -->
00:12:07,816
uh, back, you know, I think it was
349
00:12:07,816 -->
00:12:11,626
about 40% returned and ended up, uh,
350
00:12:11,656 -->
00:12:15,256
doing 250 colon cancer screenings,
351
00:12:15,316 -->
00:12:17,686
um, additional with a colonoscopy.
352
00:12:17,896 -->
00:12:18,976
And we did actually have one
353
00:12:18,976 -->
00:12:21,166
positive colon cancer, um, returned.
354
00:12:21,166 -->
00:12:22,246
That was caught early.
355
00:12:22,576 -->
00:12:25,216
Um, so had a resection and full survival.
356
00:12:25,216 -->
00:12:27,841
So I, I mean, that's just that is game
357
00:12:27,841 -->
00:12:30,001
changing, you know, clinical care where,
358
00:12:30,031 -->
00:12:32,161
uh, agentic AI solution can actually
359
00:12:32,161 -->
00:12:34,501
help, you know, help prevent, um, you
360
00:12:34,501 -->
00:12:37,021
know, further, um, further mortality.
361
00:12:37,021 -->
00:12:39,136
So I, I, I think that's pretty cool stuff.
362
00:12:40,126 -->
00:12:40,936
Krishna Gade: Wow, that's awesome.
363
00:12:40,936 -->
00:12:42,316
So automating appointments
364
00:12:42,316 -->
00:12:44,206
to clinical diagnosis.
365
00:12:44,206 -->
00:12:46,636
So, uh, what is like, maybe like,
366
00:12:46,636 -->
00:12:48,226
you know, maybe you can give us like
367
00:12:48,496 -->
00:12:50,566
audience also, like what's one use
368
00:12:50,566 -->
00:12:52,516
case that's often over hyped in,
369
00:12:52,516 -->
00:12:54,856
in, and, and also what's like, you
370
00:12:54,856 -->
00:12:57,136
think that is underappreciated, you
371
00:12:57,136 -->
00:12:58,096
know, both, you know, like in your
372
00:12:58,126 -->
00:13:00,211
experience that you've seen with AI.
373
00:13:00,231 -->
00:13:00,451
Dr. Stacey Johnston: Yes.
374
00:13:01,096 -->
00:13:03,586
Um, so one use case, so obviously
375
00:13:03,586 -->
00:13:05,596
radiology, a lot of the radiology
376
00:13:05,596 -->
00:13:07,336
use cases are fully embraced.
377
00:13:07,786 -->
00:13:10,876
Um, I don't know that, I don't know
378
00:13:10,876 -->
00:13:12,436
of any organizations that are fully
379
00:13:12,436 -->
00:13:14,476
autonomous with radiology yet.
380
00:13:14,806 -->
00:13:16,126
Um, we are using it as an
381
00:13:16,126 -->
00:13:17,806
augmentation to the workflow.
382
00:13:17,806 -->
00:13:20,206
So I think that's one that isn't
383
00:13:20,206 -->
00:13:22,966
overhyped, but is definitely used, you
384
00:13:22,966 -->
00:13:24,826
know, throughout multiple health systems.
385
00:13:25,636 -->
00:13:27,706
Um, and one that I think is
386
00:13:28,321 -->
00:13:31,501
underappreciated, um, is the augmented
387
00:13:31,501 -->
00:13:34,021
responses to the in basket messaging.
388
00:13:34,411 -->
00:13:35,641
Um, and I think part of it is
389
00:13:35,641 -->
00:13:38,641
just understanding, you know,
390
00:13:38,641 -->
00:13:40,951
the large language model and, you
391
00:13:40,951 -->
00:13:43,111
know, having to say, well, I have
392
00:13:43,111 -->
00:13:45,241
to modify the message anyway.
393
00:13:45,961 -->
00:13:48,301
But yeah, modifying a message that's
394
00:13:48,301 -->
00:13:49,711
already written up is very different
395
00:13:49,711 -->
00:13:51,331
than typing a brand new message.
396
00:13:51,331 -->
00:13:53,881
So I think that's one that we should
397
00:13:53,881 -->
00:13:57,181
continue to explore and push ourselves on.
398
00:13:57,181 -->
00:13:58,621
You know, there are studies out there
399
00:13:58,621 -->
00:14:01,081
that show that our, um, especially our
400
00:14:01,081 -->
00:14:04,291
primary care docs are getting 85% more
401
00:14:04,291 -->
00:14:07,521
in basket messages from the portal, these
402
00:14:07,531 -->
00:14:09,946
portal messages are, um, since COVID.
403
00:14:10,336 -->
00:14:12,136
And so, you know, that's additional
404
00:14:12,136 -->
00:14:13,966
burden, again, the cognitive
405
00:14:13,966 -->
00:14:15,286
burden for our physicians.
406
00:14:15,616 -->
00:14:17,956
But yet also it's not being,
407
00:14:18,316 -->
00:14:19,756
it's not being reimbursed.
408
00:14:19,966 -->
00:14:21,316
It's not, you know, it's just
409
00:14:21,346 -->
00:14:23,776
additional time for our providers.
410
00:14:24,166 -->
00:14:24,946
The other thing is we
411
00:14:24,946 -->
00:14:26,116
need to look at the pools.
412
00:14:26,186 -->
00:14:28,106
For these messages, do they, does
413
00:14:28,106 -->
00:14:29,636
this particular message really
414
00:14:29,636 -->
00:14:31,286
need to go to that provider?
415
00:14:31,286 -->
00:14:33,116
Or can it go to an MA? Can it
416
00:14:33,116 -->
00:14:34,766
go to the front office staff?
417
00:14:34,766 -->
00:14:37,586
So I think routing the messages better,
418
00:14:37,586 -->
00:14:39,896
and then when it truly is a provider
419
00:14:39,896 -->
00:14:42,536
message, having the AI help queue it up.
420
00:14:42,716 -->
00:14:45,366
Um, I, I think also some referral
421
00:14:45,366 -->
00:14:47,236
only, um, referral management
422
00:14:47,516 -->
00:14:49,886
is not as well utilized with AI.
423
00:14:50,306 -->
00:14:52,136
Um, we're, we're diving
424
00:14:52,136 -->
00:14:54,931
into that refill management.
425
00:14:54,931 -->
00:14:56,461
So refill medications, I really
426
00:14:56,461 -->
00:14:58,621
feel like as an opportunity with AI.
427
00:14:58,831 -->
00:15:00,121
Again, cognitive burden.
428
00:15:00,301 -->
00:15:02,401
What is a repetitive task, you know?
429
00:15:02,401 -->
00:15:04,591
And so can the AI queue up the refills
430
00:15:04,591 -->
00:15:06,271
that are ready so the docs aren't having
431
00:15:06,271 -->
00:15:08,671
to click on those boxes all the time.
432
00:15:08,671 -->
00:15:10,201
So there are some solutions out
433
00:15:10,201 -->
00:15:11,281
there that are doing some pretty
434
00:15:11,281 -->
00:15:13,401
cool stuff with, um, refill meds.
435
00:15:14,476 -->
00:15:14,836
Krishna Gade: Awesome.
436
00:15:15,196 -->
00:15:16,186
So, uh, we touched upon
437
00:15:16,186 -->
00:15:17,026
a few things, right?
438
00:15:17,026 -->
00:15:18,646
You know, uh, uh, you know, you
439
00:15:18,646 -->
00:15:20,236
leveraging large language models to
440
00:15:20,236 -->
00:15:23,356
respond on behalf of doctors or even
441
00:15:23,356 -->
00:15:24,976
augment, you know, their answers.
442
00:15:25,366 -->
00:15:27,556
Now, one of the problems with LLMs
443
00:15:27,616 -->
00:15:29,416
is that there are a bunch of risks.
444
00:15:29,416 -->
00:15:31,006
You know, you know, they can hallucinate
445
00:15:31,006 -->
00:15:33,046
at times, you know, they could, you know,
446
00:15:33,076 -->
00:15:35,146
leak PII information and all of that,
447
00:15:35,146 -->
00:15:37,936
or PHI in the case of healthcare, uh.
448
00:15:38,621 -->
00:15:39,671
How do you, you know, when you
449
00:15:39,671 -->
00:15:41,051
joined, you know, AI, how did you
450
00:15:41,051 -->
00:15:42,941
approach this problem of AI governance
451
00:15:43,181 -->
00:15:44,171
and, and where did you begin?
452
00:15:44,291 -->
00:15:44,501
You know,
453
00:15:46,031 -->
00:15:46,241
Dr. Stacey Johnston: uh,
454
00:15:46,241 -->
00:15:47,621
yeah, so great question.
455
00:15:47,801 -->
00:15:53,281
Um, so we at Beacon, um, upon my
456
00:15:53,681 -->
00:15:54,941
arrival didn't really have a good
457
00:15:54,941 -->
00:15:56,561
strong IS governance process.
458
00:15:56,561 -->
00:16:00,831
So it was, I was either the approvers or
459
00:16:00,831 -->
00:16:01,811
not approvers, and it was kind of just.
460
00:16:02,266 -->
00:16:03,526
Tickets were submitted and
461
00:16:03,616 -->
00:16:05,206
people didn't hear back, and it
462
00:16:05,206 -->
00:16:06,946
was just not a great process.
463
00:16:07,336 -->
00:16:09,136
So that's the, actually, the very
464
00:16:09,136 -->
00:16:10,696
first thing that I did when I came here
465
00:16:10,696 -->
00:16:12,456
was stand up an IS governanceprocess.
466
00:16:12,556 -->
00:16:14,116
So we have an executive steering
467
00:16:14,116 -->
00:16:16,066
committee, we have eight different
468
00:16:16,066 -->
00:16:17,566
advisory levels, and then we have
469
00:16:17,566 -->
00:16:18,946
multiple work groups that report
470
00:16:18,946 -->
00:16:20,476
up to the advisory councils.
471
00:16:20,866 -->
00:16:22,636
Um, one of the advisory councils that
472
00:16:22,636 -->
00:16:25,186
we did stand up as an AI council, um,
473
00:16:25,216 -->
00:16:27,826
and in that AI council we developed, um.
474
00:16:28,756 -->
00:16:30,016
Two AI policies.
475
00:16:30,376 -->
00:16:32,296
One's basically an AI policy,
476
00:16:32,296 -->
00:16:33,916
meaning that, you know, you don't
477
00:16:33,916 -->
00:16:35,686
go out and just sign up for AI.
478
00:16:35,956 -->
00:16:37,186
Um, you know, it has to come through
479
00:16:37,186 -->
00:16:39,076
the advisory council for approval.
480
00:16:39,376 -->
00:16:42,406
Um, and that, you know, if you're, if
481
00:16:42,676 -->
00:16:44,776
you, we have, uh, if you use AI, it
482
00:16:44,776 -->
00:16:46,516
has to be like a system approved AI.
483
00:16:46,516 -->
00:16:48,226
So kind of more about like how to
484
00:16:48,226 -->
00:16:51,466
use AI and then the, um, and then we
485
00:16:51,466 -->
00:16:53,506
have a, a permitted and prohibitive
486
00:16:53,626 -->
00:16:56,686
use policy, meaning I can't put.
487
00:16:57,556 -->
00:17:02,026
PHI into ChatGPT or into Open Evidence.
488
00:17:02,086 -->
00:17:04,186
Um, something along those lines, but that
489
00:17:04,186 -->
00:17:06,466
we do have permitted clinical uses, which
490
00:17:06,466 -->
00:17:08,746
is more like augmenting the workflow.
491
00:17:08,986 -->
00:17:10,726
So there's ambient listening, you
492
00:17:10,726 -->
00:17:12,226
know, the clinical oracle's clinical
493
00:17:12,226 -->
00:17:14,506
AI agent, but you might use a bridge
494
00:17:14,806 -->
00:17:17,356
or, you know, um, DAX, uh, copilot.
495
00:17:17,566 -->
00:17:18,946
So you know, some of those
496
00:17:18,946 -->
00:17:20,386
solutions that you know.
497
00:17:20,766 -->
00:17:22,356
Are augmenting your workflows.
498
00:17:22,356 -->
00:17:24,066
So you're still reviewing and you're
499
00:17:24,066 -->
00:17:26,016
signing off on that documentation.
500
00:17:26,286 -->
00:17:27,846
So in the radiology space, that's
501
00:17:27,846 -->
00:17:29,316
an augmentation to the workflow
502
00:17:29,316 -->
00:17:31,326
because it queues up the severity.
503
00:17:31,656 -->
00:17:33,636
So, you know, if an, um.
504
00:17:34,231 -->
00:17:35,191
That they get a, you know,
505
00:17:35,281 -->
00:17:36,631
hundreds of CTs a day.
506
00:17:37,021 -->
00:17:39,001
It will do the read first and then
507
00:17:39,001 -->
00:17:40,831
it will queue up to the radiologist.
508
00:17:40,831 -->
00:17:42,871
These are ones with results that
509
00:17:42,871 -->
00:17:45,241
we think you need to review first.
510
00:17:45,481 -->
00:17:47,641
Um, and so it actually helps them
511
00:17:47,641 -->
00:17:49,471
become more efficient because they
512
00:17:49,471 -->
00:17:51,691
can then, you know, read that the
513
00:17:51,691 -->
00:17:53,791
most urgent reads first, and then
514
00:17:53,791 -->
00:17:54,941
go to the, you know, more of them.
515
00:17:55,676 -->
00:17:56,336
Maintenance read.
516
00:17:56,336 -->
00:17:58,766
So, um, there's just a lot of cool
517
00:17:58,766 -->
00:18:01,016
things that we can be doing, but we need
518
00:18:01,016 -->
00:18:03,116
to make sure that we are governing it.
519
00:18:03,146 -->
00:18:04,646
So at first it comes through
520
00:18:04,646 -->
00:18:06,386
the AI Council for approval.
521
00:18:06,896 -->
00:18:08,786
There is a form that we created that
522
00:18:08,786 -->
00:18:10,736
the act, um, that the actual companies
523
00:18:10,736 -->
00:18:12,596
submit and fill out themselves.
524
00:18:12,866 -->
00:18:14,246
So it's about, you know, what
525
00:18:14,246 -->
00:18:15,686
is your data modeling look like?
526
00:18:15,806 -->
00:18:18,506
Um, What, what are you using LLM?
527
00:18:18,506 -->
00:18:19,856
Which model are you using?
528
00:18:19,856 -->
00:18:19,916
Um.
529
00:18:21,361 -->
00:18:22,291
You know, what, what are you
530
00:18:22,291 -->
00:18:24,331
monitoring for drift and bias?
531
00:18:24,421 -->
00:18:26,941
Um, where are you storing the data?
532
00:18:26,971 -->
00:18:28,381
You know, so basically some
533
00:18:28,381 -->
00:18:30,811
vendor, uh, it's like a vendor risk
534
00:18:30,811 -->
00:18:32,371
assessment, but from an AI risk.
535
00:18:32,611 -->
00:18:37,471
Um, and then we are, um, also.
536
00:18:37,906 -->
00:18:39,646
Going, uh, we've done double
537
00:18:39,646 -->
00:18:42,136
down on our AI literacy program.
538
00:18:42,466 -->
00:18:44,656
So now we have, um, we built out
539
00:18:44,656 -->
00:18:46,246
an AI literacy program and it's
540
00:18:46,246 -->
00:18:49,726
required training, um, for our
541
00:18:49,726 -->
00:18:51,766
managers on up, which is essentially
542
00:18:51,766 -->
00:18:53,446
to help us look for drift or bias
543
00:18:53,446 -->
00:18:55,786
or anything that is starting to, um,
544
00:18:56,116 -->
00:18:58,966
help us like manage these AI agents.
545
00:18:58,966 -->
00:19:00,646
You know, as you might manage
546
00:19:00,646 -->
00:19:02,176
an associate, you need to start
547
00:19:02,176 -->
00:19:04,486
managing these agents themselves too.
548
00:19:05,056 -->
00:19:05,596
Um.
549
00:19:05,936 -->
00:19:07,616
And then it also helps us understand
550
00:19:07,616 -->
00:19:08,996
like what's a permitted use, what's
551
00:19:08,996 -->
00:19:10,976
not permitted so that we don't have our
552
00:19:10,976 -->
00:19:13,556
individual team members doing something
553
00:19:13,556 -->
00:19:15,116
with AI that they shouldn't be doing.
554
00:19:16,066 -->
00:19:18,226
And then one last component of governance
555
00:19:18,226 -->
00:19:21,016
is, uh, there are a few solutions out
556
00:19:21,016 -->
00:19:23,356
there that are doing AI monitoring.
557
00:19:23,626 -->
00:19:26,686
So overlaying an another solution
558
00:19:26,686 -->
00:19:28,336
on top that is monitoring
559
00:19:28,336 -->
00:19:30,676
the AI for drift, for bias.
560
00:19:31,066 -->
00:19:32,656
And then they would escalate
561
00:19:32,656 -->
00:19:34,456
their concerns to the, you
562
00:19:34,456 -->
00:19:36,226
know, to the leadership, um,
563
00:19:36,526 -->
00:19:38,326
if, if something seems amiss.
564
00:19:38,326 -->
00:19:40,396
So, um, it's, you know, kind of like.
565
00:19:40,616 -->
00:19:42,896
You have the security, um, for
566
00:19:42,896 -->
00:19:45,896
monitoring for network, and it's, it's
567
00:19:45,896 -->
00:19:47,906
basically another security metric,
568
00:19:47,906 -->
00:19:50,246
but It's, it's monitoring for AI use.
569
00:19:51,056 -->
00:19:52,196
Krishna Gade: Yeah, makes a lot of sense.
570
00:19:52,196 -->
00:19:54,296
And, uh, very relevant to today.
571
00:19:55,461 -->
00:19:58,226
So, so I guess, you know, fundamentally,
572
00:19:58,226 -->
00:19:59,606
right, like it all comes down to
573
00:19:59,606 -->
00:20:01,406
trust and safety and governance.
574
00:20:01,436 -->
00:20:04,586
Um, what does trust in AI look from
575
00:20:04,586 -->
00:20:06,116
a hospital board's perspective?
576
00:20:06,146 -->
00:20:08,606
You know, like, because, uh, as you
577
00:20:08,606 -->
00:20:10,346
think about, uh, you know, like.
578
00:20:11,176 -->
00:20:12,646
Potentially AI recommendations
579
00:20:12,646 -->
00:20:15,616
conflicting with clinician judgment or,
580
00:20:15,616 -->
00:20:17,926
you know, AI maybe standardizing too
581
00:20:17,926 -->
00:20:21,526
much across, uh, and losing physician
582
00:20:21,526 -->
00:20:22,876
intuition and things like that.
583
00:20:22,876 -->
00:20:24,916
So, you know, how, how do you view
584
00:20:24,916 -->
00:20:26,956
of the, the world of like trust
585
00:20:26,956 -->
00:20:29,236
in AI from hospital perspective?
586
00:20:30,556 -->
00:20:32,266
Dr. Stacey Johnston: Yeah, so I, I think
587
00:20:32,266 -->
00:20:36,916
the administrators and operators, um, are.
588
00:20:37,606 -->
00:20:39,916
The leadership is most, most likely
589
00:20:39,916 -->
00:20:41,686
to embrace AI and trust it at
590
00:20:41,686 -->
00:20:43,606
this time because they've seen it
591
00:20:43,606 -->
00:20:46,036
successful in the backend processes.
592
00:20:46,366 -->
00:20:48,046
So, for instance, our revenue
593
00:20:48,046 -->
00:20:50,776
cycle team, they are on full on AI.
594
00:20:50,776 -->
00:20:52,036
They're like, the more AI
595
00:20:52,126 -->
00:20:53,896
can layer on, bring it on.
596
00:20:53,896 -->
00:20:56,266
And so, um, so they've actually
597
00:20:56,266 -->
00:20:57,316
been great to work with.
598
00:20:57,316 -->
00:20:58,366
They are actually kind of
599
00:20:58,366 -->
00:20:59,806
leading our organization with the
600
00:20:59,806 -->
00:21:01,426
amount of AI they've embraced.
601
00:21:01,811 -->
00:21:03,491
Which then made it easier to tell
602
00:21:03,491 -->
00:21:04,841
our referral team, Hey, we're gonna
603
00:21:04,841 -->
00:21:07,091
embrace AI for your referral management.
604
00:21:07,151 -->
00:21:10,001
And okay, now, now we've gotten that.
605
00:21:10,001 -->
00:21:11,801
So now let's del dive more
606
00:21:11,801 -->
00:21:12,971
into the clinical space.
607
00:21:13,451 -->
00:21:15,551
Um, the, the.
608
00:21:16,636 -->
00:21:17,926
AI that's more of an
609
00:21:17,926 -->
00:21:20,056
augmentation tool workflow.
610
00:21:20,056 -->
00:21:21,616
So as I mentioned, the in basket
611
00:21:21,616 -->
00:21:24,856
messaging, um, the clinical AI agent,
612
00:21:24,886 -->
00:21:27,016
we, we've really adopted the clinical
613
00:21:27,016 -->
00:21:29,116
AI agent, which is our ambient listening
614
00:21:29,116 -->
00:21:32,236
solution to queue up the documentation.
615
00:21:32,236 -->
00:21:36,046
So about 70 to 80% of our providers
616
00:21:36,046 -->
00:21:37,666
and the ambulatory clinics are
617
00:21:37,666 -->
00:21:39,586
using the ambient listening
618
00:21:39,586 -->
00:21:40,906
solution for their documentation.
619
00:21:41,941 -->
00:21:44,371
And about 70 to 80% of their notes are
620
00:21:44,371 -->
00:21:47,431
generated, um, with the ambient solution.
621
00:21:47,431 -->
00:21:50,731
So, um, so it's being very widely used.
622
00:21:50,791 -->
00:21:52,801
Um, part of that was we train them,
623
00:21:52,951 -->
00:21:54,781
um, we spent a lot of time training
624
00:21:54,781 -->
00:21:57,121
them on how to use it to the, to
625
00:21:57,121 -->
00:21:58,621
help them become more efficient.
626
00:21:59,041 -->
00:22:01,021
And they saw their documentation time
627
00:22:01,021 -->
00:22:02,551
go from seven and a half minutes.
628
00:22:02,566 -->
00:22:03,676
Per note to two and a
629
00:22:03,676 -->
00:22:04,636
half minutes per note.
630
00:22:04,966 -->
00:22:07,186
And so the fact that they saw the value
631
00:22:07,186 -->
00:22:10,066
in it and that they were starting to see
632
00:22:10,066 -->
00:22:11,896
that the note was actually producing a
633
00:22:11,896 -->
00:22:14,386
good quality note, so they trust that.
634
00:22:14,746 -->
00:22:16,546
Now, if I were to say, I'm gonna to
635
00:22:16,756 -->
00:22:19,186
total turn it on to be fully autonomous
636
00:22:19,186 -->
00:22:20,416
and you don't even get to read the
637
00:22:20,416 -->
00:22:21,856
note, I don't think we're there yet.
638
00:22:21,886 -->
00:22:23,416
They're not trusting it enough.
639
00:22:23,926 -->
00:22:25,516
Um, and so you.
640
00:22:25,876 -->
00:22:27,916
You know, so you've gotta move the needle
641
00:22:27,916 -->
00:22:30,466
a little bit, have them trust that one
642
00:22:30,466 -->
00:22:32,386
component and then move the needle again.
643
00:22:32,386 -->
00:22:33,766
Have them trust that component.
644
00:22:33,766 -->
00:22:36,346
And so it, it's just a natural trust
645
00:22:36,346 -->
00:22:39,346
that builds on top of each other as as
646
00:22:39,346 -->
00:22:41,236
you start to expand and expand more.
647
00:22:42,436 -->
00:22:43,996
Krishna Gade: Very interesting and,
648
00:22:43,996 -->
00:22:45,736
uh, you know, like, let's, I think
649
00:22:45,736 -->
00:22:47,386
this is a great ROI thing that,
650
00:22:47,386 -->
00:22:48,496
you know, sort of mentioned, right?
651
00:22:48,496 -->
00:22:49,936
You know, healthcare industry
652
00:22:49,936 -->
00:22:51,106
runs on thin margins.
653
00:22:51,106 -->
00:22:53,146
You know, uh, you probably have
654
00:22:53,146 -->
00:22:54,946
to justify your AI investments.
655
00:22:55,336 -->
00:22:57,196
Where have you seen like the path fastest
656
00:22:57,196 -->
00:22:59,326
payback, you know, cost reduction,
657
00:22:59,326 -->
00:23:01,546
clinician productivity, or new revenue?
658
00:23:02,701 -->
00:23:03,961
Dr. Stacey Johnston: Um, yeah,
659
00:23:03,961 -->
00:23:05,431
so I think that's still kind of
660
00:23:05,431 -->
00:23:07,351
for us on the backend pieces.
661
00:23:07,411 -->
00:23:09,421
Um, although I do have to say the
662
00:23:09,421 -->
00:23:11,311
clinical AI agent, you know, I
663
00:23:11,311 -->
00:23:12,991
remember when we first deployed it
664
00:23:12,991 -->
00:23:15,751
at another organization it was, gosh,
665
00:23:15,751 -->
00:23:17,011
I don't know if we can afford to.
666
00:23:17,721 -->
00:23:18,861
Do doctors pay for it?
667
00:23:18,861 -->
00:23:20,871
Do the, does the hospital pay for it?
668
00:23:20,871 -->
00:23:23,811
And there's no ROI, you know, maybe
669
00:23:23,811 -->
00:23:25,941
they might see one more patient per day.
670
00:23:26,151 -->
00:23:27,531
Maybe they won't, you know?
671
00:23:27,531 -->
00:23:29,841
And so, um, but what we have actually
672
00:23:29,841 -->
00:23:32,241
found is, um, we've been live for
673
00:23:32,241 -->
00:23:33,441
about a year and a half and in,
674
00:23:33,531 -->
00:23:35,061
you know, over a 12 month period of
675
00:23:35,061 -->
00:23:38,781
time, um, we had an extra $10,000.
676
00:23:38,871 -->
00:23:39,171
Um.
677
00:23:39,276 -->
00:23:42,276
Per doc revenue and just better capturing
678
00:23:42,276 -->
00:23:44,316
of the documentation of the visit.
679
00:23:44,796 -->
00:23:47,256
Um, and so there was a definitive
680
00:23:47,286 -->
00:23:49,926
ROI by going live with the AI.
681
00:23:50,256 -->
00:23:52,506
So then we're like, that's a no brainer.
682
00:23:52,506 -->
00:23:54,126
Let's just keep rolling that one out.
683
00:23:54,156 -->
00:23:56,856
So, um, so, um.
684
00:23:57,426 -->
00:23:58,121
And the other one.
685
00:23:58,296 -->
00:23:58,746
Krishna Gade: And what does
686
00:23:58,746 -->
00:24:00,036
this clinical AI agent do?
687
00:24:00,036 -->
00:24:00,336
Actually?
688
00:24:00,696 -->
00:24:01,146
Can you double?
689
00:24:01,206 -->
00:24:02,496
Dr. Stacey Johnston: It's like via bridge.
690
00:24:02,526 -->
00:24:04,476
It's within Oracle built it.
691
00:24:04,536 -->
00:24:06,366
Um, but you know, it's like the co you
692
00:24:06,366 -->
00:24:09,246
know, DAX 360, you know, the co-pilot, the
693
00:24:09,366 -->
00:24:09,546
Krishna Gade: right
694
00:24:09,636 -->
00:24:10,056
Dr. Stacey Johnston: A bridge
695
00:24:10,206 -->
00:24:12,726
that ambient, you know, it you is
696
00:24:12,726 -->
00:24:15,576
conversational AI that you have with
697
00:24:15,576 -->
00:24:17,016
your patient, the provider, and the
698
00:24:17,016 -->
00:24:18,546
patient that then queues up your
699
00:24:18,546 -->
00:24:20,376
documentation templates to review.
700
00:24:20,796 -->
00:24:21,846
Um, but that one's built
701
00:24:21,846 -->
00:24:23,676
within Oracle by Oracle.
702
00:24:23,676 -->
00:24:23,736
Um.
703
00:24:25,171 -->
00:24:28,321
And, um, the other one that
704
00:24:28,321 -->
00:24:30,631
again, received great ROI
705
00:24:30,631 -->
00:24:32,101
was the, the colonoscopy.
706
00:24:32,311 -->
00:24:35,191
So why not fully, I mean, the
707
00:24:35,191 -->
00:24:37,621
Cologuard, why not fully automate that?
708
00:24:37,621 -->
00:24:40,051
Like why can't that be a fully
709
00:24:40,051 -->
00:24:42,391
autonomous decision that if we have,
710
00:24:42,391 -->
00:24:45,181
based off of your inclusion criteria,
711
00:24:45,181 -->
00:24:47,161
and we know your exclusion criteria.
712
00:24:47,986 -->
00:24:49,576
Why isn't an a, why isn't
713
00:24:49,576 -->
00:24:51,136
an agent just automatically
714
00:24:51,136 -->
00:24:53,176
ordering the Cologuard screening?
715
00:24:53,866 -->
00:24:55,336
You know, I, I, I think
716
00:24:55,336 -->
00:24:56,806
that's a great use case.
717
00:24:56,836 -->
00:24:58,966
And then, you know, for all, for
718
00:24:58,966 -->
00:25:01,096
all of the decreased colonoscopies,
719
00:25:01,156 -->
00:25:03,916
un unnecessary colonoscopies and
720
00:25:03,916 -->
00:25:05,956
increase in appropriate colon cancer
721
00:25:05,986 -->
00:25:08,446
screening, then, um, you know, which
722
00:25:08,446 -->
00:25:10,456
is tied to a lot of the, you know,
723
00:25:10,456 -->
00:25:12,646
four and five star metrics from your.
724
00:25:13,246 -->
00:25:15,256
You know, your Medicare Advantage plans.
725
00:25:15,586 -->
00:25:17,836
Um, so you know why not
726
00:25:17,836 -->
00:25:18,886
do something like that.
727
00:25:19,996 -->
00:25:20,716
Krishna Gade: Yeah, it makes sense.
728
00:25:21,046 -->
00:25:23,026
And as you bring these multiple tools,
729
00:25:23,026 -->
00:25:24,166
right, like you mentioned, you know,
730
00:25:24,166 -->
00:25:26,926
clinical agents developed by Oracle or
731
00:25:26,926 -->
00:25:29,656
average or other, other systems, are you
732
00:25:29,656 -->
00:25:31,606
seeing like the need for a centralized
733
00:25:31,606 -->
00:25:33,586
control plane emerge where you are able
734
00:25:33,586 -->
00:25:35,686
to track all of these agents in one place
735
00:25:35,686 -->
00:25:37,786
and how know how they're performing?
736
00:25:38,506 -->
00:25:39,136
Dr. Stacey Johnston: Yeah, so
737
00:25:39,136 -->
00:25:41,686
that's, so, um, AI reports to me.
738
00:25:41,686 -->
00:25:43,666
I've, uh, formed a Department of AI
739
00:25:43,666 -->
00:25:45,436
and Transformational Technologies,
740
00:25:45,796 -->
00:25:47,476
and we are a centralized
741
00:25:47,476 -->
00:25:48,826
process at this point in time.
742
00:25:48,856 -->
00:25:52,726
However, um, you know, the recommendation
743
00:25:52,726 -->
00:25:53,866
in the studies are showing that.
744
00:25:54,136 -->
00:25:56,836
You're more likely to embrace AI if
745
00:25:56,836 -->
00:25:59,446
it's more of a federated model, um, you
746
00:25:59,446 -->
00:26:01,096
know, more of a decentralized model.
747
00:26:01,096 -->
00:26:02,296
However, I still think the
748
00:26:02,296 -->
00:26:04,006
monitoring should be centralized.
749
00:26:04,306 -->
00:26:04,396
Krishna Gade: Right.
750
00:26:04,456 -->
00:26:05,116
Dr. Stacey Johnston: And so what
751
00:26:05,116 -->
00:26:06,496
does that, you know, federated,
752
00:26:06,496 -->
00:26:08,146
decentralized model really look like?
753
00:26:08,176 -->
00:26:10,156
Well, that means that, you know,
754
00:26:10,156 -->
00:26:11,686
you've got your department's
755
00:26:11,746 -->
00:26:14,206
owning, um, you know, just kind of
756
00:26:14,206 -->
00:26:15,826
day-to-day monitoring of the AI.
757
00:26:16,131 -->
00:26:17,661
Does this still make sense?
758
00:26:17,961 -->
00:26:19,881
Am I, you know, what is my adoption rate?
759
00:26:19,881 -->
00:26:21,231
Are these metrics meeting,
760
00:26:21,261 -->
00:26:22,521
you know, my needs?
761
00:26:22,851 -->
00:26:25,191
Um, is this AI no longer needed?
762
00:26:25,191 -->
00:26:26,211
Do we turn it off?
763
00:26:26,211 -->
00:26:27,471
I mean, that's something we never
764
00:26:27,471 -->
00:26:29,061
ask ourselves in technology is
765
00:26:29,061 -->
00:26:30,591
do we turn off these systems?
766
00:26:31,071 -->
00:26:34,401
Um, and so I, I think the monitoring
767
00:26:34,431 -->
00:26:36,381
still should be a centralized process,
768
00:26:36,711 -->
00:26:39,141
but how well you adopt and deploy,
769
00:26:39,231 -->
00:26:41,751
um, the AI's agents or, you know.
770
00:26:41,901 -->
00:26:43,821
If any RPA or anything like that, I
771
00:26:43,821 -->
00:26:46,071
think we can start to move towards
772
00:26:46,071 -->
00:26:47,271
more of this federated model.
773
00:26:48,256 -->
00:26:49,306
One of the things we're talking
774
00:26:49,306 -->
00:26:50,416
about developing, and there are
775
00:26:50,416 -->
00:26:51,946
some organizations that are doing
776
00:26:51,946 -->
00:26:53,806
this really well, which is this
777
00:26:53,806 -->
00:26:55,726
kind of this AI citizen program,
778
00:26:55,726 -->
00:26:57,676
which is you can start building out
779
00:26:57,676 -->
00:26:59,626
like low code, no code yourself.
780
00:27:00,016 -->
00:27:03,046
Um, and so we have, um, you know, we
781
00:27:03,046 -->
00:27:05,686
have Microsoft Co-pilot and if you
782
00:27:05,686 -->
00:27:07,776
know, with a. You know, that higher
783
00:27:07,776 -->
00:27:11,016
level, um, um, license, you can
784
00:27:11,016 -->
00:27:12,246
actually start building out some of
785
00:27:12,246 -->
00:27:15,516
your agents and or do we use that?
786
00:27:15,516 -->
00:27:16,866
Do we use UiPath?
787
00:27:16,896 -->
00:27:17,826
You know, where you can maybe
788
00:27:17,826 -->
00:27:19,056
build out your own bots?
789
00:27:19,056 -->
00:27:22,356
And so how can we actually, um, through
790
00:27:22,356 -->
00:27:24,876
an approval process that maybe more
791
00:27:24,906 -->
00:27:27,276
centralized, then give people the
792
00:27:27,276 -->
00:27:29,436
tools to actually build their own bots.
793
00:27:29,706 -->
00:27:31,626
Um, I think that's where we wanna
794
00:27:31,626 -->
00:27:32,796
go, but we're not quite there yet.
795
00:27:33,781 -->
00:27:34,471
Krishna Gade: Yeah, makes sense.
796
00:27:34,471 -->
00:27:34,951
Makes sense.
797
00:27:35,101 -->
00:27:36,631
And then like, as part of this,
798
00:27:36,631 -->
00:27:37,861
uh, control chain of centralized
799
00:27:37,861 -->
00:27:40,111
Observability, you know, like,
800
00:27:40,171 -->
00:27:42,271
uh, is there a need for a runtime
801
00:27:42,271 -->
00:27:43,891
AI governance, like, almost like
802
00:27:43,891 -->
00:27:45,961
that redact PII before inform?
803
00:27:45,961 -->
00:27:47,131
You know, I think this is an audience
804
00:27:47,131 -->
00:27:49,141
question I'm reading out that redact the
805
00:27:49,141 -->
00:27:52,126
PHI before any information enters LLM,
806
00:27:54,266 -->
00:27:54,666
Dr. Stacey Johnston: I, yeah.
807
00:27:54,986 -->
00:27:55,186
I mean.
808
00:27:56,371 -->
00:27:58,801
Obviously I, I think you have to be
809
00:27:58,801 -->
00:28:01,021
real time because I, you know, without
810
00:28:01,021 -->
00:28:04,051
that, Um, we won't be responsive.
811
00:28:04,051 -->
00:28:06,651
So, um, It's
812
00:28:08,671 -->
00:28:10,681
It's critical for you, you know,
813
00:28:10,681 -->
00:28:11,821
you need to have kind of your
814
00:28:11,821 -->
00:28:13,681
standard processes, but there are
815
00:28:13,681 -->
00:28:15,361
gonna be times and needs where you
816
00:28:15,361 -->
00:28:17,191
need to be a little bit more agile.
817
00:28:17,431 -->
00:28:19,621
So how can you have a, you know, a
818
00:28:19,621 -->
00:28:22,141
governance that's, um, that has a
819
00:28:22,231 -->
00:28:24,511
standard process but yet is agile?
820
00:28:24,511 -->
00:28:25,951
I think that's the sweet spot.
821
00:28:25,951 -->
00:28:27,091
And I don't know that a lot of
822
00:28:27,091 -->
00:28:28,981
organizations are doing that really well.
823
00:28:28,981 -->
00:28:31,051
We're not quite in a fully agile yet.
824
00:28:31,531 -->
00:28:33,181
Um, but I think having our de
825
00:28:33,271 -->
00:28:35,251
having a dedicated AI team has
826
00:28:35,251 -->
00:28:36,601
actually made us much quicker.
827
00:28:37,306 -->
00:28:37,666
Krishna Gade: Hmm.
828
00:28:38,026 -->
00:28:38,986
Interesting, interesting.
829
00:28:39,376 -->
00:28:41,056
And then, and another related question,
830
00:28:41,056 -->
00:28:42,496
I'm just taking a couple of audience
831
00:28:42,496 -->
00:28:44,356
questions here of, you know, there are
832
00:28:44,356 -->
00:28:46,036
so many that please keep coming them.
833
00:28:46,036 -->
00:28:48,166
And, and so there's one more question
834
00:28:48,166 -->
00:28:51,316
from John Sri's on uh, hi Dr. Johnston.
835
00:28:51,316 -->
00:28:54,586
How does Beacon ensure that is robust
836
00:28:54,586 -->
00:28:56,326
in their tool evaluation while not
837
00:28:56,326 -->
00:28:58,786
acting as a bottleneck for innovation?
838
00:28:58,786 -->
00:29:00,706
I think this is kind of that trade
839
00:29:00,706 -->
00:29:02,356
off between innovation and governance.
840
00:29:02,941 -->
00:29:03,571
Dr. Stacey Johnston: Yeah, no,
841
00:29:04,021 -->
00:29:05,341
that's, that's always hard.
842
00:29:05,461 -->
00:29:09,571
Um, I, I think the organization is
843
00:29:09,571 -->
00:29:12,001
definitely clamoring for more AI, but
844
00:29:12,031 -->
00:29:14,041
you know, I brought up in our large
845
00:29:14,041 -->
00:29:15,661
system meeting this morning that,
846
00:29:16,051 -->
00:29:17,941
hey, we can go as fast as you want.
847
00:29:17,971 -->
00:29:19,831
I just need more people to do the work.
848
00:29:19,891 -->
00:29:23,311
You, you know, so, um, so is, you know,
849
00:29:23,311 -->
00:29:25,171
are these tools worth their investment?
850
00:29:25,171 -->
00:29:25,951
I think so.
851
00:29:26,871 -->
00:29:28,851
You know, if we work on these AI
852
00:29:28,851 -->
00:29:30,741
solutions, what are we not working on?
853
00:29:30,741 -->
00:29:32,571
So I think that's kind of the,
854
00:29:32,571 -->
00:29:34,221
you know, so we need, we need to
855
00:29:34,221 -->
00:29:35,901
understand the system priorities.
856
00:29:36,351 -->
00:29:39,351
Um, but also as we move into this more,
857
00:29:39,411 -->
00:29:42,561
um, federated model where if it's kind
858
00:29:42,561 -->
00:29:45,741
of this low code, no code, um, or, you
859
00:29:45,741 -->
00:29:48,021
know, a, a simple bot build program,
860
00:29:48,021 -->
00:29:49,881
then that I think helps us become more.
861
00:29:50,471 -->
00:29:52,031
Innovative and that people could
862
00:29:52,031 -->
00:29:53,441
actually be building some of their
863
00:29:53,441 -->
00:29:55,391
own agents or building their own bots.
864
00:29:55,391 -->
00:29:55,451
Um.
865
00:29:56,866 -->
00:29:59,296
I, that's I think where most organizations
866
00:29:59,296 -->
00:30:00,796
are ultimately gonna have to go.
867
00:30:00,796 -->
00:30:01,786
And the other thing is that,
868
00:30:01,846 -->
00:30:03,526
um, do you buy or do you build?
869
00:30:03,916 -->
00:30:05,416
And we're doing a hybrid of both.
870
00:30:05,416 -->
00:30:08,056
So if it's a really complex workflow or if
871
00:30:08,056 -->
00:30:10,006
we need it turned around really quickly,
872
00:30:10,276 -->
00:30:11,656
we're gonna buy, we're gonna partner
873
00:30:11,656 -->
00:30:13,426
with someone and we're gonna say, Hey,
874
00:30:13,426 -->
00:30:15,286
we need this agent built in two weeks.
875
00:30:15,676 -->
00:30:17,326
Um, can you, you know, we have it
876
00:30:17,326 -->
00:30:19,276
mapped out, you know, you can you
877
00:30:19,276 -->
00:30:22,366
build that for us versus if you, um.
878
00:30:22,996 -->
00:30:25,036
You know, if it's a, uh, a simple work
879
00:30:25,336 -->
00:30:27,946
workflow, um, you know, why don't we
880
00:30:27,946 -->
00:30:29,866
build that ourselves within our team?
881
00:30:30,136 -->
00:30:32,026
Um, so, you know, I think, I think
882
00:30:32,026 -->
00:30:34,066
it's a, you're gonna almost have
883
00:30:34,066 -->
00:30:35,926
to do all you, you're gonna have to
884
00:30:35,926 -->
00:30:40,006
be innovative and have a process.
885
00:30:40,036 -->
00:30:41,806
You're gonna need to be agile.
886
00:30:42,271 -->
00:30:44,131
And also look at cost reduction.
887
00:30:44,131 -->
00:30:45,871
I mean, you, you have to, you have
888
00:30:45,871 -->
00:30:48,481
to kind of do all, and so, you know
889
00:30:48,481 -->
00:30:50,041
what I brought up this morning was
890
00:30:50,131 -->
00:30:52,231
I think, you know, yes, we might be.
891
00:30:52,981 -->
00:30:54,991
Spending more now, but in the long run,
892
00:30:54,991 -->
00:30:57,061
it's gonna save us more money by hope, you
893
00:30:57,061 -->
00:31:00,361
know, reducing FTEs, or honestly, where
894
00:31:00,361 -->
00:31:02,341
we are is we can't even fill those FTEs.
895
00:31:02,341 -->
00:31:03,781
So we're paying contractors.
896
00:31:04,261 -->
00:31:06,601
And so rather, instead of filling
897
00:31:06,601 -->
00:31:08,701
it with contractors, can we fill
898
00:31:08,701 -->
00:31:11,011
that with an, an AI agent that can
899
00:31:11,011 -->
00:31:12,536
actually help reduce the workloads?
900
00:31:12,751 -->
00:31:14,101
You don't need as many FTEs.
901
00:31:14,131 -->
00:31:16,441
So for even it's not for us,
902
00:31:16,441 -->
00:31:17,491
it's not about reducing your
903
00:31:17,491 -->
00:31:18,901
FTEs, it's just about, you know.
904
00:31:19,531 -->
00:31:21,181
Filling the positions we have open
905
00:31:21,181 -->
00:31:22,861
or reducing our contractor spend.
906
00:31:23,581 -->
00:31:23,941
Krishna Gade: No, that's
907
00:31:23,941 -->
00:31:24,811
very, very interesting.
908
00:31:25,321 -->
00:31:27,751
And, and so, uh, I guess like for other,
909
00:31:27,991 -->
00:31:30,001
you know, I AI leaders like you working
910
00:31:30,001 -->
00:31:31,951
in healthcare, what's your advice?
911
00:31:31,951 -->
00:31:33,091
Like, what do they, how do they
912
00:31:33,091 -->
00:31:35,401
think about, uh, you know, bringing
913
00:31:35,401 -->
00:31:37,501
new AI tools to deployment?
914
00:31:37,531 -->
00:31:39,121
What is the one thing to put in
915
00:31:39,121 -->
00:31:41,581
place, you know, before they deploy
916
00:31:41,581 -->
00:31:43,141
anything in terms of AI, you know,
917
00:31:43,141 -->
00:31:44,096
in their, in their environment?
918
00:31:45,256 -->
00:31:45,856
Dr. Stacey Johnston: I do think
919
00:31:45,856 -->
00:31:48,016
having an AI governance council
920
00:31:48,016 -->
00:31:49,696
and advisory council makes sense.
921
00:31:49,696 -->
00:31:52,156
And so, um, we have physicians sitting on
922
00:31:52,156 -->
00:31:54,856
it, vice presidents, hospital presidents.
923
00:31:55,246 -->
00:31:57,316
Um, we have ethics, we have compliance,
924
00:31:57,316 -->
00:32:00,656
we have risk, we have security, um, I ask
925
00:32:00,806 -->
00:32:03,706
leaders um, so, you know, it's kind of the
926
00:32:03,736 -->
00:32:08,206
gamut of, you know, what's gonna Have HR.
927
00:32:08,566 -->
00:32:09,916
Um, so, you know, we have the, kind
928
00:32:09,916 -->
00:32:11,656
of the gamut of who needs AI and who
929
00:32:11,656 -->
00:32:13,816
should be responsible for AI, and that
930
00:32:13,816 -->
00:32:16,096
just helps to ensure that we have some
931
00:32:16,096 -->
00:32:19,156
responsibility, um, for our solutions.
932
00:32:19,516 -->
00:32:21,556
Um, but it, you know, also helps
933
00:32:21,556 -->
00:32:22,816
us understand our training and
934
00:32:22,816 -->
00:32:24,526
needs, our change management needs.
935
00:32:24,946 -->
00:32:25,456
Um.
936
00:32:26,161 -->
00:32:27,691
So I think that's the first thing.
937
00:32:27,691 -->
00:32:29,431
And then you need to understand, you
938
00:32:29,431 -->
00:32:31,411
know, the other thing we did very early
939
00:32:31,411 -->
00:32:33,601
on was at a minimum, understand what
940
00:32:33,601 -->
00:32:35,731
your permitted and prohibited uses are.
941
00:32:35,761 -->
00:32:37,861
Are you going to be, you know,
942
00:32:37,921 -->
00:32:39,091
what is your risk for fully
943
00:32:39,091 -->
00:32:40,621
autonomous in the clinical space?
944
00:32:40,621 -->
00:32:41,371
Are you there yet?
945
00:32:41,371 -->
00:32:41,941
Are you not?
946
00:32:42,001 -->
00:32:42,811
You know, you need to
947
00:32:42,811 -->
00:32:44,461
decide as an organization.
948
00:32:44,941 -->
00:32:48,271
Um, and then I would, you know,
949
00:32:48,271 -->
00:32:49,861
so those are the first two things
950
00:32:49,861 -->
00:32:51,031
I think you need to do first
951
00:32:51,031 -->
00:32:52,681
before you really fully expand.
952
00:32:53,131 -->
00:32:54,001
And then I would.
953
00:32:54,136 -->
00:32:57,196
A hundred percent what I did of standing
954
00:32:57,196 -->
00:32:59,356
up a department that this is all I did.
955
00:32:59,686 -->
00:33:02,686
Because if you add AI on top of your team
956
00:33:02,686 -->
00:33:04,846
that's already busy with code upgrades,
957
00:33:05,086 -->
00:33:07,426
maintaining your systems, the break fix,
958
00:33:07,486 -->
00:33:10,006
you know, the, you know, the request from,
959
00:33:10,036 -->
00:33:11,986
you know, I need to change this order set,
960
00:33:12,046 -->
00:33:14,206
um, this documentation template, you know,
961
00:33:14,206 -->
00:33:16,276
we have now joint commissions coming.
962
00:33:16,276 -->
00:33:17,476
So if, if.
963
00:33:17,881 -->
00:33:19,501
If you're a team that's doing that
964
00:33:19,501 -->
00:33:21,991
day-to-day maintenance and you know
965
00:33:22,051 -->
00:33:23,881
of your systems, and then they're
966
00:33:23,881 -->
00:33:25,711
also responsible for going out
967
00:33:25,711 -->
00:33:28,021
there and, and being a partner for
968
00:33:28,081 -->
00:33:30,571
implementing AI, I think it's gonna
969
00:33:30,571 -->
00:33:32,431
be very hard to move the needle.
970
00:33:32,671 -->
00:33:34,681
So I think having a team that they don't
971
00:33:34,681 -->
00:33:36,071
need to worry about that the mundane.
972
00:33:36,071 -->
00:33:37,951
Yeah, their job is just to go and
973
00:33:37,951 -->
00:33:40,051
meet with the business partners, um,
974
00:33:40,081 -->
00:33:42,541
go, they're going to conferences.
975
00:33:42,571 -->
00:33:43,891
They are, you know, meeting
976
00:33:43,891 -->
00:33:45,571
with other organizations.
977
00:33:45,841 -->
00:33:47,491
They are out there just scanning the
978
00:33:47,491 -->
00:33:49,711
environment and finding out what's new.
979
00:33:49,741 -->
00:33:51,481
They're meeting with venture capitalists.
980
00:33:51,871 -->
00:33:54,001
Um, we're meeting with, um, you know,
981
00:33:54,001 -->
00:33:55,471
we're partners with Notre Dame, so
982
00:33:55,471 -->
00:33:57,241
we meet with Notre Dame frequently.
983
00:33:57,241 -->
00:33:58,261
What are they working on?
984
00:33:58,261 -->
00:33:59,341
What are they seeing?
985
00:33:59,791 -->
00:34:00,301
Um.
986
00:34:00,636 -->
00:34:02,646
You know, try and we are setting up,
987
00:34:02,676 -->
00:34:05,256
um, some internships with Notre Dame so
988
00:34:05,256 -->
00:34:07,296
they can actually do some development
989
00:34:07,356 -->
00:34:09,276
in our, in our environment too.
990
00:34:09,276 -->
00:34:11,586
So before you get there though, you,
991
00:34:11,586 -->
00:34:14,046
you know, it's kind of that, um, you
992
00:34:14,046 -->
00:34:15,996
know, crawl, walk, run philosophy.
993
00:34:15,996 -->
00:34:16,056
Yeah.
994
00:34:16,086 -->
00:34:17,676
So you've gotta crawl first.
995
00:34:17,676 -->
00:34:19,836
So choose a couple of good select
996
00:34:19,836 -->
00:34:21,636
use cases that you felt fairly
997
00:34:21,636 -->
00:34:23,976
confident in, um, before you start
998
00:34:23,976 -->
00:34:25,926
to, you know, um, walk and run.
999
00:34:26,196 -->
00:34:27,726
The other thing that we do, I
1000
00:34:27,726 -->
00:34:29,886
forgot to mention with the AI, um.
1001
00:34:30,496 -->
00:34:33,496
Policy is we only approve at the
1002
00:34:33,496 -->
00:34:37,276
advisory council if it has a defined ROI.
1003
00:34:37,996 -->
00:34:41,086
If there is no ROI, then you need
1004
00:34:41,086 -->
00:34:42,826
to ask yourself why are we doing it?
1005
00:34:43,246 -->
00:34:45,076
But if it's soft, ROI, and it's
1006
00:34:45,076 -->
00:34:46,526
really gonna give back, you know?
1007
00:34:47,011 -->
00:34:49,351
Back to the clinical experience, then it
1008
00:34:49,351 -->
00:34:51,181
goes to the steering committee, meaning
1009
00:34:51,181 -->
00:34:54,001
that, you know, do, do the executives,
1010
00:34:54,001 -->
00:34:56,731
so the C-suite do we buy in that we wanna
1011
00:34:56,731 -->
00:34:59,341
give this to the clinical end users, even
1012
00:34:59,341 -->
00:35:02,281
though we may not get any ROI from it.
1013
00:35:02,311 -->
00:35:04,471
And so, um, we haven't actually
1014
00:35:04,471 -->
00:35:06,781
had anything that we've approved,
1015
00:35:06,871 -->
00:35:08,761
um, without a defined ROI.
1016
00:35:08,761 -->
00:35:10,501
So right now, you know.
1017
00:35:11,161 -->
00:35:12,871
Finances are tight for all of us.
1018
00:35:12,871 -->
00:35:15,271
So you, you all should be looking for
1019
00:35:15,271 -->
00:35:17,881
ROI first with your AI before really
1020
00:35:17,881 -->
00:35:19,861
going to the non ROI initiatives.
1021
00:35:20,401 -->
00:35:21,061
Krishna Gade: Yeah, makes sense.
1022
00:35:21,241 -->
00:35:22,261
So you got to measure your
1023
00:35:22,261 -->
00:35:24,121
ROI continuously and evaluate
1024
00:35:24,121 -->
00:35:25,351
how these things are working.
1025
00:35:25,351 -->
00:35:26,431
I guess that's important.
1026
00:35:27,001 -->
00:35:27,301
Uh,
1027
00:35:27,361 -->
00:35:27,601
Dr. Stacey Johnston: yeah.
1028
00:35:28,051 -->
00:35:28,171
Krishna Gade: So
1029
00:35:32,566 -->
00:35:35,556
there is, um, winning the confidence
1030
00:35:35,611 -->
00:35:37,141
of clinical, clinical staff, right,
1031
00:35:37,141 -->
00:35:38,251
when it comes to introduction of
1032
00:35:38,251 -->
00:35:39,811
new tools, especially AI, right?
1033
00:35:40,216 -->
00:35:43,066
What has been your journey, uh, in
1034
00:35:43,066 -->
00:35:45,286
terms of, uh, doing that at Beacon?
1035
00:35:45,286 -->
00:35:47,236
You know, how you successful
1036
00:35:47,236 -->
00:35:48,436
or any lessons learned?
1037
00:35:48,706 -->
00:35:48,796
And
1038
00:35:49,696 -->
00:35:49,996
Dr. Stacey Johnston: I'm sorry,
1039
00:35:49,996 -->
00:35:51,316
it broke up for just a second.
1040
00:35:51,316 -->
00:35:52,031
Could you ask the question
1041
00:35:52,031 -->
00:35:52,191
Krishna Gade: again?
1042
00:35:52,191 -->
00:35:52,391
Yeah.
1043
00:35:52,391 -->
00:35:53,836
So the question was basically how are
1044
00:35:53,836 -->
00:35:55,486
you, uh, you know, one of the problems
1045
00:35:55,486 -->
00:35:57,196
in, in healthcare is like, how do you
1046
00:35:57,196 -->
00:35:59,386
win the confidence of AI clinical staff
1047
00:35:59,476 -->
00:36:01,696
and who might be skeptical about AI
1048
00:36:01,696 -->
00:36:03,991
and the tools and, and, you know, yeah.
1049
00:36:04,636 -->
00:36:05,806
And what has been your
1050
00:36:05,806 -->
00:36:07,126
experience doing that at Beacon?
1051
00:36:07,131 -->
00:36:07,401
Yeah.
1052
00:36:07,606 -->
00:36:07,966
Dr. Stacey Johnston: Yeah.
1053
00:36:07,966 -->
00:36:10,816
So I think having, um, key
1054
00:36:10,816 -->
00:36:12,376
physician partners, um,
1055
00:36:12,406 -->
00:36:14,026
understanding their use cases.
1056
00:36:14,026 -->
00:36:15,646
So I actually spoke at a physician
1057
00:36:15,646 -->
00:36:18,016
meeting last night and kind of talked
1058
00:36:18,016 -->
00:36:19,486
about the future of healthcare and
1059
00:36:19,606 -->
00:36:21,196
the future of healthcare technology.
1060
00:36:21,226 -->
00:36:22,846
Obviously AI was probably a number
1061
00:36:22,846 -->
00:36:24,706
one thing we talked about and.
1062
00:36:24,941 -->
00:36:26,951
People were, you know, some were
1063
00:36:27,011 -->
00:36:28,361
reticent, but some were like,
1064
00:36:28,391 -->
00:36:30,311
Hey, I have really great ideas.
1065
00:36:30,311 -->
00:36:31,391
I'm like, submit 'em to me.
1066
00:36:31,391 -->
00:36:34,241
You know, if it's a low hanging fruit, we
1067
00:36:34,241 -->
00:36:36,761
could put that in, you know, from the time
1068
00:36:36,761 -->
00:36:39,851
of request to an agent being built is, you
1069
00:36:39,851 -->
00:36:42,221
know, probably about, you know, four to.
1070
00:36:42,901 -->
00:36:44,341
Four to eight weeks,
1071
00:36:44,341 -->
00:36:45,601
depending on the complexity.
1072
00:36:45,961 -->
00:36:46,351
Right?
1073
00:36:46,351 -->
00:36:48,511
And so why not, you know, if that,
1074
00:36:48,511 -->
00:36:49,891
if there's gonna be ROI there, I
1075
00:36:49,891 -->
00:36:51,241
don't see why we couldn't do that.
1076
00:36:51,631 -->
00:36:52,711
Um, if it's obviously more
1077
00:36:52,711 -->
00:36:54,451
complicated workflow or it requires
1078
00:36:54,451 -->
00:36:56,251
integration into our EHR, that's
1079
00:36:56,251 -->
00:36:58,051
gonna be a longer, you know, ask.
1080
00:36:58,561 -->
00:36:58,951
Um.
1081
00:36:59,231 -->
00:37:02,141
But the, but you need to deliver
1082
00:37:02,141 -->
00:37:03,881
on your promises, and then they
1083
00:37:03,881 -->
00:37:06,371
need to see the value of what they,
1084
00:37:06,491 -->
00:37:08,021
you know, what we've turned on.
1085
00:37:08,501 -->
00:37:09,821
So, you know, one of the things I
1086
00:37:09,821 -->
00:37:11,441
always say is don't implement AI for
1087
00:37:11,441 -->
00:37:13,121
AI sake to say you're implementing
1088
00:37:13,121 -->
00:37:15,311
AI, but it's gonna add, it's
1089
00:37:15,311 -->
00:37:17,351
gonna be meaningless to the docs.
1090
00:37:17,381 -->
00:37:18,731
Or make sure you know where
1091
00:37:18,731 -->
00:37:20,081
that data's pulling from.
1092
00:37:20,081 -->
00:37:21,616
So again, if you have.
1093
00:37:22,996 -->
00:37:24,676
Different dataset, definitions, and
1094
00:37:24,676 -->
00:37:26,116
you're pulling from the wrong dataset.
1095
00:37:26,116 -->
00:37:26,956
It's the data.
1096
00:37:26,956 -->
00:37:28,306
You know, the docs aren't gonna
1097
00:37:28,306 -->
00:37:29,476
trust the data and therefore
1098
00:37:29,476 -->
00:37:30,886
they're not gonna use the systems.
1099
00:37:31,396 -->
00:37:33,736
Um, and it needs to be easy to use,
1100
00:37:33,766 -->
00:37:35,716
you know, just naturally intuitive.
1101
00:37:35,716 -->
00:37:37,306
It needs to be part of their workflow.
1102
00:37:37,366 -->
00:37:39,346
Asking them to go out of their
1103
00:37:39,346 -->
00:37:41,986
EHR to use another system that's
1104
00:37:42,136 -->
00:37:44,476
gonna also not help your case.
1105
00:37:44,476 -->
00:37:46,546
And so it needs to be seamless,
1106
00:37:46,576 -->
00:37:47,836
you know, integrating into your
1107
00:37:47,836 -->
00:37:49,516
system as much as possible.
1108
00:37:49,846 -->
00:37:50,056
Um.
1109
00:37:50,791 -->
00:37:52,171
And, and it just needs to
1110
00:37:52,171 -->
00:37:53,911
provide some value to them.
1111
00:37:53,941 -->
00:37:56,761
Either, you know, um, they can see
1112
00:37:56,761 -->
00:37:59,041
more patients or they go home earlier.
1113
00:37:59,041 -->
00:38:01,831
They, um, less cognitive burden,
1114
00:38:01,891 -->
00:38:03,871
you know, one less click, you know,
1115
00:38:03,991 -->
00:38:05,911
throughout a day adds up, you know,
1116
00:38:06,271 -->
00:38:08,401
a significant, um, time savings.
1117
00:38:08,401 -->
00:38:10,951
So, you know, the, you're gonna
1118
00:38:10,951 -->
00:38:13,021
gain the trust by doing solutions
1119
00:38:13,021 -->
00:38:14,036
that are easy to use for them.
1120
00:38:15,271 -->
00:38:15,631
Krishna Gade: Awesome.
1121
00:38:15,631 -->
00:38:17,371
Yeah, and I think once you're able
1122
00:38:17,371 -->
00:38:19,111
to trust your AI, then you can start
1123
00:38:19,111 -->
00:38:21,391
thinking about customer facing AI as well.
1124
00:38:21,391 -->
00:38:21,631
Right.
1125
00:38:22,021 -->
00:38:23,911
That's actually an interesting, uh,
1126
00:38:23,971 -->
00:38:25,591
you know, question from an audience,
1127
00:38:27,661 -->
00:38:29,671
you know, how is there, how
1128
00:38:29,671 -->
00:38:30,546
are the con, you know, how
1129
00:38:30,546 -->
00:38:31,801
are you tracking the concerns?
1130
00:38:31,801 -->
00:38:32,671
You know, how are you sort of
1131
00:38:32,701 -->
00:38:35,521
overcoming the concerns about AI
1132
00:38:35,521 -->
00:38:37,771
agents directly work, you know.
1133
00:38:38,176 -->
00:38:40,666
Working with patient and, and sort of,
1134
00:38:40,666 -->
00:38:42,796
uh, and potential patient reactions,
1135
00:38:42,796 -->
00:38:44,326
you know, dealing with AI agents,
1136
00:38:44,326 -->
00:38:46,246
you know, what has been, you know, I
1137
00:38:46,246 -->
00:38:48,016
don't know if you are there yet, or
1138
00:38:48,016 -->
00:38:49,111
like, how are you thinking about it?
1139
00:38:49,951 -->
00:38:49,951
Dr. Stacey Johnston:
1140
00:38:49,951 -->
00:38:51,121
We're, we're looking at it.
1141
00:38:51,121 -->
00:38:53,581
I mean, we are investigating the AI
1142
00:38:53,581 -->
00:38:56,101
agents making phone calls and so, you
1143
00:38:56,101 -->
00:38:59,881
know, studies have shown, um, that that
1144
00:39:00,061 -->
00:39:03,691
empathy from an AI agent call is actually
1145
00:39:03,691 -->
00:39:06,481
higher than, you know, a person calling.
1146
00:39:06,931 -->
00:39:10,146
And, um, you know, so, so that's probably
1147
00:39:10,146 -->
00:39:12,601
where we're gonna start with our AI,
1148
00:39:12,961 -->
00:39:15,361
true interaction with the patient.
1149
00:39:16,321 -->
00:39:16,411
Um.
1150
00:39:17,011 -->
00:39:19,411
Now there's other things that AI
1151
00:39:19,411 -->
00:39:21,001
could be doing and that I'd like to
1152
00:39:21,001 -->
00:39:23,671
pursue, which is really AI detecting
1153
00:39:24,001 -->
00:39:26,791
kind of emotional or mental crises.
1154
00:39:26,791 -->
00:39:28,171
You know, when they're calling
1155
00:39:28,171 -->
00:39:29,521
to make an appointment, there's
1156
00:39:29,611 -->
00:39:31,261
AI solutions out there that can
1157
00:39:31,351 -->
00:39:33,241
detect, you know, suicidal risks.
1158
00:39:33,271 -->
00:39:35,701
So is that something that you, you
1159
00:39:35,701 -->
00:39:37,291
know, wanna look at as well too?
1160
00:39:37,681 -->
00:39:40,291
Um, but I, I think, you know, I think.
1161
00:39:40,581 -->
00:39:42,681
AI agents calling patients
1162
00:39:42,681 -->
00:39:43,881
is pretty cutting edge.
1163
00:39:43,881 -->
00:39:43,941
Hmm.
1164
00:39:44,241 -->
00:39:45,531
Um, you know, the other
1165
00:39:45,531 -->
00:39:47,061
thing that, you know, AI.
1166
00:39:48,001 -->
00:39:49,531
Um, can be doing, there's a lot of
1167
00:39:49,531 -->
00:39:52,441
solutions out there that, um, for in your
1168
00:39:52,441 -->
00:39:55,081
inpatient stays, you know, AI and falls.
1169
00:39:55,081 -->
00:39:56,911
A lot, a lot of people are doing AI with
1170
00:39:56,911 -->
00:39:59,191
fall predictive modeling with camera use.
1171
00:39:59,551 -->
00:40:01,231
But what about if you could take that
1172
00:40:01,231 -->
00:40:03,931
and then that, um, if the, um, you know,
1173
00:40:03,931 -->
00:40:06,456
the patient is about to um, stand up,
1174
00:40:06,456 -->
00:40:08,646
the agent itself calls into the room
1175
00:40:08,646 -->
00:40:10,536
and says, Hey, please sit back down.
1176
00:40:10,926 -->
00:40:12,276
Um, or do you need me to
1177
00:40:12,276 -->
00:40:13,596
call your nurse for you?
1178
00:40:13,626 -->
00:40:13,896
You know?
1179
00:40:13,896 -->
00:40:16,086
So having that agent interact with the
1180
00:40:16,086 -->
00:40:18,456
patient actually in the, the clinical
1181
00:40:18,456 -->
00:40:21,336
setting in the care room, um, that's
1182
00:40:21,336 -->
00:40:23,286
out there, I mean that's, we haven't
1183
00:40:23,556 -->
00:40:24,996
done, we haven't done that yet, but
1184
00:40:24,996 -->
00:40:26,226
that's definitely on our roadmap.
1185
00:40:27,006 -->
00:40:27,726
Krishna Gade: Very interesting.
1186
00:40:28,536 -->
00:40:29,136
Very interesting.
1187
00:40:29,196 -->
00:40:31,056
Uh, I guess, uh, you know,
1188
00:40:31,416 -->
00:40:33,126
coming back to sort of, uh.
1189
00:40:33,781 -->
00:40:36,331
You know, copilots and autopilots in
1190
00:40:36,331 -->
00:40:38,611
medicine, you know what, what, you know,
1191
00:40:38,616 -->
00:40:40,771
apart, apart from like administrative
1192
00:40:40,771 -->
00:40:42,241
assistance and clinical diagnosis,
1193
00:40:42,241 -->
00:40:43,561
where do you see the world going?
1194
00:40:43,561 -->
00:40:44,851
You know, for example, if you have to
1195
00:40:45,451 -->
00:40:47,671
rebuild, you know, the entire healthcare
1196
00:40:47,671 -->
00:40:50,311
system from scratch today, you know,
1197
00:40:50,311 -->
00:40:51,961
with AI, what would you do differently?
1198
00:40:53,011 -->
00:40:53,401
Dr. Stacey Johnston: Uh,
1199
00:40:53,551 -->
00:40:55,291
yeah, so it's interesting.
1200
00:40:55,381 -->
00:40:59,941
Um, so Oracle is rebuilding their or
1201
00:41:00,211 -->
00:41:03,241
total EHR with this new AI semantic
1202
00:41:03,241 -->
00:41:06,121
layer and we are going live with it
1203
00:41:06,121 -->
00:41:08,221
in our ambulatory clinics this summer.
1204
00:41:08,221 -->
00:41:10,441
So at least in pilot, you know, a pilot,
1205
00:41:10,441 -->
00:41:13,741
a few clinics, um, so it'll be interesting
1206
00:41:13,741 -->
00:41:16,241
to see if that significantly changes.
1207
00:41:16,251 -->
00:41:18,291
But some of the AI things that we
1208
00:41:18,291 -->
00:41:20,331
need to embrace within the EHR.
1209
00:41:20,811 -->
00:41:23,001
Um, again, the augmented responses
1210
00:41:23,001 -->
00:41:25,011
to the patient messages, number one.
1211
00:41:25,041 -->
00:41:27,261
Number two, um, summarizer.
1212
00:41:27,261 -->
00:41:29,211
So kind of what happened between,
1213
00:41:29,271 -->
00:41:31,161
from the time I saw you, you
1214
00:41:31,161 -->
00:41:32,971
know, six months ago to now.
1215
00:41:33,491 -->
00:41:34,891
You know, was there admission?
1216
00:41:34,946 -->
00:41:36,446
What, you know, what um,
1217
00:41:36,506 -->
00:41:37,706
what images were done?
1218
00:41:37,706 -->
00:41:38,906
Did they follow up on
1219
00:41:38,906 -->
00:41:39,986
any of their care gaps?
1220
00:41:39,986 -->
00:41:41,846
What care gaps still need to be done?
1221
00:41:42,266 -->
00:41:43,946
You know, so basically kind of
1222
00:41:43,946 -->
00:41:45,566
what happened to now and how
1223
00:41:45,566 -->
00:41:47,336
can I help keep you healthy?
1224
00:41:47,666 -->
00:41:50,096
Like layering the EHR on top of
1225
00:41:50,096 -->
00:41:52,346
large language model, um, to be
1226
00:41:52,346 -->
00:41:54,086
able to queue that up combined
1227
00:41:54,086 -->
00:41:55,616
with the ambient solutioning.
1228
00:41:55,916 -->
00:41:58,376
So then you can review the, you
1229
00:41:58,376 -->
00:41:59,786
know, the note summarizer and say,
1230
00:41:59,816 -->
00:42:01,316
okay, I see that you, you know,
1231
00:42:01,316 -->
00:42:03,916
visited that Ad for chest pain.
1232
00:42:03,916 -->
00:42:05,236
Did you follow up with a cardiologist?
1233
00:42:05,236 -->
00:42:05,716
No, I didn't.
1234
00:42:05,716 -->
00:42:06,946
Well, let me place this
1235
00:42:06,946 -->
00:42:08,236
referral for you now.
1236
00:42:08,236 -->
00:42:09,316
You know, let's get you referred.
1237
00:42:09,316 -->
00:42:10,096
Do you have a particular
1238
00:42:10,096 -->
00:42:11,296
cardiologist you want?
1239
00:42:11,686 -->
00:42:13,336
Let's go ahead and order an echo
1240
00:42:13,366 -->
00:42:16,036
and, Um, you know, follow up with
1241
00:42:16,036 -->
00:42:17,896
me in a month that conversation.
1242
00:42:18,231 -->
00:42:19,701
Should be actually placing the
1243
00:42:19,701 -->
00:42:21,951
orders as you're speaking instead
1244
00:42:21,951 -->
00:42:23,661
of you then getting the document.
1245
00:42:23,661 -->
00:42:24,681
Right now it's just queuing
1246
00:42:24,681 -->
00:42:25,701
it up in the document.
1247
00:42:25,701 -->
00:42:26,571
You read the document, you
1248
00:42:26,571 -->
00:42:28,071
submit, it still saves you time.
1249
00:42:28,371 -->
00:42:29,661
But wouldn't it be great if then it
1250
00:42:29,661 -->
00:42:31,551
also actually placed those orders,
1251
00:42:31,551 -->
00:42:33,231
or at a minimum, queued them up into
1252
00:42:33,231 -->
00:42:35,211
your in basket so that you just had to
1253
00:42:35,211 -->
00:42:37,371
sign off on them instead of actually
1254
00:42:37,371 -->
00:42:39,021
having to manually type in the orders.
1255
00:42:39,021 -->
00:42:41,061
And because it was using a large language
1256
00:42:41,061 -->
00:42:42,921
model, it could review to make sure
1257
00:42:42,921 -->
00:42:44,301
you have the right CPT code and the
1258
00:42:44,301 -->
00:42:47,601
right diagnosis code, and you know, um.
1259
00:42:47,876 -->
00:42:49,856
And it also has access to your
1260
00:42:49,946 -->
00:42:51,596
referral management system, so
1261
00:42:51,596 -->
00:42:53,036
it knows which doctor to send
1262
00:42:53,036 -->
00:42:54,356
it to and the right address.
1263
00:42:54,356 -->
00:42:56,996
And then the patient gets a text, um,
1264
00:42:57,056 -->
00:42:58,616
you know, 30 minutes later and say,
1265
00:42:58,616 -->
00:43:00,416
Hey, we got your referral processed.
1266
00:43:00,746 -->
00:43:02,606
Um, you know, please, uh, you
1267
00:43:02,606 -->
00:43:04,436
know, we'll text you again when we
1268
00:43:04,436 -->
00:43:05,636
have an appointment made for you.
1269
00:43:05,666 -->
00:43:07,886
So, um, or text in here.
1270
00:43:07,886 -->
00:43:09,566
You know, use your texting solution
1271
00:43:09,566 -->
00:43:11,156
to choose which time you want.
1272
00:43:11,156 -->
00:43:13,766
So that's where we need to go.
1273
00:43:13,766 -->
00:43:15,026
We're not there yet, but I
1274
00:43:15,026 -->
00:43:17,426
definitely, um, think there are some.
1275
00:43:17,551 -->
00:43:19,231
Organizations that are doing bits
1276
00:43:19,231 -->
00:43:20,581
and pieces, like we're doing the
1277
00:43:20,581 -->
00:43:23,281
texting and the online scheduling.
1278
00:43:23,431 -->
00:43:25,021
Um, but we are not fully,
1279
00:43:25,711 -->
00:43:27,151
Um, the note summarizer.
1280
00:43:27,151 -->
00:43:28,711
We won't go live with till this summer.
1281
00:43:29,461 -->
00:43:30,001
Krishna Gade: Interesting.
1282
00:43:30,121 -->
00:43:30,571
Interesting.
1283
00:43:31,111 -->
00:43:32,521
And there was also a question on
1284
00:43:32,521 -->
00:43:34,561
variables, uh, from the audience, uh,
1285
00:43:35,191 -->
00:43:38,671
you know, from, um, Dr. Johnston, how
1286
00:43:38,671 -->
00:43:41,071
are you envisioning utilizing lifestyle
1287
00:43:41,071 -->
00:43:43,021
variable data from patients in their care?
1288
00:43:43,021 -->
00:43:44,311
Variable devices like smart
1289
00:43:44,311 -->
00:43:45,481
watches, rings, wristbands,
1290
00:43:45,481 -->
00:43:48,181
continuously collect data, and Yeah.
1291
00:43:48,601 -->
00:43:49,831
And then how can these data sets
1292
00:43:49,831 -->
00:43:51,841
be utilized by leveraging AI
1293
00:43:51,841 -->
00:43:53,021
tools to enhance patient care?
1294
00:43:53,566 -->
00:43:53,626
Dr. Stacey Johnston: I
1295
00:43:53,626 -->
00:43:54,766
mean, great question.
1296
00:43:54,766 -->
00:43:55,696
So I mean, that's where I
1297
00:43:55,696 -->
00:43:56,746
think the sky's the limit.
1298
00:43:56,746 -->
00:43:58,666
So, you know, we've got all of
1299
00:43:58,666 -->
00:44:00,856
this, this IOT world, you know, so
1300
00:44:00,856 -->
00:44:02,446
we have all of this patient data.
1301
00:44:02,446 -->
00:44:04,156
So whether it's a continuous glucose
1302
00:44:04,156 -->
00:44:07,036
monitor, blood pressure cuff, um, you
1303
00:44:07,036 -->
00:44:08,896
know, we've got, you know, these telemetry
1304
00:44:08,896 -->
00:44:10,546
devices, your smart watch, you know,
1305
00:44:10,546 -->
00:44:12,436
so all of this wearable data, right?
1306
00:44:12,436 -->
00:44:14,746
And so coming in, um.
1307
00:44:15,306 -->
00:44:17,646
And so it can come into the EHR
1308
00:44:17,646 -->
00:44:19,296
and ingest it pretty easily.
1309
00:44:19,326 -->
00:44:20,586
That's not the hard part.
1310
00:44:20,646 -->
00:44:22,716
What the hard part is from where
1311
00:44:22,716 -->
00:44:24,126
I started at the very beginning
1312
00:44:24,516 -->
00:44:26,076
is the workflow around that.
1313
00:44:26,076 -->
00:44:27,996
So if you have someone's whose blood
1314
00:44:27,996 -->
00:44:30,546
pressure's reading at, uh, you know, 200.
1315
00:44:30,711 -->
00:44:31,671
What are you gonna do?
1316
00:44:31,671 -->
00:44:34,821
Like who's monitoring that alert and does
1317
00:44:34,821 -->
00:44:36,291
that alert come to the primary care doc?
1318
00:44:36,291 -->
00:44:37,611
Well, what if he's off that week?
1319
00:44:37,971 -->
00:44:38,931
Does it, you know, does it
1320
00:44:38,931 -->
00:44:40,341
go to a case management list?
1321
00:44:40,371 -->
00:44:41,691
You know, 'cause we, you can
1322
00:44:41,691 -->
00:44:43,131
embed it into a registry.
1323
00:44:43,401 -->
00:44:44,421
So is there someone that's
1324
00:44:44,421 -->
00:44:46,011
always monitoring this registry?
1325
00:44:46,011 -->
00:44:47,601
Are they, is it a nurse practitioner
1326
00:44:47,601 -->
00:44:49,011
that's just monitor, like
1327
00:44:49,191 -->
00:44:50,991
managing a hypertension clinic?
1328
00:44:50,991 -->
00:44:53,091
So there are some organizations
1329
00:44:53,091 -->
00:44:54,591
that are doing this really well in
1330
00:44:54,591 -->
00:44:56,961
the wearable space where we are not
1331
00:44:56,961 -->
00:44:58,761
there yet, just being totally honest.
1332
00:44:58,761 -->
00:44:59,091
But, um.
1333
00:44:59,971 -->
00:45:00,931
You know, I, I do think the
1334
00:45:00,931 -->
00:45:02,431
wearables are one components.
1335
00:45:02,431 -->
00:45:03,631
Then I think the other that I
1336
00:45:03,631 -->
00:45:04,531
talked about last night with
1337
00:45:04,531 -->
00:45:06,001
the physicians is genomics.
1338
00:45:06,301 -->
00:45:07,996
Um, genomics is, you know, um.
1339
00:45:09,121 -->
00:45:10,591
I think the studies show that every
1340
00:45:10,591 -->
00:45:12,481
three or four out of five, um, drugs
1341
00:45:12,481 -->
00:45:14,371
now have some sort of genomic marker
1342
00:45:14,731 -->
00:45:16,651
being tested as it's going through FDA
1343
00:45:16,651 -->
00:45:19,981
approval and, you know, depression, meds,
1344
00:45:20,041 -->
00:45:22,951
um, antipsychotics, um, blood thinners.
1345
00:45:22,981 -->
00:45:25,141
I mean, a lot of medications right
1346
00:45:25,141 -->
00:45:26,611
now have some sort of genetic
1347
00:45:26,611 -->
00:45:28,651
testing and that's additional
1348
00:45:28,651 -->
00:45:29,791
cognitive burden to our docs.
1349
00:45:30,041 -->
00:45:32,291
There is no way they are gonna remember
1350
00:45:32,561 -->
00:45:33,941
everything that needs to be done
1351
00:45:33,941 -->
00:45:35,891
for your patients at any given time.
1352
00:45:36,251 -->
00:45:38,111
So embedding that genetic
1353
00:45:38,111 -->
00:45:40,061
data to alert you and say, oh,
1354
00:45:40,061 -->
00:45:41,351
you're about to order Pradaxa.
1355
00:45:41,351 -->
00:45:43,641
This patient isn't, uh, a candidate for
1356
00:45:43,641 -->
00:45:45,731
Pradaxa because of their genetic profile,
1357
00:45:46,061 -->
00:45:47,771
please consider one of these others.
1358
00:45:47,771 -->
00:45:49,751
And then the solution actually queues
1359
00:45:49,751 -->
00:45:51,701
up what the other recommendations are.
1360
00:45:52,031 -->
00:45:54,311
So I think between both the wearables and
1361
00:45:54,311 -->
00:45:56,741
the genetic data, you can now move from.
1362
00:45:57,006 -->
00:45:59,616
Predictive analytics to truly prescriptive
1363
00:45:59,616 -->
00:46:01,926
care, which is the right patient at the
1364
00:46:01,926 -->
00:46:04,506
right time with the right care, um, you
1365
00:46:04,506 -->
00:46:06,606
know, by, you know, really surrounding
1366
00:46:06,606 -->
00:46:08,076
them with, you know, the data and
1367
00:46:08,076 -->
00:46:09,966
analytics and the AI on top of that.
1368
00:46:10,836 -->
00:46:11,646
Krishna Gade: Wow, that's amazing.
1369
00:46:12,126 -->
00:46:13,476
I think that gets me into the future
1370
00:46:13,476 -->
00:46:14,736
of healthcare delivery, right?
1371
00:46:14,736 -->
00:46:15,966
Like, you know, fast forward five
1372
00:46:15,966 -->
00:46:18,276
to 10 years, what does a hospital
1373
00:46:18,276 -->
00:46:20,706
look like with deeply embedded AI?
1374
00:46:20,736 -->
00:46:21,516
You know, according to you.
1375
00:46:22,756 -->
00:46:23,746
Dr. Stacey Johnston: Yeah, so I think
1376
00:46:23,746 -->
00:46:25,936
what will end up happening is, you know,
1377
00:46:25,936 -->
00:46:27,706
there's a lot of consumer grade, um
1378
00:46:28,771 -->
00:46:30,421
Things that are happening right now, like
1379
00:46:30,421 -->
00:46:32,401
blood pressure cups and you know, and
1380
00:46:32,401 -->
00:46:34,771
then, you know, all the wearables and
1381
00:46:35,161 -->
00:46:37,771
even, um, telemedicine visits are, you
1382
00:46:37,771 -->
00:46:39,841
know, I think gonna continue to expand.
1383
00:46:39,841 -->
00:46:41,131
So I think in the primary care
1384
00:46:41,131 -->
00:46:43,231
space, you are gonna actually
1385
00:46:43,231 -->
00:46:44,791
see sticker patients in the.
1386
00:46:45,156 -->
00:46:46,896
Less sick patients are gonna just
1387
00:46:46,896 -->
00:46:48,516
kind of do their care at home.
1388
00:46:48,816 -->
00:46:50,886
You know, even prescriptions that
1389
00:46:50,946 -->
00:46:53,436
required an actual visit in a prescription
1390
00:46:53,706 -->
00:46:55,206
can now be bought over the counter.
1391
00:46:55,206 -->
00:46:57,696
So there's a, you know, like, um,
1392
00:46:57,756 -->
00:46:59,316
Omeprazole used to be a prescription
1393
00:46:59,316 -->
00:47:00,636
only, now it's bought over the counter.
1394
00:47:00,636 -->
00:47:03,156
So there's, and also with a large
1395
00:47:03,156 -->
00:47:04,656
language model and ChatGPT people
1396
00:47:05,376 -->
00:47:07,716
are gonna start to self care at home.
1397
00:47:08,196 -->
00:47:08,496
I think.
1398
00:47:08,701 -->
00:47:10,021
And then, and then it's
1399
00:47:10,021 -->
00:47:11,191
going to become sicker.
1400
00:47:11,191 -->
00:47:12,811
Patients gonna be seeing that primary
1401
00:47:12,811 -->
00:47:15,361
care office and they're gonna manage
1402
00:47:15,361 -->
00:47:18,211
those at, at home as best as possible.
1403
00:47:18,391 -->
00:47:19,621
And therefore your inpatient
1404
00:47:19,621 -->
00:47:21,511
days are gonna be even sicker.
1405
00:47:21,781 -->
00:47:23,821
And so, you know, I think you're.
1406
00:47:24,211 -->
00:47:26,281
What we used to see in the inpatient
1407
00:47:26,281 -->
00:47:28,651
space, some of these people that we would
1408
00:47:28,651 -->
00:47:30,061
keep a little bit longer because they
1409
00:47:30,061 -->
00:47:32,011
need some extra oxygen and we're gonna
1410
00:47:32,011 -->
00:47:34,021
take care of those patients at home.
1411
00:47:34,291 -->
00:47:36,511
And that, that truly would then leave the
1412
00:47:36,511 -->
00:47:38,461
beds for the, the sickest of the sick.
1413
00:47:38,491 -->
00:47:39,481
So I think we're gonna
1414
00:47:39,841 -->
00:47:41,281
really see the acuity in our
1415
00:47:41,281 -->
00:47:43,051
hospitals going up as well too.
1416
00:47:43,441 -->
00:47:45,721
Um, but on the flip side, I think
1417
00:47:45,721 -->
00:47:47,311
we're gonna have more automation.
1418
00:47:47,521 -->
00:47:48,421
We're gonna have, you know,
1419
00:47:48,541 -->
00:47:50,071
these cameras in the rooms and
1420
00:47:50,071 -->
00:47:51,901
the AI predicting the care.
1421
00:47:52,201 -->
00:47:53,401
Um, you know, um.
1422
00:47:54,301 -->
00:47:57,241
AI right now without a, without an actual
1423
00:47:57,241 -->
00:47:59,401
blood pressure cuff, is able to start
1424
00:47:59,401 -->
00:48:01,291
detecting blood pressure, temperature,
1425
00:48:01,321 -->
00:48:03,871
heart rate, um, just by cameras, you know?
1426
00:48:03,871 -->
00:48:06,331
So I think there's just more we would
1427
00:48:06,331 -->
00:48:08,821
be able to do, um, with the tech
1428
00:48:09,121 -->
00:48:11,431
technical aspects and the technology
1429
00:48:11,431 -->
00:48:13,831
surrounding, you know, the patient in
1430
00:48:13,831 -->
00:48:15,961
the individual hospital room so that you
1431
00:48:15,961 -->
00:48:17,821
know, that hospital room of the future.
1432
00:48:18,181 -->
00:48:20,071
Um, so that, uh, so you're
1433
00:48:20,071 -->
00:48:22,741
monitoring these sicker patients, um.
1434
00:48:23,416 -->
00:48:25,276
Kind of more around the clock instead
1435
00:48:25,276 -->
00:48:27,046
of relying on people always going
1436
00:48:27,046 -->
00:48:28,426
in and looking at them all the time.
1437
00:48:29,146 -->
00:48:29,926
Krishna Gade: Yeah, it makes sense.
1438
00:48:30,196 -->
00:48:31,906
And are there gonna be some new roles
1439
00:48:31,906 -->
00:48:33,556
in healthcare because of these changes?
1440
00:48:33,556 -->
00:48:34,816
Are there, like, are there's gonna be
1441
00:48:34,816 -->
00:48:37,216
a job job shifting as well happening
1442
00:48:37,216 -->
00:48:38,326
in healthcare because of this?
1443
00:48:38,416 -->
00:48:38,926
Dr. Stacey Johnston: Yeah, I think
1444
00:48:38,926 -->
00:48:40,276
we're gonna obviously continue
1445
00:48:40,276 -->
00:48:41,806
to have to upskill our folks.
1446
00:48:41,866 -->
00:48:44,206
Um, so, you know, everyone's gonna need
1447
00:48:44,206 -->
00:48:45,736
to practice to their top of the license.
1448
00:48:45,736 -->
00:48:47,356
But don't forget, we're also going,
1449
00:48:47,626 -->
00:48:48,736
you know, already in a nursing
1450
00:48:48,736 -->
00:48:50,566
shortage and a physician shortage.
1451
00:48:50,836 -->
00:48:52,756
So how are we gonna face both, you
1452
00:48:52,756 -->
00:48:54,766
know, the clinical caregiver shortage?
1453
00:48:55,001 -->
00:48:56,381
Well, we're gonna need to upskill our
1454
00:48:56,381 -->
00:48:58,151
CNAs, we're gonna need to upskill,
1455
00:48:58,181 -->
00:49:00,011
um, our, you know, transport team.
1456
00:49:00,011 -->
00:49:02,561
So, you know, the more we can upskill the
1457
00:49:02,561 -->
00:49:05,531
folks that we may have, um, that they can
1458
00:49:05,531 -->
00:49:07,481
maybe go into some of these other roles.
1459
00:49:07,511 -->
00:49:09,671
Um, I know some hospital
1460
00:49:09,671 -->
00:49:11,591
systems are looking at, um.
1461
00:49:12,391 -->
00:49:14,101
You know, partnering with community
1462
00:49:14,101 -->
00:49:16,021
colleges that, you know, they come
1463
00:49:16,021 -->
00:49:18,001
directly outta community college to
1464
00:49:18,001 -->
00:49:20,011
go into some of these, um, you know,
1465
00:49:20,011 -->
00:49:22,561
these roles, but I think upskill and
1466
00:49:22,771 -->
00:49:24,781
definitely upskilling on your IT team.
1467
00:49:24,781 -->
00:49:27,271
So how to manage an AI model and, you
1468
00:49:27,271 -->
00:49:28,951
know, building, are you, you know, a
1469
00:49:28,951 -->
00:49:31,351
builder or a buy program and you know.
1470
00:49:31,821 -->
00:49:33,231
Your smaller community hospitals
1471
00:49:33,231 -->
00:49:35,031
may continue to be a BY program.
1472
00:49:35,031 -->
00:49:36,981
I think us at Beacon managing a
1473
00:49:36,981 -->
00:49:38,361
large language model is probably
1474
00:49:39,051 -->
00:49:40,581
not gonna happen anytime soon.
1475
00:49:40,581 -->
00:49:42,291
So that's why we partner, but you
1476
00:49:42,291 -->
00:49:44,931
know, um, I, you know, upskilling
1477
00:49:44,931 -->
00:49:46,401
your team, but upskilling the
1478
00:49:46,401 -->
00:49:47,661
clinical people as well too.
1479
00:49:48,471 -->
00:49:48,651
Krishna Gade: Awesome.
1480
00:49:49,131 -->
00:49:50,301
Okay, so we are almost at
1481
00:49:50,301 -->
00:49:51,171
the end of the webinar.
1482
00:49:51,201 -->
00:49:52,461
Uh, let's actually wrap
1483
00:49:52,461 -->
00:49:53,631
this up with like, uh.
1484
00:49:54,286 -->
00:49:55,246
You know, lightning round.
1485
00:49:55,336 -->
00:49:56,806
So like, you know, one or two
1486
00:49:56,806 -->
00:49:58,876
word answers Dr. Johnston.
1487
00:49:58,876 -->
00:50:01,366
So, uh, ask, quickly, ask these questions.
1488
00:50:01,366 -->
00:50:02,986
You know, one AI myth in
1489
00:50:02,986 -->
00:50:04,546
healthcare for you right now?
1490
00:50:04,546 -->
00:50:05,626
Dr. Stacey Johnston: Uh, that I
1491
00:50:05,626 -->
00:50:06,706
think it's gonna get rid of the
1492
00:50:06,706 -->
00:50:09,886
caregivers anytime soon in the future.
1493
00:50:09,886 -->
00:50:11,296
Maybe five, 10 years from now.
1494
00:50:11,296 -->
00:50:12,736
But right now we're not seeing that.
1495
00:50:12,886 -->
00:50:13,186
Krishna Gade: Okay.
1496
00:50:13,276 -->
00:50:13,576
Great.
1497
00:50:14,116 -->
00:50:15,676
And one company doing it, right?
1498
00:50:16,846 -->
00:50:19,786
Dr. Stacey Johnston: Uh, we've, I, I had
1499
00:50:19,786 -->
00:50:21,436
to say right now I'm really impressed
1500
00:50:21,436 -->
00:50:23,536
with the Oracle and their clinical AI
1501
00:50:23,536 -->
00:50:26,386
agent and, um, you know, so we're doubling
1502
00:50:26,386 -->
00:50:28,306
down on our Oracle experience and going
1503
00:50:28,306 -->
00:50:31,396
all in with their new, um, agentic EHR.
1504
00:50:32,746 -->
00:50:32,836
Krishna Gade: Okay.
1505
00:50:32,836 -->
00:50:33,826
And one capability that
1506
00:50:33,826 -->
00:50:34,906
you wish existed today?
1507
00:50:36,466 -->
00:50:37,546
Dr. Stacey Johnston: Uh, I, I
1508
00:50:37,586 -->
00:50:40,376
really think just being like totally
1509
00:50:40,376 -->
00:50:42,301
keyboardless, like ambient everything.
1510
00:50:42,301 -->
00:50:43,261
Just being able to have a
1511
00:50:43,261 -->
00:50:44,641
conversation with the computer,
1512
00:50:44,911 -->
00:50:46,441
a conversation with the patient.
1513
00:50:46,681 -->
00:50:48,601
Just you know, your thought process
1514
00:50:48,661 -->
00:50:50,581
and you know, queuing up your orders,
1515
00:50:50,611 -->
00:50:52,531
everything being totally ambient.
1516
00:50:52,861 -->
00:50:53,911
I think it's coming,
1517
00:50:53,911 -->
00:50:54,961
but we're not there yet.
1518
00:50:55,561 -->
00:50:55,771
Krishna Gade: Got it.
1519
00:50:55,771 -->
00:50:58,156
And this is probably interesting, AI
1520
00:50:58,156 -->
00:51:00,691
as a co-pilot or autopilot in medicine.
1521
00:51:01,696 -->
00:51:01,696
Dr. Stacey Johnston:
1522
00:51:01,696 -->
00:51:02,896
Uh, I think it's both.
1523
00:51:02,896 -->
00:51:04,516
I think it, for the most part
1524
00:51:04,516 -->
00:51:05,956
should be a co-pilot in the clinical
1525
00:51:05,956 -->
00:51:07,486
space, but I don't see any reason
1526
00:51:07,486 -->
00:51:08,746
why it couldn't be autopilot in
1527
00:51:08,746 -->
00:51:10,096
some of the backend processes.
1528
00:51:10,846 -->
00:51:11,026
Krishna Gade: Okay.
1529
00:51:11,596 -->
00:51:13,096
So finally, if you went back
1530
00:51:13,096 -->
00:51:14,716
to being a physician today, how
1531
00:51:14,716 -->
00:51:16,066
would you use AI differently?
1532
00:51:17,446 -->
00:51:18,706
Dr. Stacey Johnston: Uh, definitely,
1533
00:51:18,856 -->
00:51:20,986
you know, I would embrace it, you
1534
00:51:20,986 -->
00:51:22,816
know, is, you know, seeing more and
1535
00:51:22,816 -->
00:51:26,011
more patients, um mostly just to
1536
00:51:26,011 -->
00:51:28,201
finish on time and go home on time.
1537
00:51:28,231 -->
00:51:30,661
You know, I, the EHR is definitely
1538
00:51:30,661 -->
00:51:32,881
hard work and it adds more clicks
1539
00:51:32,881 -->
00:51:35,311
to the docs, and so if I could use
1540
00:51:35,311 -->
00:51:37,681
AI to make my life easier, I, I
1541
00:51:37,831 -->
00:51:39,331
would a hundred percent embrace it.
1542
00:51:40,096 -->
00:51:40,516
Krishna Gade: Awesome.
1543
00:51:41,206 -->
00:51:43,051
Well, thank you so much Dr. Johnston.
1544
00:51:43,126 -->
00:51:44,836
Uh, thanks for your valuable time
1545
00:51:44,836 -->
00:51:46,336
here with our audience and, uh, I
1546
00:51:46,336 -->
00:51:48,076
learned a ton about AI in healthcare
1547
00:51:48,076 -->
00:51:49,756
and your journey from being a
1548
00:51:49,756 -->
00:51:51,826
physician to now a technologist.
1549
00:51:51,826 -->
00:51:54,226
And, uh, some great insights and thanks.
1550
00:51:54,226 -->
00:51:54,226
Dr. Stacey Johnston:
1551
00:51:54,226 -->
00:51:55,486
Oh, thank you so much.
1552
00:51:56,026 -->
00:51:56,296
Krishna Gade: Awesome.
1553
00:51:56,986 -->
00:51:58,666
But that's it for this,
1554
00:51:58,726 -->
00:52:00,226
uh, month AI Explained.
1555
00:52:00,226 -->
00:52:02,266
Folks, uh, we'll come back to
1556
00:52:02,266 -->
00:52:03,526
you with another great guest
1557
00:52:03,526 -->
00:52:05,536
on, um, uh, in the month in, I
1558
00:52:05,536 -->
00:52:06,676
think in the next few weeks then.
1559
00:52:07,036 -->
00:52:07,441
See you all.
1560
00:52:09,331 -->
00:52:09,571
Dr. Stacey Johnston: Great.
1561
00:52:09,571 -->
00:52:09,866
Thank you.
