WEBVTT

340b994f-eda6-44de-a022-5799228740fc-0
00:00:20.800 --> 00:00:21.680
Welcome everyone.

75829464-7912-4e34-8141-5ec1de949658-0
00:00:24.080 --> 00:00:27.200
This past year has been a turning point.

b0269144-f8c2-4c06-b8a4-4ebcbc942151-0
00:00:27.240 --> 00:00:31.617
I'd say every health system is talking
about AI,

b0269144-f8c2-4c06-b8a4-4ebcbc942151-1
00:00:31.617 --> 00:00:35.280
but I'd say most are maybe still talking.

237b6c83-f029-4dc9-b02a-9301339561d3-0
00:00:35.680 --> 00:00:41.256
Only a few are actually changing how
either care is delivered,

237b6c83-f029-4dc9-b02a-9301339561d3-1
00:00:41.256 --> 00:00:47.452
how patient experience is transformed,
how operations is transformed,

237b6c83-f029-4dc9-b02a-9301339561d3-2
00:00:47.452 --> 00:00:49.399
or how data gets used.

c62001cf-49d9-4b23-8375-da0380ad3bc3-0
00:00:50.000 --> 00:00:56.382
So today we're going beyond the hype,
asking what truly separates the leaders

c62001cf-49d9-4b23-8375-da0380ad3bc3-1
00:00:56.382 --> 00:01:02.110
from the laggards in healthcare AI,
and how we can make sure Canadian

c62001cf-49d9-4b23-8375-da0380ad3bc3-2
00:01:02.110 --> 00:01:06.120
healthcare specifically doesn't get left
behind.

57f35f40-89e3-46a2-ae7b-e69397dba41c-0
00:01:08.840 --> 00:01:13.208
Let's go into getting to learn a little
bit more about the panels,

57f35f40-89e3-46a2-ae7b-e69397dba41c-1
00:01:13.208 --> 00:01:17.120
but before we do that,
some logistics here in housekeeping.

315cde6d-6c4c-4f06-a67d-66b619af228d-0
00:01:17.800 --> 00:01:21.000
There's a chat function that you'll see
on the bottom left.

6a4e468c-bd91-4520-a2f3-d55abce7607c-0
00:01:21.840 --> 00:01:24.720
You could use that to sort of react and
participate.

9e742499-9f9c-4103-8e33-23e558243622-0
00:01:24.720 --> 00:01:28.320
So please feel free to use it through the
session.

f6bf4049-a94f-494f-a6a0-8a5c659b5733-0
00:01:28.600 --> 00:01:32.273
There's also Q&A,
so please submit your questions to the

f6bf4049-a94f-494f-a6a0-8a5c659b5733-1
00:01:32.273 --> 00:01:34.200
speakers by using that function.

cfc96ea3-e2e6-4397-a318-4c5164714b6c-0
00:01:34.440 --> 00:01:40.706
And I'll make sure to go in there and
pull out a few questions and be able to

cfc96ea3-e2e6-4397-a318-4c5164714b6c-1
00:01:40.706 --> 00:01:44.000
pose them to the speakers as we go along.

c3fcd84d-8dc2-4933-a822-fa4fe1148907-0
00:01:44.280 --> 00:01:48.320
So please keep it interactive and look
for your questions as we go along.

edc083e1-d396-40f1-b038-8df35ff02052-0
00:01:49.760 --> 00:01:55.112
OK, so having said that,
I'm going to get our panelists to choose

edc083e1-d396-40f1-b038-8df35ff02052-1
00:01:55.112 --> 00:02:01.520
themselves and I'm going to start with a
question that I'll ask each panelist.

b0df979d-34b5-4cd5-af4d-39d52126c4a5-0
00:02:01.720 --> 00:02:06.920
And that question is, well,
let's start broad.

5dfe8218-df53-402e-9f6d-8f7f9e11bbe7-0
00:02:07.560 --> 00:02:13.680
When you look at healthcare AI right now,
what's one thing that impresses you and

5dfe8218-df53-402e-9f6d-8f7f9e11bbe7-1
00:02:13.680 --> 00:02:15.920
one thing that frustrates you?

ea5fd95d-2ce7-4351-95f7-d0110d13f2df-0
00:02:16.160 --> 00:02:18.949
So chance,
if you want to start with that and it's

ea5fd95d-2ce7-4351-95f7-d0110d13f2df-1
00:02:18.949 --> 00:02:22.560
just kindly introduce yourself and your
role in the organization.

7d4360f8-93ab-493b-b5b5-b7f6c2253548-0
00:02:22.560 --> 00:02:25.160
And then if you could answer that,
that'd be great.

5f5ce49b-f04f-43c8-abd9-a5d5dfd424e0-0
00:02:28.120 --> 00:02:29.520
Absolutely.

4b8c0776-99df-4e1b-8729-caf944eb4080-0
00:02:30.280 --> 00:02:30.640
Yeah.

18898e0e-20ef-4f43-9b5a-0b5b39e870e4-0
00:02:30.640 --> 00:02:32.874
Well,
thank you for inviting me to this panel

18898e0e-20ef-4f43-9b5a-0b5b39e870e4-1
00:02:32.874 --> 00:02:35.400
and I'm really looking forward to the
conversation.

8d470796-450c-440f-86e0-3002f157c96e-0
00:02:36.960 --> 00:02:37.760
My name is Chance.

d32e0848-ec95-4f80-a88a-1d8c9128f21a-0
00:02:37.760 --> 00:02:41.531
I am the Executive Director of Technology
and Analytics and the Chief Privacy

d32e0848-ec95-4f80-a88a-1d8c9128f21a-1
00:02:41.531 --> 00:02:44.240
Officer of the Canadian Medical
Protective Association.

be65adbc-ebcd-4291-b8b4-5330cd346b39-0
00:02:44.800 --> 00:02:49.568
So for those of you who don't know,
know what my organization does is we

be65adbc-ebcd-4291-b8b4-5330cd346b39-1
00:02:49.568 --> 00:02:53.683
provide medical legal assistance to most
physicians in Canada,

be65adbc-ebcd-4291-b8b4-5330cd346b39-2
00:02:53.683 --> 00:02:55.120
I think 9798% of them.

1e9c0bc8-e2c8-4a36-9af5-da5dd7a7f14e-0
00:02:55.920 --> 00:03:00.769
And so anytime there's a, a complaint,
maybe a lawsuit brought against

1e9c0bc8-e2c8-4a36-9af5-da5dd7a7f14e-1
00:03:00.769 --> 00:03:03.160
physicians, we provide the defense.

c72d6535-dfa2-4637-875f-82a28b04cfed-0
00:03:03.160 --> 00:03:06.556
And if you know,
it's found that the care didn't meet the

c72d6535-dfa2-4637-875f-82a28b04cfed-1
00:03:06.556 --> 00:03:08.840
standard, we, we pay patients promptly.

1fa97eff-9bb6-4767-b90f-f6e0be3a74b4-0
00:03:09.640 --> 00:03:12.880
And so, you know,
our interest in an AI is twofold.

e6a2693c-ddcd-4998-8de8-2e2893261af9-0
00:03:13.520 --> 00:03:17.348
On the one hand, we're,
we're trying to think through how we can

e6a2693c-ddcd-4998-8de8-2e2893261af9-1
00:03:17.348 --> 00:03:21.707
provide advice to physicians and
healthcare on how to use this technology

e6a2693c-ddcd-4998-8de8-2e2893261af9-2
00:03:21.707 --> 00:03:22.120
safely.

edc8d204-d9e1-4729-be9a-cba426717de6-0
00:03:22.720 --> 00:03:27.305
The other thing is, if you can imagine,
every suit, complaint,

edc8d204-d9e1-4729-be9a-cba426717de6-1
00:03:27.305 --> 00:03:32.545
piece of advice that we provide flows
through our organization and this

edc8d204-d9e1-4729-be9a-cba426717de6-2
00:03:32.545 --> 00:03:36.840
provides us with a massive amount of data
and information.

a1960683-9ddc-46bc-9e3b-679642f9bff6-0
00:03:37.760 --> 00:03:42.283
And so internally we're,
we're looking at how can we analyze this

a1960683-9ddc-46bc-9e3b-679642f9bff6-1
00:03:42.283 --> 00:03:47.971
information to improve the research that
we provide to healthcare and to transform

a1960683-9ddc-46bc-9e3b-679642f9bff6-2
00:03:47.971 --> 00:03:49.000
our operations.

1a0ead50-b0fe-4ae2-bac9-8ea663528875-0
00:03:49.800 --> 00:03:51.680
And so that's kind of what we've been
doing.

8722cdbc-ed44-4370-a773-7f374b14cf00-0
00:03:52.400 --> 00:03:58.834
So that's kind of like a a quick rundown
on my organization and our interest in AI

8722cdbc-ed44-4370-a773-7f374b14cf00-1
00:03:58.834 --> 00:04:05.346
To your very interesting question on the
progress side, what impresses me, I think,

8722cdbc-ed44-4370-a773-7f374b14cf00-2
00:04:05.346 --> 00:04:09.299
you know,
I've got a couple degrees in in computer

8722cdbc-ed44-4370-a773-7f374b14cf00-3
00:04:09.299 --> 00:04:09.920
science.

02659016-fb40-4790-95b7-71bf03aa4995-0
00:04:09.920 --> 00:04:16.132
And so just that the quantum change in
how we deal with text information and

02659016-fb40-4790-95b7-71bf03aa4995-1
00:04:16.132 --> 00:04:19.360
visual information is it's mind blowing.

89ce7524-2672-4fca-b723-9aba09d17c8e-0
00:04:19.360 --> 00:04:24.760
This is opening up so many new avenues of
how you can use data and technology.

afcd0397-0c82-416b-b0a1-edd39f04d4a5-0
00:04:25.800 --> 00:04:31.493
The thing that frustrates me,
it's all around data and the protection

afcd0397-0c82-416b-b0a1-edd39f04d4a5-1
00:04:31.493 --> 00:04:33.120
of it less and less.

6303676b-1a7c-4db9-9280-f0fc112f5ee8-0
00:04:33.200 --> 00:04:38.240
I am clear on whether I trust AI vendors
to protect my data.

370bc3d0-944a-4a88-82b1-8a8eb50be214-0
00:04:39.320 --> 00:04:42.720
And so those would be my my 2 pieces and
the answer to your question.

0e911497-3747-47a0-8de8-eed03bec928e-0
00:04:42.720 --> 00:04:43.640
Eric, Excellent.

92485887-f89b-4492-ac7b-ed5e40150d06-0
00:04:45.880 --> 00:04:46.720
Devin, you're next.

f2e5ac00-80f5-474f-9b18-f8e9b2f7908a-0
00:04:47.520 --> 00:04:48.120
Yeah, sure.

2fa65ae6-0e85-4918-916a-a724e1135e62-0
00:04:48.120 --> 00:04:49.960
And it's so great to be here with
everyone.

b865af0f-21ea-4b82-9843-1a0db9c755ee-0
00:04:49.960 --> 00:04:54.498
I'm already getting text messages of
people sending me messages excited about

b865af0f-21ea-4b82-9843-1a0db9c755ee-1
00:04:54.498 --> 00:04:55.080
the panel.

918ea016-2d03-452e-800e-694d2814ec19-0
00:04:55.080 --> 00:04:56.760
And so my name is Devin.

85384a69-1cb7-4c77-91ee-a07b43d2541b-0
00:04:56.760 --> 00:05:01.396
I'm an emergency physician at the
Hospital for Sick Children or Sick Kids

85384a69-1cb7-4c77-91ee-a07b43d2541b-1
00:05:01.396 --> 00:05:01.960
Hospital.

6000965c-e43f-4f8e-947c-cc250fd23593-0
00:05:02.400 --> 00:05:03.840
I'm also a computer scientist.

16246a44-f7ac-4be3-9349-5ba382c36497-0
00:05:03.840 --> 00:05:08.019
And so throughout my training,
just having some really tough patient

16246a44-f7ac-4be3-9349-5ba382c36497-1
00:05:08.019 --> 00:05:12.987
cases and frustrations around some of the
limitations in the way we as clinicians

16246a44-f7ac-4be3-9349-5ba382c36497-2
00:05:12.987 --> 00:05:16.440
can deliver care just got me really angry,
to be honest.

09be7abc-5a48-4836-a3f6-fb0fd2063987-0
00:05:16.440 --> 00:05:21.400
And wanted to think through how to change
the way we deliver care at a large scale.

a4ab34a0-6c90-4046-8d27-dabc0ba2cd53-0
00:05:21.400 --> 00:05:24.987
And so ended up doing my master's of
computer science from the University of

a4ab34a0-6c90-4046-8d27-dabc0ba2cd53-1
00:05:24.987 --> 00:05:25.360
Toronto.

626275ac-23a2-45aa-b345-2dc176ab166b-0
00:05:25.840 --> 00:05:30.888
And that's led me to now at Sick Kids Co
lead something called Sky or Sick Kids AI,

626275ac-23a2-45aa-b345-2dc176ab166b-1
00:05:30.888 --> 00:05:33.953
which is an or Sick Kids AI service,
I should say,

626275ac-23a2-45aa-b345-2dc176ab166b-2
00:05:33.953 --> 00:05:38.821
which is an enterprise wide service that
looks through transforming every pillar

626275ac-23a2-45aa-b345-2dc176ab166b-3
00:05:38.821 --> 00:05:42.847
of the hospital with AI,
keeping a keen lens on responsible AI and

626275ac-23a2-45aa-b345-2dc176ab166b-4
00:05:42.847 --> 00:05:47.775
thinking through how do we operationalize
responsible AI both into operations and

626275ac-23a2-45aa-b345-2dc176ab166b-5
00:05:47.775 --> 00:05:50.240
into the bedside to improve patient care.

288cf095-2d8c-4e5d-a7d7-71cbd255133d-0
00:05:50.600 --> 00:05:55.262
I'm also the Co founder and CEO of a
healthcare technology company called Hero

288cf095-2d8c-4e5d-a7d7-71cbd255133d-1
00:05:55.262 --> 00:05:55.440
AI.

3766771a-a290-4c90-aa7e-1a07bdb71976-0
00:05:56.200 --> 00:05:59.200
They just sort of spun out of sick kids
and is doing some incredible work.

b5a72e4b-b5ac-4715-b887-35dee257eb64-0
00:05:59.200 --> 00:06:04.333
And so I see this incredible opportunity
for artificial intelligence with my

b5a72e4b-b5ac-4715-b887-35dee257eb64-1
00:06:04.333 --> 00:06:07.000
clinician hat on, to expand my capacity.

dd06a2f4-a151-451e-948c-ca1b2dfd586d-0
00:06:07.320 --> 00:06:10.302
You know,
when I'm busy providing urgent care in

dd06a2f4-a151-451e-948c-ca1b2dfd586d-1
00:06:10.302 --> 00:06:15.170
one place in the emergency department,
the fact that AI can now process data in

dd06a2f4-a151-451e-948c-ca1b2dfd586d-2
00:06:15.170 --> 00:06:20.161
real time and begin to apply aspects of
automation to care for patients who might

dd06a2f4-a151-451e-948c-ca1b2dfd586d-3
00:06:20.161 --> 00:06:24.665
be still waiting for me to advance their
care forward, to improve safety,

dd06a2f4-a151-451e-948c-ca1b2dfd586d-4
00:06:24.665 --> 00:06:28.073
to improve that patient experience,
improve wait times,

dd06a2f4-a151-451e-948c-ca1b2dfd586d-5
00:06:28.073 --> 00:06:29.959
I think is incredibly powerful.

b0039bc7-dfa4-41b4-a377-2feffcba8c5e-0
00:06:30.920 --> 00:06:31.320
To be honest.

ad1c9d64-2de2-4673-ad08-032fcabcb8e8-0
00:06:31.320 --> 00:06:33.280
I don't get frustrated by many things.

d552a78f-85fe-4ce0-8221-9bd0681a3c77-0
00:06:33.280 --> 00:06:38.259
And so I'm not really frustrated by the
AI journey because I think it makes sense

d552a78f-85fe-4ce0-8221-9bd0681a3c77-1
00:06:38.259 --> 00:06:43.360
that there's going to be slow adoption in
a risk averse industry and rightfully so.

2a0f4e5f-06da-4f6e-9b3d-08bcfe30cdef-0
00:06:43.360 --> 00:06:48.095
And so where I see frustrations is rather
opportunities to think through creative

2a0f4e5f-06da-4f6e-9b3d-08bcfe30cdef-1
00:06:48.095 --> 00:06:52.022
ways on how to innovate on privacy,
the regulatory, the governance,

2a0f4e5f-06da-4f6e-9b3d-08bcfe30cdef-2
00:06:52.022 --> 00:06:56.641
and how do you create the processes and
procedures in place to build the runway

2a0f4e5f-06da-4f6e-9b3d-08bcfe30cdef-3
00:06:56.641 --> 00:07:00.279
to land AI technologies to the bedside
safely and effectively.

c5a13a5c-0ee9-47ca-869e-8bab63382782-0
00:07:01.160 --> 00:07:03.200
And so it's a bit about me and I'll,
I'll pass it back.

06dd2bc6-614e-4e6b-b1b1-673eeda849c0-0
00:07:03.240 --> 00:07:03.800
Thank you so much.

899ff009-1846-4aa3-a616-3a1aff6dca02-0
00:07:04.560 --> 00:07:06.040
Excellent moment.

cd737452-a0f3-47c0-9a58-d76bf124a1c3-0
00:07:07.400 --> 00:07:09.160
Hi everyone, pleasure to be here.

3ee572bc-9144-4ba3-9909-6cd1b1f9e40f-0
00:07:10.800 --> 00:07:13.242
I,
I head up a program currently at Unity

3ee572bc-9144-4ba3-9909-6cd1b1f9e40f-1
00:07:13.242 --> 00:07:17.370
Health Toronto as Vice President of Data
Science and Events analytics,

3ee572bc-9144-4ba3-9909-6cd1b1f9e40f-2
00:07:17.370 --> 00:07:21.498
but where we have a team of about 30
people that develop and deploy AI

3ee572bc-9144-4ba3-9909-6cd1b1f9e40f-3
00:07:21.498 --> 00:07:22.080
solutions.

ff794881-e280-4378-b9fa-e33dda3c3b5f-0
00:07:23.040 --> 00:07:26.304
We've developed and deployed over 50 AI
solutions,

ff794881-e280-4378-b9fa-e33dda3c3b5f-1
00:07:26.304 --> 00:07:29.120
the largest of any hospital in the
country.

f7101147-f137-415a-a924-8df9712ca9ec-0
00:07:29.320 --> 00:07:32.800
So we have kind of experience working
with AI solutions.

650c08dc-3054-432b-8717-1d305dde59d3-0
00:07:33.320 --> 00:07:34.240
And I think that excites.

c3e14381-74e5-4fbf-b7cf-2b6e0ec599f4-0
00:07:34.280 --> 00:07:38.414
What excites me the most is we've
actually seen real tangible benefits from

c3e14381-74e5-4fbf-b7cf-2b6e0ec599f4-1
00:07:38.414 --> 00:07:42.385
these solutions all the way from
mortality reductions to cost savings to

c3e14381-74e5-4fbf-b7cf-2b6e0ec599f4-2
00:07:42.385 --> 00:07:43.800
efficiency gains and such.

7dbdc513-9e52-4d6a-a664-d4f21d971842-0
00:07:44.680 --> 00:07:48.604
So I think there's a ton of potential
with these solutions,

7dbdc513-9e52-4d6a-a664-d4f21d971842-1
00:07:48.604 --> 00:07:53.640
but the frustration frustrating part is
there's this notion that I can solve

7dbdc513-9e52-4d6a-a664-d4f21d971842-2
00:07:53.640 --> 00:07:54.360
everything.

40221a2a-4746-466e-8820-f3aa52685656-0
00:07:54.960 --> 00:07:56.560
It can't in many cases.

090f8710-88e6-48a2-97fe-48ceb9c563de-0
00:07:56.960 --> 00:07:59.915
You know,
we see problems that require process map

090f8710-88e6-48a2-97fe-48ceb9c563de-1
00:07:59.915 --> 00:08:04.610
and it's more of a process issue that it
is an AI issue and we just have to stop

090f8710-88e6-48a2-97fe-48ceb9c563de-2
00:08:04.610 --> 00:08:09.304
doing a few things rather than relying on
AI and some Co opulated way of solving

090f8710-88e6-48a2-97fe-48ceb9c563de-3
00:08:09.304 --> 00:08:10.000
the problem.

d8851d23-406c-4855-857b-68d315ef0786-0
00:08:10.840 --> 00:08:16.020
So we're pretty what we're pretty focused
on making sure we get the right problems

d8851d23-406c-4855-857b-68d315ef0786-1
00:08:16.020 --> 00:08:20.640
to solve with AI because there are many
that just won't be able to solve.

ba7022b5-5290-453a-836e-155f8a5df14f-0
00:08:21.640 --> 00:08:27.179
But if you can get it right and not have
this notion of AI can be thrown at

ba7022b5-5290-453a-836e-155f8a5df14f-1
00:08:27.179 --> 00:08:30.969
everything,
I think you have something powerful and

ba7022b5-5290-453a-836e-155f8a5df14f-2
00:08:30.969 --> 00:08:31.480
muscle.

55cb94dd-76af-432f-9f85-95c92ca6ca4c-0
00:08:34.400 --> 00:08:36.440
I'm really happy to be able to join today
as well.

4dcdd700-97cb-45c6-bdcc-78183b4989a4-0
00:08:36.440 --> 00:08:39.634
My name is Marsha Bryan,
I'm the director of data analytics and

4dcdd700-97cb-45c6-bdcc-78183b4989a4-1
00:08:39.634 --> 00:08:42.680
quality at Ontario Shores Center for
Mental Health Sciences.

be664130-dc4f-4bc6-97c9-0e1939cf8cd7-0
00:08:43.440 --> 00:08:47.014
We are a standalone mental health
hospital located in Whitby,

be664130-dc4f-4bc6-97c9-0e1939cf8cd7-1
00:08:47.014 --> 00:08:51.280
just outside of Toronto in Ontario that
offers specialized assessment and

be664130-dc4f-4bc6-97c9-0e1939cf8cd7-2
00:08:51.280 --> 00:08:55.200
treatment for those living with complex
and serious mental illness.

7ef27465-d23f-4c59-99e9-3e75a8a7da32-0
00:08:56.600 --> 00:08:59.825
I think that I,
I've loved the answers that the other

7ef27465-d23f-4c59-99e9-3e75a8a7da32-1
00:08:59.825 --> 00:09:01.080
panelists have given.

a5e9c333-597f-4d79-9f8c-18dec320c7d0-0
00:09:01.080 --> 00:09:04.666
I think for me, I,
I think that one of the things that

a5e9c333-597f-4d79-9f8c-18dec320c7d0-1
00:09:04.666 --> 00:09:09.947
impresses me the most right now is again,
just that speed at which we are seeing

a5e9c333-597f-4d79-9f8c-18dec320c7d0-2
00:09:09.947 --> 00:09:14.120
the advancements with the,
the technology and the capabilities.

8e538c5d-0fb8-46e3-8cae-0bc5d02dfc77-0
00:09:15.400 --> 00:09:16.960
I am an optimist by nature.

37ddda0c-5ff7-4fdc-a01e-9cb9b16fd1a0-0
00:09:16.960 --> 00:09:23.424
So I do think that framing as Devin did,
that the frustrations are really part of

37ddda0c-5ff7-4fdc-a01e-9cb9b16fd1a0-1
00:09:23.424 --> 00:09:29.967
the journey that we need to go through in
order to realize the benefits of AI in a

37ddda0c-5ff7-4fdc-a01e-9cb9b16fd1a0-2
00:09:29.967 --> 00:09:36.037
healthcare setting where we need to
ensure that we're protecting individuals

37ddda0c-5ff7-4fdc-a01e-9cb9b16fd1a0-3
00:09:36.037 --> 00:09:39.978
and, and,
and looking at ways that we can do that

37ddda0c-5ff7-4fdc-a01e-9cb9b16fd1a0-4
00:09:39.978 --> 00:09:41.239
in a secure way.

32218f4a-74ea-41fe-af55-60164b8e9d64-0
00:09:41.560 --> 00:09:44.944
Understanding that at this time,
a lot of the,

32218f4a-74ea-41fe-af55-60164b8e9d64-1
00:09:44.944 --> 00:09:50.992
the hype and the information that's being
shared about opportunities may not always

32218f4a-74ea-41fe-af55-60164b8e9d64-2
00:09:50.992 --> 00:09:56.824
be aligned with our Canadian regulations,
with our guidelines in Ontario that so

32218f4a-74ea-41fe-af55-60164b8e9d64-3
00:09:56.824 --> 00:10:01.000
there's a there's,
there's extra work in order to kind of

32218f4a-74ea-41fe-af55-60164b8e9d64-4
00:10:01.000 --> 00:10:04.960
navigate that in terms of the work that
we need to do.

65bf0d6d-4ce2-4614-8325-91118e04f8d0-0
00:10:07.240 --> 00:10:11.003
But OK,
so we're already diving into a bit of

65bf0d6d-4ce2-4614-8325-91118e04f8d0-1
00:10:11.003 --> 00:10:17.301
test sort of value versus risk and
feasibility and where to start and how to

65bf0d6d-4ce2-4614-8325-91118e04f8d0-2
00:10:17.301 --> 00:10:18.120
get there.

48fbe291-baf9-4463-aa5f-3eaa0839eb18-0
00:10:18.400 --> 00:10:20.880
So maybe I'll start with my first
question.

4f51f098-865e-4be2-afdc-ac3a4d14f533-0
00:10:20.880 --> 00:10:26.883
This is direct to the moment You built
one of the most mature analytic shops in

4f51f098-865e-4be2-afdc-ac3a4d14f533-1
00:10:26.883 --> 00:10:28.760
healthcare, AI in Canada.

4bd823e5-48b2-4cce-b2d6-e27809807cc8-0
00:10:29.360 --> 00:10:34.768
Yet we're finding a lot of the hospitals
are still sort of stuck in,

4bd823e5-48b2-4cce-b2d6-e27809807cc8-1
00:10:34.768 --> 00:10:41.040
if I may say so, dashboarding and sort of,
you know, getting the data together.

2fd6c530-db63-4fa5-8b4e-f90f2fb34de8-0
00:10:41.560 --> 00:10:48.127
What's the biggest mindset shift or other
shifts that separates a data-driven or a

2fd6c530-db63-4fa5-8b4e-f90f2fb34de8-1
00:10:48.127 --> 00:10:52.480
data hospital to an AI native one or an
AI leader one?

f3e0f5d1-6b32-41d6-aa90-fcb6ff980afa-0
00:10:54.280 --> 00:10:55.800
Yeah, I think it's multiple.

6981a991-8a34-4052-ad9f-ac27126d0324-0
00:10:56.440 --> 00:10:58.080
There are multiple things that actually
drive this.

23a6768a-0b17-4349-9bd9-ee0796b31fb5-0
00:10:58.080 --> 00:11:03.465
The first is going to be AI literacy and
understanding what AI is and what AI

23a6768a-0b17-4349-9bd9-ee0796b31fb5-1
00:11:03.465 --> 00:11:03.880
needs.

8eb2e712-f62e-4108-98c3-7f96bbb743f5-0
00:11:04.880 --> 00:11:09.808
And I think the second big thing is
vision around what you want to be as a

8eb2e712-f62e-4108-98c3-7f96bbb743f5-1
00:11:09.808 --> 00:11:10.400
hospital.

08cbabfe-498b-4c96-9734-514fa0c9af7c-0
00:11:10.400 --> 00:11:13.984
Do you want to be an innovator where
you're paying for failures and have to

08cbabfe-498b-4c96-9734-514fa0c9af7c-1
00:11:13.984 --> 00:11:14.880
invest quite a bit?

aa1b4abb-ca69-450a-ac65-c884336dbbc6-0
00:11:15.400 --> 00:11:19.344
Or a lagger that waits for everyone to
fail where you don't have to make those

aa1b4abb-ca69-450a-ac65-c884336dbbc6-1
00:11:19.344 --> 00:11:22.640
investments and you can pick the winners
or somewhere in between.

d251da13-424d-47c6-a072-bc31f6c11d25-0
00:11:23.480 --> 00:11:26.987
And based on that vision and that
strategy, the investment,

d251da13-424d-47c6-a072-bc31f6c11d25-1
00:11:26.987 --> 00:11:31.604
a lot of hospitals just simply don't make
the investment because if you if you

d251da13-424d-47c6-a072-bc31f6c11d25-2
00:11:31.604 --> 00:11:35.754
build local solutions, for example,
you need to set up infrastructure,

d251da13-424d-47c6-a072-bc31f6c11d25-3
00:11:35.754 --> 00:11:39.320
compute infrastructure,
OPS infrastructure to enable the AI.

73166b99-c504-489f-971c-ef108c122da5-0
00:11:39.600 --> 00:11:43.175
And you also need to have people who have
expertise in this space to develop

73166b99-c504-489f-971c-ef108c122da5-1
00:11:43.175 --> 00:11:43.640
solutions.

e63d8c71-9f17-4b39-b75a-5b0e901c981b-0
00:11:45.440 --> 00:11:47.034
But again,
if you're going to be a laggard,

e63d8c71-9f17-4b39-b75a-5b0e901c981b-1
00:11:47.034 --> 00:11:48.520
the investment isn't going to be as much.

f89c60f3-1b88-4a03-8476-4203b9f0279d-0
00:11:48.560 --> 00:11:49.880
We have to figure that out first.

d371eb01-762b-4f83-ae8d-2daee26d41ec-0
00:11:50.880 --> 00:11:53.360
So selecting the archetype on what do you
want to be?

feb947f4-9677-4b41-be65-57cce76e45c1-0
00:11:53.600 --> 00:11:59.075
Do you want to be that first out of the
gate or try something that's novel versus

feb947f4-9677-4b41-be65-57cce76e45c1-1
00:11:59.075 --> 00:12:01.880
maybe be the fast adopter or be a laggard?

12ccaaae-8f96-4598-8e95-227da888d668-0
00:12:01.880 --> 00:12:06.435
And that's just a choice to make and then
align your investments accordingly,

12ccaaae-8f96-4598-8e95-227da888d668-1
00:12:06.435 --> 00:12:07.720
but be clear about it.

9d877d47-59e5-4a54-8dde-d2d105a5ef92-0
00:12:07.840 --> 00:12:12.186
And then obviously the literacy piece
around making sure everyone's sort of

9d877d47-59e5-4a54-8dde-d2d105a5ef92-1
00:12:12.186 --> 00:12:15.160
aligned on what it can do,
cannot do as a hospital.

cf59355c-7618-4933-ad38-3ce51998366b-0
00:12:15.480 --> 00:12:17.900
And actually,
maybe I'll add one other thing is defined

cf59355c-7618-4933-ad38-3ce51998366b-1
00:12:17.900 --> 00:12:18.160
value.

2cf8f3d5-2c98-46a2-94a9-9e56382eb704-0
00:12:18.160 --> 00:12:20.294
That's something we find a lot of
hospitals don't do,

2cf8f3d5-2c98-46a2-94a9-9e56382eb704-1
00:12:20.294 --> 00:12:21.520
a lot of institutions don't do.

ccfdeb45-9cb8-4750-99c4-c48ff0834502-0
00:12:22.320 --> 00:12:24.320
And it seems like a simple thing, right?

a385ba83-3c64-4b41-957d-6e1a62e6ec58-0
00:12:24.320 --> 00:12:26.760
People will go with quintuble aims and
things like that.

79fd8e7f-3e4c-4e6e-ac46-a3cfa085e652-0
00:12:27.280 --> 00:12:28.880
I don't know what to do with the
quintuple aim.

f8699c23-b075-4a24-8da4-239670d2fc1e-0
00:12:29.680 --> 00:12:32.400
You have to be very laser focused and be
specific.

6810c6a5-7ecf-4c5a-9848-e089a3af77d3-0
00:12:33.840 --> 00:12:36.423
For example,
with our value definition at Unity,

6810c6a5-7ecf-4c5a-9848-e089a3af77d3-1
00:12:36.423 --> 00:12:40.376
we we pick things that are measurable so
we can show ROI and that's really

6810c6a5-7ecf-4c5a-9848-e089a3af77d3-2
00:12:40.376 --> 00:12:41.800
important to organizations.

8a2ac349-c2b0-4278-890d-183fc8dfc455-0
00:12:42.120 --> 00:12:44.880
So we picked 5 metrics that you can
actually count to measure.

d418b778-69a3-4fe9-a518-41d38054c59d-0
00:12:45.320 --> 00:12:48.693
We didn't pick things like patient
experience or quality of life,

d418b778-69a3-4fe9-a518-41d38054c59d-1
00:12:48.693 --> 00:12:51.606
which you should,
but I don't know how to measure it and

d418b778-69a3-4fe9-a518-41d38054c59d-2
00:12:51.606 --> 00:12:55.388
how to tell people, oh, you know,
the two point change on a short form 36

d418b778-69a3-4fe9-a518-41d38054c59d-3
00:12:55.388 --> 00:12:57.279
people don't know what to do with it.

fa1ac7bc-a41b-463a-8277-18d71d241da1-0
00:12:59.600 --> 00:13:00.160
Thanks for that.

60eef9ee-c4a9-4d0d-ab5c-57378126345b-0
00:13:00.640 --> 00:13:06.670
So maybe then I'll pass it to Devin for
this question from from your lens,

60eef9ee-c4a9-4d0d-ab5c-57378126345b-1
00:13:06.670 --> 00:13:11.976
we're seeing, you know,
executives or physician leaders sometimes

60eef9ee-c4a9-4d0d-ab5c-57378126345b-2
00:13:11.976 --> 00:13:16.560
being ended up ending up being a bit of
that risk break.

8c7664f8-9298-478d-9131-a18561b0170f-0
00:13:16.960 --> 00:13:23.312
And but what would it look like for a a
clinician champion or an executive in a,

8c7664f8-9298-478d-9131-a18561b0170f-1
00:13:23.312 --> 00:13:29.822
in a healthcare setting to actually play
offense on AI to drive safe, responsible,

8c7664f8-9298-478d-9131-a18561b0170f-2
00:13:29.822 --> 00:13:32.960
but innovate instead of slowing it down.

b6750ff7-b1b8-4a4f-bc7b-141900307323-0
00:13:33.000 --> 00:13:38.240
What is that that makes someone play
offense on AI in healthcare?

c1a5fe82-f42a-410b-98ca-c9ce15b97ca2-0
00:13:39.160 --> 00:13:41.760
Yeah, I,
I think it's such such an interesting

c1a5fe82-f42a-410b-98ca-c9ce15b97ca2-1
00:13:41.760 --> 00:13:44.360
question in the way you're you're
phrasing it.

e89495ed-b9cc-41d8-abff-c11dd5878b0a-0
00:13:44.360 --> 00:13:50.320
And I'd want to lean on the the physician
lines and that clinical aspect of things.

7bf12074-85d6-4693-ae5e-fd2542efcd26-0
00:13:50.640 --> 00:13:52.700
I mean,
a lot of the motivators to the

7bf12074-85d6-4693-ae5e-fd2542efcd26-1
00:13:52.700 --> 00:13:55.182
technologies that I try to champion and
build,

7bf12074-85d6-4693-ae5e-fd2542efcd26-2
00:13:55.182 --> 00:13:59.461
they come from real patient stories and
cases like it's a real problem that this

7bf12074-85d6-4693-ae5e-fd2542efcd26-3
00:13:59.461 --> 00:14:03.000
type of patient might be waiting too long
and we want to solve it.

ab1621ca-c73a-4cf6-a14f-44ea555d71d5-0
00:14:03.000 --> 00:14:08.009
It's not even necessarily about the AI as
it is about as an organization saying we

ab1621ca-c73a-4cf6-a14f-44ea555d71d5-1
00:14:08.009 --> 00:14:12.897
don't want to allow a patient who comes
in with an acute mental health crisis to

ab1621ca-c73a-4cf6-a14f-44ea555d71d5-2
00:14:12.897 --> 00:14:16.760
wait 10 hours for care,
that we just say that's not acceptable.

9d3a086f-59ce-46a7-a219-a0a1bf48a1a8-0
00:14:17.560 --> 00:14:20.695
And so it starts with that,
that problem space,

9d3a086f-59ce-46a7-a219-a0a1bf48a1a8-1
00:14:20.695 --> 00:14:25.464
that applying that sort of value
statement to the problem to say we want

9d3a086f-59ce-46a7-a219-a0a1bf48a1a8-2
00:14:25.464 --> 00:14:26.640
to solve for this.

25909914-d9a0-412c-ad0c-27d22961f0e2-0
00:14:27.000 --> 00:14:30.757
And that allows you to go into this
offensive mode, if you will,

25909914-d9a0-412c-ad0c-27d22961f0e2-1
00:14:30.757 --> 00:14:35.092
because it's very easy to rally people
around an organization to say, yes,

25909914-d9a0-412c-ad0c-27d22961f0e2-2
00:14:35.092 --> 00:14:37.520
here is a problem we want to solve for it.

468a9b3b-f185-49e6-8d75-0052d5166b9d-0
00:14:38.040 --> 00:14:43.331
AI then becomes one part of the solution
potentially to solving that problem

468a9b3b-f185-49e6-8d75-0052d5166b9d-1
00:14:43.331 --> 00:14:46.080
paired with integration into a workflow.

44f88d27-5da0-4c1a-a8ec-20c100b5832f-0
00:14:46.400 --> 00:14:51.108
And so I think that that's what's key,
but that's also very much easier said

44f88d27-5da0-4c1a-a8ec-20c100b5832f-1
00:14:51.108 --> 00:14:51.720
than done.

02cbdfcc-162e-42bc-9df3-dc7c936c37d4-0
00:14:51.840 --> 00:14:55.304
The journey at our organization,
Sick Kids,

02cbdfcc-162e-42bc-9df3-dc7c936c37d4-1
00:14:55.304 --> 00:14:59.320
has been over 6 to 8 years now building
solutions.

3ceaea62-9158-4bad-9532-75d59117d450-0
00:14:59.320 --> 00:15:00.560
We actually had what Muhammad said.

56a631a9-f222-4d20-8058-d647114e8988-0
00:15:00.560 --> 00:15:03.826
We had the investment, the technology,
we had the computer scientists and the

56a631a9-f222-4d20-8058-d647114e8988-1
00:15:03.826 --> 00:15:04.120
people.

c68c6bd4-4cb8-4140-9f74-27214082d0f6-0
00:15:04.760 --> 00:15:07.600
Building the AI actually became quite
easy for us.

9ceca51e-8d6c-4fb8-a14a-91077bf66595-0
00:15:07.600 --> 00:15:09.230
You know,
we could spin up new models pretty

9ceca51e-8d6c-4fb8-a14a-91077bf66595-1
00:15:09.230 --> 00:15:09.520
quickly.

8b5691c2-079e-408e-b7f2-3dc356a2ad79-0
00:15:09.840 --> 00:15:13.688
What we realized what was missing that
even if you were in this offensive mode

8b5691c2-079e-408e-b7f2-3dc356a2ad79-1
00:15:13.688 --> 00:15:17.244
and you had the institution aligned on,
let's try to solve this problem,

8b5691c2-079e-408e-b7f2-3dc356a2ad79-2
00:15:17.244 --> 00:15:19.680
we didn't have a runway to land the plane,
right?

9a8e81a4-a1f6-4f4f-af8a-691f54f7758c-0
00:15:20.280 --> 00:15:22.905
We didn't have the processes,
the policies,

9a8e81a4-a1f6-4f4f-af8a-691f54f7758c-1
00:15:22.905 --> 00:15:26.783
the procedures for health leaders and the
executive to say, yes,

9a8e81a4-a1f6-4f4f-af8a-691f54f7758c-2
00:15:26.783 --> 00:15:31.080
we feel confident that we can safely land
this AI plane to the bedside.

73cbfb70-2120-4b18-83b3-7a7ae3953d0a-0
00:15:31.360 --> 00:15:32.920
There was a void of that.

3081768b-61c4-46c1-9e70-ae9acc9ad88a-0
00:15:33.200 --> 00:15:39.240
And so and that's really what Sky Service
has filled is put in the runway in place.

a2d03691-50a1-4ae6-812f-01498c86bc27-0
00:15:39.960 --> 00:15:43.560
I love that you're sort of putting that
patient's story up front and the problem

a2d03691-50a1-4ae6-812f-01498c86bc27-1
00:15:43.560 --> 00:15:45.160
statement right in the middle of it.

b290da4e-a6e3-419e-878d-c3524dac504a-0
00:15:45.160 --> 00:15:48.400
And then if you sort of grow from there,
then you can play offense.

e100ab64-52c6-459f-9056-47aacfd0981b-0
00:15:48.400 --> 00:15:51.946
You can you,
you've got something to solve for that's

e100ab64-52c6-459f-9056-47aacfd0981b-1
00:15:51.946 --> 00:15:54.640
a lot more surgically, you know, precise.

74224339-6882-43f4-9e06-ba82e5380465-0
00:15:55.120 --> 00:15:58.829
I'll just say something really quickly is
that sometimes then in the heat of the

74224339-6882-43f4-9e06-ba82e5380465-1
00:15:58.829 --> 00:16:01.120
journey,
there's differencing different opinions.

3dee44f9-28ec-4194-a8d9-a1309357b191-0
00:16:01.320 --> 00:16:02.360
People are passionate.

587d00f0-c96b-4185-835b-166f952bb43e-0
00:16:02.560 --> 00:16:05.905
You're pulling people out of silos and
putting them into a place to work

587d00f0-c96b-4185-835b-166f952bb43e-1
00:16:05.905 --> 00:16:07.280
together with strong opinions.

3dc9c054-8421-4e6e-a037-3da252c2db9c-0
00:16:07.680 --> 00:16:13.120
Going back to the patient's story in
those moments is really powerful, right?

88b9b0ca-9b3a-4f02-b0b6-2f4779cd5dd1-0
00:16:13.440 --> 00:16:17.200
Because it really aligns people again,
to like what they're trying to solve for.

b931b5b1-ce95-4867-9ace-05b759f9f040-0
00:16:18.960 --> 00:16:24.115
So then maybe to Marsha,
mental health is is often seen as one of

b931b5b1-ce95-4867-9ace-05b759f9f040-1
00:16:24.115 --> 00:16:27.240
the hardest areas to play offense in AI.

d86c7d8d-57bb-4d1e-b23a-e8e1f7a6d967-0
00:16:28.560 --> 00:16:33.173
What has Ontario shows done differently
that puts it ahead of others trying to do

d86c7d8d-57bb-4d1e-b23a-e8e1f7a6d967-1
00:16:33.173 --> 00:16:33.680
the same?

c262b850-5744-4b8a-ad6b-d826214d5bc1-0
00:16:36.400 --> 00:16:36.720
Thanks.

30c05dbc-8e97-4939-9094-cf69ac28fdcf-0
00:16:36.760 --> 00:16:41.054
I think that there's a number of factors
that have positioned Ontario Shores well

30c05dbc-8e97-4939-9094-cf69ac28fdcf-1
00:16:41.054 --> 00:16:41.840
in this regard.

098f8660-2cc2-4e80-9a16-7672ca06b214-0
00:16:43.080 --> 00:16:46.280
I'll highlight a few that just comes to
to mind.

5aa2563e-7cef-4853-9069-6f798e8f3988-0
00:16:46.600 --> 00:16:50.986
One is our ongoing leadership and
commitment to implementation of evidence

5aa2563e-7cef-4853-9069-6f798e8f3988-1
00:16:50.986 --> 00:16:53.560
based practices and measurement based
care.

4957684f-c07c-4849-88d6-85e58a189fdf-0
00:16:53.760 --> 00:16:57.196
And I think for this,
why I think that's important is it's

4957684f-c07c-4849-88d6-85e58a189fdf-1
00:16:57.196 --> 00:17:01.680
allowed us to collect information over
time that allows us to understand the

4957684f-c07c-4849-88d6-85e58a189fdf-2
00:17:01.680 --> 00:17:06.281
services that we're delivering and the
effectiveness and the outcomes of those

4957684f-c07c-4849-88d6-85e58a189fdf-3
00:17:06.281 --> 00:17:10.998
services that will feed and power work
that we want to do to derive new insights

4957684f-c07c-4849-88d6-85e58a189fdf-4
00:17:10.998 --> 00:17:12.280
with analytics and AI.

0aa56371-3f63-4627-bef9-0e72b1580ab0-0
00:17:13.000 --> 00:17:16.708
I think that it's secondly,
our experience maturity with the advanced

0aa56371-3f63-4627-bef9-0e72b1580ab0-1
00:17:16.708 --> 00:17:18.880
use of technology and advanced analytics.

bff1a9e1-0aee-429e-b4a4-29d4ba87dcf4-0
00:17:19.200 --> 00:17:22.235
We've had a long term senior leadership
commitment,

bff1a9e1-0aee-429e-b4a4-29d4ba87dcf4-1
00:17:22.235 --> 00:17:27.080
prioritization and investment around the
use of technology and advanced analytics.

20c215c9-c66e-44b8-8fad-53efcf47d8c0-0
00:17:27.080 --> 00:17:32.847
We're very proud with our recognition
with the HEM Stage 7 in terms of our

20c215c9-c66e-44b8-8fad-53efcf47d8c0-1
00:17:32.847 --> 00:17:39.000
maturity and using the EMR both with
inpatients, outpatients and our analytics.

f0a3937d-c0ca-44ef-94bc-a92ba8f1532a-0
00:17:39.000 --> 00:17:43.560
And that's that's been part of us working
overtime many years.

2e3b53c3-5151-4ed7-a5f0-bb6414cc096b-0
00:17:43.600 --> 00:17:47.120
As the other plans have said is,
is it's a journey of building that

2e3b53c3-5151-4ed7-a5f0-bb6414cc096b-1
00:17:47.120 --> 00:17:49.760
maturity and that culture across the
organization.

b80d6a67-b2ac-4e6a-8fac-1b566c9db76d-0
00:17:50.280 --> 00:17:54.559
I think on top of that is,
is building that enhanced capability

b80d6a67-b2ac-4e6a-8fac-1b566c9db76d-1
00:17:54.559 --> 00:17:57.970
around AI,
having the structures that would enable

b80d6a67-b2ac-4e6a-8fac-1b566c9db76d-2
00:17:57.970 --> 00:18:02.517
governance, accountability,
decision making in a way that allows us

b80d6a67-b2ac-4e6a-8fac-1b566c9db76d-3
00:18:02.517 --> 00:18:03.320
to progress.

a89d13bb-a7cc-40f4-8b3b-691b46d163f6-0
00:18:03.320 --> 00:18:06.128
So we have,
we have a framework that allows us to

a89d13bb-a7cc-40f4-8b3b-691b46d163f6-1
00:18:06.128 --> 00:18:08.880
address these opportunities and to
support them.

0012c826-b0e5-4d84-96ef-b257ec27dc4f-0
00:18:09.480 --> 00:18:13.909
And then I think the final piece that for
me and I think the teams in the

0012c826-b0e5-4d84-96ef-b257ec27dc4f-1
00:18:13.909 --> 00:18:17.680
organization here at Ontario Shores is
this is purpose driven.

07c05558-b331-43b8-bd6f-6a9077987187-0
00:18:18.160 --> 00:18:22.873
So going back to the patient,
we have a bold and transforming vision

07c05558-b331-43b8-bd6f-6a9077987187-1
00:18:22.873 --> 00:18:27.040
where we are are part of that includes
revolutionizing care.

951af632-38d5-4c6a-9e67-1a020412ff7b-0
00:18:27.320 --> 00:18:33.214
It's a commitment to our patients that we
we believe that we can change care in a

951af632-38d5-4c6a-9e67-1a020412ff7b-1
00:18:33.214 --> 00:18:38.965
way that we can improve outcomes and the
experiences and that's the orientation

951af632-38d5-4c6a-9e67-1a020412ff7b-2
00:18:38.965 --> 00:18:42.560
that I think the other enablers help us
progress.

d475e9f1-6601-4172-81bc-4c2375fc9f34-0
00:18:43.240 --> 00:18:46.560
Nice sort of perfect mix of all the
ingredients coming together.

718e8e8e-ee0f-4b25-8c8c-48d10daa4ed4-0
00:18:47.680 --> 00:18:51.880
Chance you see the malpractice side of AI
as well.

2f3dd286-a9b4-43ee-9eac-2769d1e5e51b-0
00:18:51.920 --> 00:18:56.688
That's sort of one of your
responsibilities as CMPA when you're sort

2f3dd286-a9b4-43ee-9eac-2769d1e5e51b-1
00:18:56.688 --> 00:19:01.457
of interacting with physicians,
be it on your board or your members,

2f3dd286-a9b4-43ee-9eac-2769d1e5e51b-2
00:19:01.457 --> 00:19:05.880
what are you hearing about what keeps
them up that night on AI?

9a125f0e-3280-4262-a563-de40d8ed6c9c-0
00:19:05.880 --> 00:19:09.660
And,
and So what do we need to do to address

9a125f0e-3280-4262-a563-de40d8ed6c9c-1
00:19:09.660 --> 00:19:10.080
that?

5625fc0d-d669-4b18-9c1e-073f118f92aa-0
00:19:15.880 --> 00:19:18.640
So super interesting question.

4433a253-76a0-414a-b070-ad826b094c2f-0
00:19:18.640 --> 00:19:24.170
I'll preface it first with saying in
healthcare,

4433a253-76a0-414a-b070-ad826b094c2f-1
00:19:24.170 --> 00:19:30.040
there hasn't been much around AI causing
poor care.

95b48f7b-24f0-4ec8-b3f0-aad3eb13e629-0
00:19:30.600 --> 00:19:34.880
I think the adoption of AI hasn't quite
got there yet.

b6cfd72a-960f-47c3-9144-1f22be02bc4c-0
00:19:34.880 --> 00:19:39.720
So we think we're going to run into that
maybe down the road a little bit.

8798bdf4-c832-427a-862a-7958a5fb4c94-0
00:19:40.560 --> 00:19:46.767
I think that the core gap that we're,
that we're seeing around the adoption of

8798bdf4-c832-427a-862a-7958a5fb4c94-1
00:19:46.767 --> 00:19:51.639
AI is something that I think is being
spoken to a little bit,

8798bdf4-c832-427a-862a-7958a5fb4c94-2
00:19:51.639 --> 00:19:53.840
which is AI literacy, right?

fa835025-5b64-41f6-81e8-eca0b49abde6-0
00:19:54.320 --> 00:19:56.080
What, what does that actually mean?

d876ef7d-506a-44ee-a628-6a3261a2d1dc-0
00:19:57.240 --> 00:19:59.781
I,
I believe a lot of healthcare has maybe a

d876ef7d-506a-44ee-a628-6a3261a2d1dc-1
00:19:59.781 --> 00:20:02.040
bad experience with the rollout of Emrs.

9faf9718-e33e-4e83-827d-222c1e4908e1-0
00:20:02.800 --> 00:20:04.400
And so this new technology is coming.

add35e22-541f-4c07-bde2-aab1d718795d-0
00:20:04.400 --> 00:20:06.400
It's like, oh, no, here it comes again.

96657d64-2d61-497f-bf60-2e3b392b6534-0
00:20:06.920 --> 00:20:11.880
And so that seems to be a precursor of
just understanding what it is.

9a3783bc-18d5-4f36-9d6c-7f53bc984a54-0
00:20:12.400 --> 00:20:17.400
I think we're in a phase right now of
expanded hype where everything has AI in

9a3783bc-18d5-4f36-9d6c-7f53bc984a54-1
00:20:17.400 --> 00:20:18.160
front of it.

7ed4b604-7247-49d3-a279-8b3455f955f7-0
00:20:18.840 --> 00:20:23.080
I was actually sent a notification that I
can buy AI shoes.

6680d340-88c8-4342-b56b-b07d0b822f82-0
00:20:23.640 --> 00:20:24.640
I don't know what that means.

3299fdc3-ea0b-4b57-8c8c-a4315abe7693-0
00:20:25.000 --> 00:20:32.075
And so even just understanding what AI is,
does it pose risk to patients and then

3299fdc3-ea0b-4b57-8c8c-a4315abe7693-1
00:20:32.075 --> 00:20:35.440
you know how how to mitigate that risk.

308a2871-5897-41db-98e0-f4d154e77142-0
00:20:36.040 --> 00:20:39.560
What are your responsibilities as a,
as a physician or healthcare provider?

ce0a6e4a-d323-4a53-8c15-e8c2de211ffa-0
00:20:40.480 --> 00:20:44.040
And then just understanding the
additional risks that AI may pose.

d42d7077-44e7-42c0-ba58-b48d926eb51f-0
00:20:44.320 --> 00:20:47.887
And generally that's around data,
data protection,

d42d7077-44e7-42c0-ba58-b48d926eb51f-1
00:20:47.887 --> 00:20:52.225
that's around understanding what AI does
or doesn't do right,

d42d7077-44e7-42c0-ba58-b48d926eb51f-2
00:20:52.225 --> 00:20:57.122
depending on the application and what are
the limits, you know, what,

d42d7077-44e7-42c0-ba58-b48d926eb51f-3
00:20:57.122 --> 00:21:00.480
what patient populations is it
appropriate for?

e8cf87ed-7bb2-4470-87a8-202afa3de3b4-0
00:21:00.480 --> 00:21:02.000
What is it not appropriate for?

2131e3f1-5d3d-4559-862f-aa88f84973f1-0
00:21:02.000 --> 00:21:05.223
So there's this,
this piece of literacy that I think is,

2131e3f1-5d3d-4559-862f-aa88f84973f1-1
00:21:05.223 --> 00:21:06.920
is that's the current barrier.

40794d79-fe5f-41f0-9dd2-4b46e7f063d3-0
00:21:07.200 --> 00:21:12.126
And in terms of what keeps us up at night,
we know that physicians are going to be

40794d79-fe5f-41f0-9dd2-4b46e7f063d3-1
00:21:12.126 --> 00:21:16.400
sued for using AI and listening to it and
sued for not listening to AI.

108ffc94-c1e3-4abb-93bc-c49d8554ca0f-0
00:21:16.400 --> 00:21:17.680
Like we know that's going to happen.

0819311e-bacd-4dfa-9255-ca99fe003c3a-0
00:21:17.680 --> 00:21:19.200
So that's fine.

bdcc1f5c-55ec-45c9-8bae-c8ed3b02d3fa-0
00:21:19.200 --> 00:21:22.058
We're we're there to protect them,
but I think the AI literacy,

bdcc1f5c-55ec-45c9-8bae-c8ed3b02d3fa-1
00:21:22.058 --> 00:21:25.586
that's the big sort of hurdle for like
broader implementation getting a lot of

bdcc1f5c-55ec-45c9-8bae-c8ed3b02d3fa-2
00:21:25.586 --> 00:21:26.480
value in healthcare.

cf03cbc0-64a9-4ec7-849e-ac60b64945ff-0
00:21:27.680 --> 00:21:28.320
Very interesting.

c9daadd3-d240-46ae-b671-b2018662b1bf-0
00:21:30.600 --> 00:21:32.840
Maybe I'll direct this one back to you,
Devin.

17abdeb4-d541-490a-b466-18d6413ced2f-0
00:21:33.560 --> 00:21:39.200
You're both a clinician and a data leader,
AI leader, computer scientist.

3c63bc40-dcd0-4a75-96af-9f3667317364-0
00:21:39.480 --> 00:21:46.468
When does AI stop being a tool and start
becoming a a a colleague, a copilot,

3c63bc40-dcd0-4a75-96af-9f3667317364-1
00:21:46.468 --> 00:21:50.680
whatever you want to call it in the care
team?

0057c8ac-9978-4e66-86ac-7b11fc6515cc-0
00:21:51.520 --> 00:21:53.200
Are you folks thinking about that?

1c65cd0c-7faa-4635-929f-7509a79d4861-0
00:21:53.240 --> 00:21:57.904
Are you hearing out in the market on that
sort of leap that that we might be able

1c65cd0c-7faa-4635-929f-7509a79d4861-1
00:21:57.904 --> 00:21:58.360
to take?

90dc44f4-87fc-482b-935a-14bcd991a27c-0
00:21:59.000 --> 00:22:00.440
Yeah, definitely.

49056f96-07bf-424f-80ab-8cbb92664d38-0
00:22:00.440 --> 00:22:04.836
So I can tell you right now the
technology has the capacity to do that in

49056f96-07bf-424f-80ab-8cbb92664d38-1
00:22:04.836 --> 00:22:06.440
in, in certain ways, right.

48192064-252f-4172-8543-88c4d145d516-0
00:22:06.440 --> 00:22:12.965
Like you could see how you could start to
get real time data from a resuscitation

48192064-252f-4172-8543-88c4d145d516-1
00:22:12.965 --> 00:22:13.920
potentially.

61785306-f131-476e-ab81-ae59f3643562-0
00:22:14.200 --> 00:22:19.435
And just like a helpful nurse might give
you a tap on the shoulder and say, hey,

61785306-f131-476e-ab81-ae59f3643562-1
00:22:19.435 --> 00:22:24.024
you forgot to do the glucose or you
forgot to do this from a technical

61785306-f131-476e-ab81-ae59f3643562-2
00:22:24.024 --> 00:22:24.800
perspective.

4acaf4e3-5447-4952-a93a-658aa79002e4-0
00:22:25.600 --> 00:22:28.696
And, and,
and as we have the infrastructure to

4acaf4e3-5447-4952-a93a-658aa79002e4-1
00:22:28.696 --> 00:22:34.230
actually try to bring something like that
to life, we're not doing it yet, but the,

4acaf4e3-5447-4952-a93a-658aa79002e4-2
00:22:34.230 --> 00:22:38.249
the technology is there to,
to facilitate that level of care

4acaf4e3-5447-4952-a93a-658aa79002e4-3
00:22:38.249 --> 00:22:39.039
integration.

9df89d59-c81d-425d-9591-a0336147e763-0
00:22:39.800 --> 00:22:44.050
And if you think about it,
it then starts to be that AI is a member

9df89d59-c81d-425d-9591-a0336147e763-1
00:22:44.050 --> 00:22:44.800
of the team.

0995afb9-80fe-401c-b40a-9f26a6f8d7f3-0
00:22:45.320 --> 00:22:51.321
And I think then when we start to get
into this era where AI is starting to

0995afb9-80fe-401c-b40a-9f26a6f8d7f3-1
00:22:51.321 --> 00:22:54.480
truly add aspects of automation to care.

c3c5432e-a6d2-40bc-bb3b-ff385f5c353d-0
00:22:54.680 --> 00:22:57.179
For example,
we have a use case in the emergency

c3c5432e-a6d2-40bc-bb3b-ff385f5c353d-1
00:22:57.179 --> 00:23:00.647
department where right after a patient
arrives and finishes triage,

c3c5432e-a6d2-40bc-bb3b-ff385f5c353d-2
00:23:00.647 --> 00:23:03.198
if they are having an acute mental health
crisis,

c3c5432e-a6d2-40bc-bb3b-ff385f5c353d-3
00:23:03.198 --> 00:23:07.278
an AI tool is automating the consultation
to psychiatry before they've seen the

c3c5432e-a6d2-40bc-bb3b-ff385f5c353d-4
00:23:07.278 --> 00:23:07.840
emerged Dr.

9a6b7504-4d92-47b8-878e-bd5f654652ef-0
00:23:08.040 --> 00:23:11.520
It's dramatically reduced the time from
arrival to care delivery from a

9a6b7504-4d92-47b8-878e-bd5f654652ef-1
00:23:11.520 --> 00:23:12.680
psychiatric perspective.

4d5ab2f5-7fa3-42b5-ad24-e5c21ed9413b-0
00:23:13.600 --> 00:23:16.760
And in a sense there's a block of
automation there that's happening.

05a63c14-054e-4acc-804f-8359f7ea5422-0
00:23:17.160 --> 00:23:21.867
As that automation starts to increase,
we'll have to make sure that we have the

05a63c14-054e-4acc-804f-8359f7ea5422-1
00:23:21.867 --> 00:23:26.221
structures in place to critically
appraise these technologies before they

05a63c14-054e-4acc-804f-8359f7ea5422-2
00:23:26.221 --> 00:23:30.458
turn on and to ensure that the post
deployment monitoring is incredibly

05a63c14-054e-4acc-804f-8359f7ea5422-3
00:23:30.458 --> 00:23:31.400
rigorous, right?

64ea4f48-f65b-493f-a6f7-b82636557915-0
00:23:31.720 --> 00:23:35.486
That we're ensuring that the models
performance is maintained and that it's

64ea4f48-f65b-493f-a6f7-b82636557915-1
00:23:35.486 --> 00:23:38.560
it's performing equitably amongst
different subgroups, right.

51e6641f-ae24-49f4-828c-06ffa610f3c5-0
00:23:38.840 --> 00:23:41.874
But if you were to say the computer
scientists in me,

51e6641f-ae24-49f4-828c-06ffa610f3c5-1
00:23:41.874 --> 00:23:43.280
we can do that right now.

bde218b3-a644-4950-b6fa-0fc8318cd318-0
00:23:43.520 --> 00:23:44.600
That's all possible.

ca48c684-46c2-44f2-92e1-da8e68efb70a-0
00:23:44.960 --> 00:23:50.421
It's more about ensuring the ecosystem to
which we start to deploy these tools and

ca48c684-46c2-44f2-92e1-da8e68efb70a-1
00:23:50.421 --> 00:23:55.488
increase the level of automation and
impacting capacity building is ready to

ca48c684-46c2-44f2-92e1-da8e68efb70a-2
00:23:55.488 --> 00:23:58.120
do so safely, responsibly and ethically.

27813ee4-7590-425f-a27e-a0f7bcfd347c-0
00:23:58.720 --> 00:23:58.960
Yeah.

cd7f88dc-d975-483b-b43a-5a707249a0e0-0
00:23:59.400 --> 00:24:04.613
So beyond obviously you need to set up
the the data infrastructure to be able to

cd7f88dc-d975-483b-b43a-5a707249a0e0-1
00:24:04.613 --> 00:24:09.441
learn about this patient in a way that
will allow the right insights to be

cd7f88dc-d975-483b-b43a-5a707249a0e0-2
00:24:09.441 --> 00:24:12.080
surfaced by whatever your agent or AI is.

202019cd-ef61-4d27-af6b-164ae2eee011-0
00:24:12.320 --> 00:24:17.797
But then having that infrastructure,
processes and clinical workflows ready

202019cd-ef61-4d27-af6b-164ae2eee011-1
00:24:17.797 --> 00:24:20.320
for it in the right ethical domain.

0bde1c98-1682-414c-b48f-d7bf92885b60-0
00:24:20.360 --> 00:24:23.179
All that needs to come together,
which is something that we're finding now

0bde1c98-1682-414c-b48f-d7bf92885b60-1
00:24:23.179 --> 00:24:23.480
as well.

3a7c9639-3e8d-4dc6-b6e9-29eb1019ceb0-0
00:24:24.000 --> 00:24:26.400
Mohammed,
there's a question from on chat for you.

e82e4ce3-1d05-45e2-96ec-9ce7637e239d-0
00:24:27.400 --> 00:24:28.640
You mentioned AI literacy.

e0b2c4e3-2bc3-46d0-a59d-1a14f2af4b44-0
00:24:28.680 --> 00:24:29.800
Let's get a little tactical.

3fad34ea-407e-40fc-972d-956529ba4d67-0
00:24:30.000 --> 00:24:33.225
And that's really important,
as we hear from every organization,

3fad34ea-407e-40fc-972d-956529ba4d67-1
00:24:33.225 --> 00:24:36.252
right from top down,
AI literacy is important or else you're

3fad34ea-407e-40fc-972d-956529ba4d67-2
00:24:36.252 --> 00:24:37.840
sort of playing a guessing game.

504a057f-691f-4b09-b878-e35a04ff6723-0
00:24:38.120 --> 00:24:42.680
Any recommendation on how you might have
tactically achieved that within Unity?

b1a5112f-c0d6-4bc2-95ee-445cb0c9b0ab-0
00:24:42.680 --> 00:24:44.680
What might be some tactics that you can
share?

989649a8-b060-4a6b-b331-503610d27d65-0
00:24:45.520 --> 00:24:46.400
Yeah, absolutely.

fbc5e08c-6b0c-49b7-a572-062deeb40a5a-0
00:24:47.160 --> 00:24:49.680
You know,
many years ago when it wasn't such a

fbc5e08c-6b0c-49b7-a572-062deeb40a5a-1
00:24:49.680 --> 00:24:52.146
buzzword,
there are actually some pretty good

fbc5e08c-6b0c-49b7-a572-062deeb40a5a-2
00:24:52.146 --> 00:24:54.720
resources on the web,
been pretty easy to find.

7674d6d5-9749-45c6-972f-6a2579af6edd-0
00:24:54.720 --> 00:24:58.160
Now there's a lot of resources on the web,
much more than they used to be.

8b22d5ec-ba2b-4d64-bc9e-1d3febee52fb-0
00:24:58.160 --> 00:24:59.160
Some are terrific.

d837c7dc-3e81-4b52-8a20-58591e729d91-0
00:24:59.160 --> 00:25:00.560
Others, no, not so much.

df5956d9-98a8-4321-a2f9-87e841c628bc-0
00:25:01.680 --> 00:25:06.457
So how do you wade through that to learn
with good resources and incredible

df5956d9-98a8-4321-a2f9-87e841c628bc-1
00:25:06.457 --> 00:25:06.960
sources?

9e23663b-a711-44ee-8615-8ea270ac8ab1-0
00:25:07.400 --> 00:25:08.440
That's a bit of a challenge.

e6993edc-3b35-49a1-9c04-df49163c28c1-0
00:25:08.440 --> 00:25:11.899
So I know the government's making many
more investments now in terms of a

e6993edc-3b35-49a1-9c04-df49163c28c1-1
00:25:11.899 --> 00:25:12.320
literacy.

6f9d292b-f263-4a48-b198-8e3b815de15e-0
00:25:12.320 --> 00:25:16.838
So for example, at our institution,
we have the Health AI Academy as a

6f9d292b-f263-4a48-b198-8e3b815de15e-1
00:25:16.838 --> 00:25:17.920
national program.

14521b8b-25c2-4ded-8947-ce99f7f20114-0
00:25:18.840 --> 00:25:21.120
So people can actually sign up for that.

cdc7feb6-d45b-4240-8da3-35587be8b171-0
00:25:21.120 --> 00:25:24.760
It's it is limited and we've actually
just closed for this year.

6890212e-a330-46ee-a0cb-b29f59411261-0
00:25:24.760 --> 00:25:29.440
I hope it continues, but for most of us,
I'm going to put a link in the chat.

3be06984-ddab-4c87-b7bd-4142f7e05816-0
00:25:30.720 --> 00:25:35.283
This is the center we run and Devin
actually is a key leader in this center T

3be06984-ddab-4c87-b7bd-4142f7e05816-1
00:25:35.283 --> 00:25:39.320
care of the Integrity Center for AI
Research and Education Medicine.

c67611f4-ff48-4b36-995c-23b72164bace-0
00:25:40.280 --> 00:25:41.600
It is free for everybody.

a439a019-36de-43c8-bbac-88e530a61569-0
00:25:42.120 --> 00:25:45.827
We have an education section in there
where we actually have experts that

a439a019-36de-43c8-bbac-88e530a61569-1
00:25:45.827 --> 00:25:46.880
curate the materials.

fa5abcea-d4ea-4276-b836-f5591dfc47c0-0
00:25:47.360 --> 00:25:51.840
So we only put in high quality,
really good learning resources.

3ce39110-5685-4eaa-9d70-1910eecadceb-0
00:25:52.440 --> 00:25:56.789
Many of them are free, some are paid,
but you can choose and it'll go according

3ce39110-5685-4eaa-9d70-1910eecadceb-1
00:25:56.789 --> 00:25:58.040
to your learning style.

4c89252f-dd9c-4377-9911-1dc556945942-0
00:25:58.320 --> 00:26:00.400
Their videos, their articles you can read.

123bb748-827e-4d26-a42d-364cd9534221-0
00:26:00.920 --> 00:26:03.549
There are podcasts,
we have speaker series,

123bb748-827e-4d26-a42d-364cd9534221-1
00:26:03.549 --> 00:26:06.000
There's tons of learning materials there.

9875d94c-a04b-443c-9062-a169d42d3d7c-0
00:26:08.320 --> 00:26:09.560
Great, thanks for sharing that.

93dac9a4-e0d1-4a42-9164-a0af8368d52f-0
00:26:10.240 --> 00:26:11.920
Another question on chat.

17d4c6f0-476d-49a7-9739-67845936bf69-0
00:26:12.280 --> 00:26:15.120
Maybe chance we'll take go go with you on
this one.

2242333e-742a-4f2a-8c55-d5da2671eb6f-0
00:26:16.080 --> 00:26:21.685
Our healthcare system right now is,
is operating with far limited resources,

2242333e-742a-4f2a-8c55-d5da2671eb6f-1
00:26:21.685 --> 00:26:23.360
especially financially.

806c893d-a14d-4dcf-8558-23f021613c2c-0
00:26:23.960 --> 00:26:26.840
And you know a lot of that.

6a09bb29-6b23-432b-831b-cbd5cc0262ba-0
00:26:26.840 --> 00:26:33.089
This just requires funding to try things
out sometimes and and a bit of ROI that

6a09bb29-6b23-432b-831b-cbd5cc0262ba-1
00:26:33.089 --> 00:26:39.031
may be a bit of a gas or you know,
probability based ROI and you need to try

6a09bb29-6b23-432b-831b-cbd5cc0262ba-2
00:26:39.031 --> 00:26:39.880
things out.

d56ab47f-d463-4431-9e07-4ad2517658c9-0
00:26:41.160 --> 00:26:46.873
What are some sort of lessons in in
ensuring that resources are invested

d56ab47f-d463-4431-9e07-4ad2517658c9-1
00:26:46.873 --> 00:26:48.360
wisely, not wasted?

49be8335-9727-4d3d-b0aa-2bc22f3d07a4-0
00:26:49.920 --> 00:26:51.640
How are you thinking about that?

5cb300dc-ba8f-4c74-bb26-6b4659b665b6-0
00:26:51.640 --> 00:26:59.800
Sorry that that was to me, yes.

8c40217b-daf2-4497-a13e-e6e76b4e7c0d-0
00:27:02.160 --> 00:27:06.524
So I think flipping now to like how we
use AI internally,

8c40217b-daf2-4497-a13e-e6e76b4e7c0d-1
00:27:06.524 --> 00:27:11.866
what we did was we did a number of pilots,
we got them off the ground,

8c40217b-daf2-4497-a13e-e6e76b4e7c0d-2
00:27:11.866 --> 00:27:16.380
but the proof of concept on the technical
side was working,

8c40217b-daf2-4497-a13e-e6e76b4e7c0d-3
00:27:16.380 --> 00:27:19.239
but we couldn't get to implementation.

cc11d89a-6e47-4422-8b65-52e8860ab1fc-0
00:27:20.560 --> 00:27:25.856
And what we found was for us,
it didn't match with the culture of our

cc11d89a-6e47-4422-8b65-52e8860ab1fc-1
00:27:25.856 --> 00:27:26.840
organization.

ccef90d7-49c0-49bc-9dfc-bf4d28a79f1c-0
00:27:27.000 --> 00:27:30.520
And this is where it can get kind of
difficult.

fea90e49-484c-4d79-af57-b717b59f6421-0
00:27:30.560 --> 00:27:34.346
I think one of the parts is you have ROI,
and maybe that's efficiency,

fea90e49-484c-4d79-af57-b717b59f6421-1
00:27:34.346 --> 00:27:36.800
maybe that's cost,
maybe that's patient care.

1b55f161-b61c-430e-988f-88675c7cce7c-0
00:27:37.560 --> 00:27:41.538
And I'm fairly certain for every
organization, even on this panel,

1b55f161-b61c-430e-988f-88675c7cce7c-1
00:27:41.538 --> 00:27:45.160
what is most important to the
organization actually changes.

1e793593-8085-4c9c-844b-01335d667ebf-0
00:27:46.560 --> 00:27:52.314
So my advice would be figure out you get
a laundry list of the potential benefits

1e793593-8085-4c9c-844b-01335d667ebf-1
00:27:52.314 --> 00:27:54.560
of any technology, including AI.

2e5662b7-2e7b-4083-a1fa-f33f6df52a2e-0
00:27:54.920 --> 00:27:58.687
Find the ones that actually matter to you
because it's not going to be the whole

2e5662b7-2e7b-4083-a1fa-f33f6df52a2e-1
00:27:58.687 --> 00:27:58.920
list.

566b9855-fa86-4102-ae8e-033e20515c23-0
00:27:59.400 --> 00:28:03.728
Pick the one or two and make sure that
that's the part that you really hone in

566b9855-fa86-4102-ae8e-033e20515c23-1
00:28:03.728 --> 00:28:07.838
on because it'll really help you to
identify which ones are going to have,

566b9855-fa86-4102-ae8e-033e20515c23-2
00:28:07.838 --> 00:28:10.578
you know,
are going to be able to actually thrive

566b9855-fa86-4102-ae8e-033e20515c23-3
00:28:10.578 --> 00:28:12.880
in your environment and which ones aren't.

57120841-8174-4046-ab17-32ebf6258150-0
00:28:13.200 --> 00:28:17.173
And we actually went down that path where
we thought a number of different pilots

57120841-8174-4046-ab17-32ebf6258150-1
00:28:17.173 --> 00:28:18.240
were going to succeed.

e0aa72ed-d245-4a20-9c50-28c1e9e1af4f-0
00:28:18.600 --> 00:28:20.843
And what we found is when we tried to get
to implementation,

e0aa72ed-d245-4a20-9c50-28c1e9e1af4f-1
00:28:20.843 --> 00:28:22.720
they didn't because they didn't match the
culture.

44f52023-0c9b-4b8a-bfb7-b03d461e9d7d-0
00:28:23.840 --> 00:28:28.459
And so interestingly for us trying to
find that right ROI metric that fits the

44f52023-0c9b-4b8a-bfb7-b03d461e9d7d-1
00:28:28.459 --> 00:28:31.500
culture is,
is actually a little bit of a challenge

44f52023-0c9b-4b8a-bfb7-b03d461e9d7d-2
00:28:31.500 --> 00:28:35.360
and takes a little bit of work and a fair
amount of consultation.

b5242396-9dd2-43ff-81e6-5eb9a7b82c2c-0
00:28:35.360 --> 00:28:38.892
But I think once you get that right,
it does make the path to getting

b5242396-9dd2-43ff-81e6-5eb9a7b82c2c-1
00:28:38.892 --> 00:28:41.920
different technologies implemented work
quite a bit better.

2682fe44-4ebd-4d1d-bba0-3539f5b1d915-0
00:28:42.160 --> 00:28:44.600
So I think,
I think there's no cookie cutter approach.

bbfe5d40-f870-4986-9b8b-18b4c1b358ef-0
00:28:44.600 --> 00:28:50.440
You really have to for your organization
find what is the are the return you want.

11c56600-414f-4a51-bf64-f2d4fc5597e7-0
00:28:50.680 --> 00:28:52.840
And I think it differs by different
organizations.

3b69c503-e239-42d2-9419-3b77c9172a42-0
00:28:52.840 --> 00:28:58.446
So that'd be give an example of what can
you give an example of what CMP as ROI

3b69c503-e239-42d2-9419-3b77c9172a42-1
00:28:58.446 --> 00:29:01.880
metric might be that fits in the culture
of CMP?

bc0ce9cf-e7f2-4ce0-91aa-aac4f44a3273-0
00:29:02.480 --> 00:29:02.880
Sure.

33f929ef-3807-4788-9132-1c8e68fa6004-0
00:29:02.880 --> 00:29:05.963
I mean, so in looking at our at our,
our piece, we're,

33f929ef-3807-4788-9132-1c8e68fa6004-1
00:29:05.963 --> 00:29:09.440
we're very interested in providing better
service to members.

1d2afa58-954a-4b3f-b676-a938052b3e25-0
00:29:09.800 --> 00:29:14.396
So does the AI actually improve the
physician experience when they interact

1d2afa58-954a-4b3f-b676-a938052b3e25-1
00:29:14.396 --> 00:29:14.880
with us?

e01c7d9a-9126-478f-b1ed-46ea4aa6d6cc-0
00:29:14.880 --> 00:29:15.440
That's one.

3a7f72a7-ee4d-4c01-9e6f-d54ee38489ff-0
00:29:16.120 --> 00:29:21.327
The other one that really resonated is we
know that physicians are facing a lot of

3a7f72a7-ee4d-4c01-9e6f-d54ee38489ff-1
00:29:21.327 --> 00:29:22.520
technology changes.

a5a6a0c6-8d15-4b77-902e-ac732a6bb84b-0
00:29:22.520 --> 00:29:25.520
And so we looked at pilots that actually
mirror that.

7856e4f4-481b-4074-8efc-b6860a5930f9-0
00:29:25.520 --> 00:29:28.840
So help us understand the technologies
that are being used.

3516593d-1b68-4727-b997-03fbf0b068d4-0
00:29:28.840 --> 00:29:32.800
So Devin talked a little bit about
clinical decision support tools, right?

f6bf97c4-a600-4c0c-b6e4-c1e5c5ff8747-0
00:29:33.040 --> 00:29:36.685
So what can we have on our end that kind
of helps us understand how these

f6bf97c4-a600-4c0c-b6e4-c1e5c5ff8747-1
00:29:36.685 --> 00:29:40.429
technologies work so that when we provide
advice back to physicians, again,

f6bf97c4-a600-4c0c-b6e4-c1e5c5ff8747-2
00:29:40.429 --> 00:29:41.760
it's better member service.

3d9b961b-28fb-4b60-91ad-d49acae9e5ae-0
00:29:42.120 --> 00:29:47.760
So for us, that was super important,
maybe less so on productivity, right?

35ef42d4-6361-495d-99d5-f7cbbe8c6ca4-0
00:29:47.760 --> 00:29:50.195
Like, you know,
that's another metric you could look at,

35ef42d4-6361-495d-99d5-f7cbbe8c6ca4-1
00:29:50.195 --> 00:29:52.460
but for us,
it was really all about providing better

35ef42d4-6361-495d-99d5-f7cbbe8c6ca4-2
00:29:52.460 --> 00:29:53.400
service to physicians.

18aefed3-81ba-407e-8482-9473999a0e54-0
00:29:53.400 --> 00:29:55.760
So for us,
that was the return we're looking for.

a4ff80d7-af4e-43f0-9343-7be702b1d8a4-0
00:29:56.000 --> 00:29:59.259
Maybe a little harder to measure,
maybe a little bit more of a qualitative

a4ff80d7-af4e-43f0-9343-7be702b1d8a4-1
00:29:59.259 --> 00:29:59.520
story.

20967181-394a-4866-bbee-627dabd70c9f-0
00:29:59.760 --> 00:30:03.342
But that's the thing that we've found
really kind of helps us get AI

20967181-394a-4866-bbee-627dabd70c9f-1
00:30:03.342 --> 00:30:07.080
initiatives from proof of concept into
something we can operationalize.

001ec418-a7da-43d6-898e-2c5ff0b192ba-0
00:30:10.840 --> 00:30:13.592
Marsha,
going back to the laggards and the

001ec418-a7da-43d6-898e-2c5ff0b192ba-1
00:30:13.592 --> 00:30:15.640
leaders question, is it So what?

7b3ea977-c151-443b-ad09-02d553e3c7fd-0
00:30:15.640 --> 00:30:19.170
What makes an organization or I guess you
know,

7b3ea977-c151-443b-ad09-02d553e3c7fd-1
00:30:19.170 --> 00:30:24.319
what's the biggest reason some
organizations remain those AI laggards

7b3ea977-c151-443b-ad09-02d553e3c7fd-2
00:30:24.319 --> 00:30:26.600
and sort of take a lot of time?

27ab5fad-fe27-4af9-9646-8ade1a3b9294-0
00:30:27.160 --> 00:30:31.952
Maybe it's a choice they've made or maybe
that's something that they're struggling

27ab5fad-fe27-4af9-9646-8ade1a3b9294-1
00:30:31.952 --> 00:30:34.320
with and we can go to share data quality.

0d8a1e10-606c-423f-af3a-56a82f06fb22-0
00:30:35.360 --> 00:30:36.800
Yes, we don't have that enough.

e6823e14-f5d9-40c0-ba4c-8d868341d393-0
00:30:37.080 --> 00:30:40.548
Maybe that's one of the reason,
but maybe that's a little bit too

e6823e14-f5d9-40c0-ba4c-8d868341d393-1
00:30:40.548 --> 00:30:42.440
granular and you can solve for that.

0b0fcda8-ba19-4fb1-82f4-005f2029cdd0-0
00:30:42.440 --> 00:30:47.096
Or is it simply lack of conviction that
we're not sure if this is something we

0b0fcda8-ba19-4fb1-82f4-005f2029cdd0-1
00:30:47.096 --> 00:30:48.040
should be doing?

d53ee02c-fd7b-4dde-a17f-201ce0568ea4-0
00:30:50.640 --> 00:30:51.680
That's a really good question.

db1699ee-1871-439d-ad70-a7d5b6c54e3c-0
00:30:51.680 --> 00:30:56.731
I think that the the pace at which
organizations will embark on and achieve

db1699ee-1871-439d-ad70-a7d5b6c54e3c-1
00:30:56.731 --> 00:31:01.517
value from the use of AI systems,
I think the biggest driver on that is

db1699ee-1871-439d-ad70-a7d5b6c54e3c-2
00:31:01.517 --> 00:31:05.240
going to be their organizational
readiness and culture.

2bb14efc-da8f-43d8-8a7d-f63c01b75162-0
00:31:06.400 --> 00:31:08.960
I I think within that is the piece that
you mentioned.

b56efa31-973f-42e4-9169-1c6a37cc5ec6-0
00:31:09.200 --> 00:31:14.863
You need to have that strong foundation
of data stewardship and you need to have

b56efa31-973f-42e4-9169-1c6a37cc5ec6-1
00:31:14.863 --> 00:31:20.316
that capability across your organization
in order to effectively identify the

b56efa31-973f-42e4-9169-1c6a37cc5ec6-2
00:31:20.316 --> 00:31:24.161
opportunities,
the technology and that the application

b56efa31-973f-42e4-9169-1c6a37cc5ec6-3
00:31:24.161 --> 00:31:25.559
within your context.

a96132f7-6d87-4f9a-bb71-9dedcdb112f1-0
00:31:25.560 --> 00:31:30.966
So even if you as an organization decide
that you're not going to be involved in

a96132f7-6d87-4f9a-bb71-9dedcdb112f1-1
00:31:30.966 --> 00:31:35.638
the development of AI systems,
you still need that capability and and

a96132f7-6d87-4f9a-bb71-9dedcdb112f1-2
00:31:35.638 --> 00:31:40.777
capacity in order to be able to identify
the opportunities to understand the

a96132f7-6d87-4f9a-bb71-9dedcdb112f1-3
00:31:40.777 --> 00:31:46.183
context in which the the solutions that
are being presented or being proposed to

a96132f7-6d87-4f9a-bb71-9dedcdb112f1-4
00:31:46.183 --> 00:31:48.919
be developed for you are being developed.

7b4b1345-2e17-4f7b-bbf6-31ce9d307852-0
00:31:49.520 --> 00:31:53.653
Have they been developed with the
appropriate data that's representative of

7b4b1345-2e17-4f7b-bbf6-31ce9d307852-1
00:31:53.653 --> 00:31:54.360
your context?

ef52e16f-5b46-4c71-9930-b418d4403940-0
00:31:54.360 --> 00:31:57.040
I know I can speak from a mental health
context.

e2c2ad99-b098-4898-90c8-df2dc89590c5-0
00:31:57.040 --> 00:32:00.585
I look at different opportunities with
products and you know,

e2c2ad99-b098-4898-90c8-df2dc89590c5-1
00:32:00.585 --> 00:32:04.360
they may not have been trained in mental
health service delivery.

c1809d75-9d62-46aa-bc3b-471f556bc7aa-0
00:32:05.040 --> 00:32:08.624
It may not have been in specialized
mental health delivery,

c1809d75-9d62-46aa-bc3b-471f556bc7aa-1
00:32:08.624 --> 00:32:12.924
often not with data In Canada,
we're situated in Ontarios and often not

c1809d75-9d62-46aa-bc3b-471f556bc7aa-2
00:32:12.924 --> 00:32:14.000
with Ontario data.

eaa0658c-3d05-4066-a509-a7f02f7d29ad-0
00:32:14.880 --> 00:32:21.346
And so I think it's the having the
ability to be able to situate that to to

eaa0658c-3d05-4066-a509-a7f02f7d29ad-1
00:32:21.346 --> 00:32:28.324
identify how to test that appropriately
in your context in order to see where the

eaa0658c-3d05-4066-a509-a7f02f7d29ad-2
00:32:28.324 --> 00:32:32.919
best value is,
I think will be one of the the the key

eaa0658c-3d05-4066-a509-a7f02f7d29ad-3
00:32:32.919 --> 00:32:33.600
drivers.

b3cce2ee-b0ad-4c40-9c8a-095047bbadc3-0
00:32:35.880 --> 00:32:39.160
Devin,
I'm going to ask you a very specific

b3cce2ee-b0ad-4c40-9c8a-095047bbadc3-1
00:32:39.160 --> 00:32:43.560
question around something that feels very
promising on AI.

22e57970-88ae-45da-9815-2df8b3695819-0
00:32:44.160 --> 00:32:47.040
Sickens is known for its research and
analytics.

9a11680e-beaa-4567-ba69-8212f8d03719-0
00:32:47.040 --> 00:32:51.765
Broadly,
how far are we from AI that truly learns

9a11680e-beaa-4567-ba69-8212f8d03719-1
00:32:51.765 --> 00:32:59.516
from every child's experience and makes
it a real time learning health system for

9a11680e-beaa-4567-ba69-8212f8d03719-2
00:32:59.516 --> 00:32:59.800
us?

f7740100-c090-441d-8962-609e99a0b4d5-0
00:33:00.040 --> 00:33:03.841
We struggle with this concept of a
learning health system and truly,

f7740100-c090-441d-8962-609e99a0b4d5-1
00:33:03.841 --> 00:33:06.650
you know,
bringing those outcomes that we realized

f7740100-c090-441d-8962-609e99a0b4d5-2
00:33:06.650 --> 00:33:11.113
back into the data and then learning from
the data and then applying to clinical

f7740100-c090-441d-8962-609e99a0b4d5-3
00:33:11.113 --> 00:33:12.160
practice, etcetera.

6b141466-7453-4881-8b0d-4122c9866128-0
00:33:13.600 --> 00:33:17.889
Can AI when you think about HNTKI
specifically where it sort of learns by

6b141466-7453-4881-8b0d-4122c9866128-1
00:33:17.889 --> 00:33:21.600
itself and it's performance and action
and feeds back the data.

e5dbde78-56b9-42b7-8827-f010e573cc9a-0
00:33:22.000 --> 00:33:25.400
Is there a seed there for us to
transition to a learning health system?

b5b46ab6-6fcb-4962-860b-b6bf8165fc83-0
00:33:26.280 --> 00:33:29.520
Yeah,
I'll say that on the 2030 strategic

b5b46ab6-6fcb-4962-860b-b6bf8165fc83-1
00:33:29.520 --> 00:33:36.000
priority for sick kids precision medicine
is is underscored again and again and and

b5b46ab6-6fcb-4962-860b-b6bf8165fc83-2
00:33:36.000 --> 00:33:38.160
a commitment and investment.

40399d18-d6f3-45a9-942a-90997a4d51ec-0
00:33:38.160 --> 00:33:41.717
I think it was Muhammad saying, you know,
making that commitment and that

40399d18-d6f3-45a9-942a-90997a4d51ec-1
00:33:41.717 --> 00:33:43.160
investment as an organization.

702a944f-f0c0-4bca-99c3-ff22bf51ea5c-0
00:33:43.840 --> 00:33:48.031
That's exactly what sick kids is doing
and trying to think through how to not do

702a944f-f0c0-4bca-99c3-ff22bf51ea5c-1
00:33:48.031 --> 00:33:51.085
it only for sick kids,
but to then figure this out and and

702a944f-f0c0-4bca-99c3-ff22bf51ea5c-2
00:33:51.085 --> 00:33:52.120
disseminate broadly.

d7570794-db56-4848-b43e-f9e8721fee31-0
00:33:53.000 --> 00:33:57.259
What you're describing sounds really
complicated because it's quite broad,

d7570794-db56-4848-b43e-f9e8721fee31-1
00:33:57.259 --> 00:33:57.600
right.

1e7ff737-d65f-4ce8-ae9d-d4832e8ec9c2-0
00:33:57.600 --> 00:34:00.274
Anytime I hear the word learning how the
system,

1e7ff737-d65f-4ce8-ae9d-d4832e8ec9c2-1
00:34:00.274 --> 00:34:03.440
it almost just feels like overwhelming
boiling the ocean.

371b3095-4886-44cd-8904-b3b1b5ce2536-0
00:34:03.760 --> 00:34:07.705
And I actually internally at sick kids
when we try to solve these problems,

371b3095-4886-44cd-8904-b3b1b5ce2536-1
00:34:07.705 --> 00:34:11.287
it's it's exactly that like,
let's try and take that concept of what

371b3095-4886-44cd-8904-b3b1b5ce2536-2
00:34:11.287 --> 00:34:15.441
you just described and apply it to like a
meaningful problem where that type of

371b3095-4886-44cd-8904-b3b1b5ce2536-3
00:34:15.441 --> 00:34:19.594
learning health system set up and that
type of agentic learning from your chart

371b3095-4886-44cd-8904-b3b1b5ce2536-4
00:34:19.594 --> 00:34:22.917
and then reassessing and,
and applying aspects of automake here

371b3095-4886-44cd-8904-b3b1b5ce2536-5
00:34:22.917 --> 00:34:23.800
actually matters.

a3ecc7a0-71f2-462f-9e12-702a38f3a626-0
00:34:23.800 --> 00:34:26.860
We're we're it's not just technology for
technology,

a3ecc7a0-71f2-462f-9e12-702a38f3a626-1
00:34:26.860 --> 00:34:29.920
but this specific nuance use cases where
it matters.

de34536c-2813-448f-8727-00434707dcaa-0
00:34:30.400 --> 00:34:31.280
This is going to sound.

d5e59591-68ba-4bc4-8e0c-bb2f184a4c15-0
00:34:31.760 --> 00:34:36.560
It's not a precision medicine use case,
but it's an important one for us to learn.

714eab34-2f96-44f8-b3ba-748ae6ad8d8d-0
00:34:37.040 --> 00:34:40.400
And it's around predicting patient census
in the EED.

2c245229-40be-49a4-9ad7-1cb5af4e49aa-0
00:34:40.400 --> 00:34:41.200
Super boring.

31657146-a0a8-4bc2-b528-06e4006bdd60-0
00:34:41.200 --> 00:34:42.120
Everyone's done that.

77316ffd-0c17-4640-8133-b832d9b5bb79-0
00:34:42.400 --> 00:34:43.880
But here's how we've set it up.

7db6cdb3-d4ed-4f67-90fd-a065e7876a07-0
00:34:43.880 --> 00:34:47.632
We've totally over engineered the problem
because we're anticipating learning to do

7db6cdb3-d4ed-4f67-90fd-a065e7876a07-1
00:34:47.632 --> 00:34:49.240
it in a way that you just described.

428eddd2-66a6-42fc-af42-bcb1ff48446d-0
00:34:49.520 --> 00:34:52.720
But this model retrains every day, right?

3389896c-98db-4975-bff9-b554bba42692-0
00:34:53.000 --> 00:34:54.760
It then self selects.

31717701-7d39-41cd-b070-76cbcef38bbe-0
00:34:54.760 --> 00:34:59.360
It's the best version of itself based on
certain evaluation metrics.

cf173c76-7d43-4bc1-b145-09c440c528ea-0
00:34:59.720 --> 00:35:04.330
It then is engineered to self deploy
itself into practice, make predictions,

cf173c76-7d43-4bc1-b145-09c440c528ea-1
00:35:04.330 --> 00:35:09.241
audit itself from an ML OPS perspective,
and then iterate through the cycle again

cf173c76-7d43-4bc1-b145-09c440c528ea-2
00:35:09.241 --> 00:35:09.840
and again.

ba4d48c1-7b9e-4d1b-8079-a0ed21847ba0-0
00:35:10.080 --> 00:35:12.400
I know that that sounds, you know,
Oh well,

ba4d48c1-7b9e-4d1b-8079-a0ed21847ba0-1
00:35:12.400 --> 00:35:14.720
you're doing that for predicting Ed
census.

a23d198e-aa54-463b-9b98-118d8bf460e2-0
00:35:15.320 --> 00:35:19.075
But the reason why we've chose such a
banana use case to dive into such deep

a23d198e-aa54-463b-9b98-118d8bf460e2-1
00:35:19.075 --> 00:35:22.440
and sophisticated learning engineering
because it was safe to do so.

e0a76370-6671-43ff-a0c6-bc3811c7c523-0
00:35:23.040 --> 00:35:28.016
And we see then this being the way in
which we can scale these types of

e0a76370-6671-43ff-a0c6-bc3811c7c523-1
00:35:28.016 --> 00:35:33.200
methodologies to more patient specific
precision medicine based use cases.

ad66a3c5-a222-4fbb-b9be-56ee93b86bcf-0
00:35:33.200 --> 00:35:35.600
And a key_to that was sustainability.

fca3be5c-279e-43fa-80d3-85fc12a7cde7-0
00:35:35.840 --> 00:35:39.262
Like it's going to be really hard to have
all these humans manually supervising all

fca3be5c-279e-43fa-80d3-85fc12a7cde7-1
00:35:39.262 --> 00:35:40.200
of these things, right?

5e1bafc6-b62e-460f-97e7-b39d97b8a174-0
00:35:40.200 --> 00:35:46.240
And so benign, simple use case,
but incredible technology behind it.

f77a3d1c-e715-4314-9838-74c635077579-0
00:35:46.240 --> 00:35:51.187
Then as we evolve over these next five
years towards our pursuit of precision

f77a3d1c-e715-4314-9838-74c635077579-1
00:35:51.187 --> 00:35:54.930
medicine at Sick Kids,
we will start to increasingly apply

f77a3d1c-e715-4314-9838-74c635077579-2
00:35:54.930 --> 00:35:56.960
exactly what you just said, Raj.

989e4011-b894-4f12-b115-e091aac99b39-0
00:35:56.960 --> 00:35:59.847
And and again,
I can't_the importance of the policy that

989e4011-b894-4f12-b115-e091aac99b39-1
00:35:59.847 --> 00:36:02.278
governance,
the procedures like no way could we

989e4011-b894-4f12-b115-e091aac99b39-2
00:36:02.278 --> 00:36:06.279
entertain even tackling what you just
described without those pieces in place.

118afa4b-c2ee-454f-84d2-f5225994d3ca-0
00:36:07.000 --> 00:36:08.600
And we're proud to have them in place now.

748bbe03-8c22-4a84-bd41-b9bce69ac301-0
00:36:11.040 --> 00:36:12.800
Great Mohammed, you talked about value.

be8e0917-ecb2-4415-897b-73ff9fe32703-0
00:36:13.600 --> 00:36:15.520
And then there's the measurement of the
value.

cb1e6963-59dd-4b4d-8316-695de66d3cc9-0
00:36:15.520 --> 00:36:20.777
This is important and sometimes it's it's
it's it's not clear and we need to be

cb1e6963-59dd-4b4d-8316-695de66d3cc9-1
00:36:20.777 --> 00:36:24.720
clear about how value is measured and
then that measure it.

532ba6d7-dc49-4671-9bc5-8ebadcdb8304-0
00:36:25.240 --> 00:36:31.172
But then is that enough to just measure
it and what do we do to communicate that

532ba6d7-dc49-4671-9bc5-8ebadcdb8304-1
00:36:31.172 --> 00:36:36.298
and show that and be able to get adoption
from it and buy in from it,

532ba6d7-dc49-4671-9bc5-8ebadcdb8304-2
00:36:36.298 --> 00:36:40.399
more investment from it and perhaps
having more change.

85e5bcc5-977b-4feb-8b9c-6038a76373e6-0
00:36:41.480 --> 00:36:43.080
Let's say we were able to measure it.

7ca55b7a-46aa-4571-a711-4a83ecf9346d-0
00:36:43.960 --> 00:36:46.240
Is that enough or how do where do we go
from there?

8b8cc4f4-ee8e-4753-8d32-43fdf0500eaa-0
00:36:47.440 --> 00:36:47.600
Yeah.

53999afe-647a-4677-8a22-b9b889d923fd-0
00:36:47.600 --> 00:36:51.891
I think the,
the first step is to identify kind of who

53999afe-647a-4677-8a22-b9b889d923fd-1
00:36:51.891 --> 00:36:55.480
your key funding sources to be honest,
right.

1b92886a-2807-47a7-ae4d-493b045312b5-0
00:36:56.400 --> 00:36:59.326
If it's the organization and the senior
executive,

1b92886a-2807-47a7-ae4d-493b045312b5-1
00:36:59.326 --> 00:37:03.974
it's really important to understand what
do they value because at the end of the

1b92886a-2807-47a7-ae4d-493b045312b5-2
00:37:03.974 --> 00:37:06.614
day,
that's where the sustainability is going

1b92886a-2807-47a7-ae4d-493b045312b5-3
00:37:06.614 --> 00:37:07.360
to come from.

b9bb1fbc-9c35-414e-9f01-d49bc2269ac1-0
00:37:07.520 --> 00:37:09.160
That's where your resources are going to
come from.

aab4babf-f954-402a-99c7-42a6b4bdcb05-0
00:37:10.360 --> 00:37:15.208
So if they tell you like this day or
patient throughput or flow is really

aab4babf-f954-402a-99c7-42a6b4bdcb05-1
00:37:15.208 --> 00:37:20.318
important to us and we're willing to pay
for that, it then begs the question,

aab4babf-f954-402a-99c7-42a6b4bdcb05-2
00:37:20.318 --> 00:37:21.760
how do I measure flow?

0ea40281-1217-4393-bdc4-023aaa0d983d-0
00:37:21.760 --> 00:37:23.960
How do I measure throughput in a
meaningful way?

d1541753-d3f4-4131-9f9e-da5fa5260cd4-0
00:37:24.200 --> 00:37:26.280
But you'll say, yeah, here's some dollars.

816a35fa-dd0a-4558-ab2b-056a395ef4d4-0
00:37:27.680 --> 00:37:31.880
But if your organization says I'll pay
for mortality reductions, great.

eed99602-030e-448b-817c-6dda464f3674-0
00:37:32.640 --> 00:37:33.240
How much?

a477e716-8b3f-4d90-88c3-9b86cdde7559-0
00:37:34.560 --> 00:37:37.181
If they say,
actually I want to see hard cost

a477e716-8b3f-4d90-88c3-9b86cdde7559-1
00:37:37.181 --> 00:37:40.202
reductions,
that's a lot easier because then you can

a477e716-8b3f-4d90-88c3-9b86cdde7559-2
00:37:40.202 --> 00:37:44.819
say, all right, I saved you 10 million,
you're giving me 5 net net you're you're

a477e716-8b3f-4d90-88c3-9b86cdde7559-3
00:37:44.819 --> 00:37:45.560
up 5 billion.

8d8392c3-2a18-4ebf-bd95-baab7133ed62-0
00:37:46.520 --> 00:37:50.853
So understanding that kind of dynamic
around first of all,

8d8392c3-2a18-4ebf-bd95-baab7133ed62-1
00:37:50.853 --> 00:37:56.802
what is value to the people who actually
have the resources becomes important #2

8d8392c3-2a18-4ebf-bd95-baab7133ed62-2
00:37:56.802 --> 00:38:02.899
how do you develop solutions around that
definition of value #3 how you execute on

8d8392c3-2a18-4ebf-bd95-baab7133ed62-3
00:38:02.899 --> 00:38:07.600
those solutions to then be able to show
that level of of value.

de737a19-622b-4628-b507-6b69f9974691-0
00:38:08.160 --> 00:38:11.459
So it's it's not only understanding what
metric it is by,

de737a19-622b-4628-b507-6b69f9974691-1
00:38:11.459 --> 00:38:13.280
by how much you need to improve.

84b1e7e4-14cb-4c69-b99f-2cee5defe507-0
00:38:14.080 --> 00:38:17.588
So for example,
if somebody said I value simply cost

84b1e7e4-14cb-4c69-b99f-2cee5defe507-1
00:38:17.588 --> 00:38:20.832
reductions, well,
if you're going to pay $5,000,

84b1e7e4-14cb-4c69-b99f-2cee5defe507-2
00:38:20.832 --> 00:38:25.400
000 for a solution that saves you $20,
that's not a good investment.

ec6deec3-c5e8-441a-8845-77ab6830ad68-0
00:38:26.720 --> 00:38:28.680
That magnitude makes a difference as well.

51452a97-71bc-431a-a6b7-ba8c8ca43a7e-0
00:38:32.080 --> 00:38:34.191
Yeah,
that helps a lot in terms of just

51452a97-71bc-431a-a6b7-ba8c8ca43a7e-1
00:38:34.191 --> 00:38:38.097
aligning the right metric and, and,
and making sure that that that metric

51452a97-71bc-431a-a6b7-ba8c8ca43a7e-2
00:38:38.097 --> 00:38:38.520
matters.

8e5c746e-3ecf-467c-897f-efa02e37f595-0
00:38:39.800 --> 00:38:45.062
So just then maybe shifting gears a
little bit into generative AI

8e5c746e-3ecf-467c-897f-efa02e37f595-1
00:38:45.062 --> 00:38:51.600
specifically and we think about searching,
summarizing and being able to use Gen.

dfa036f9-376d-4212-8cad-9204fc416469-0
00:38:51.600 --> 00:38:55.610
AI for,
for use cases within healthcare AI use

dfa036f9-376d-4212-8cad-9204fc416469-1
00:38:55.610 --> 00:39:01.840
sort of an open question for anyone out
there for for all the panelists.

e04b31ae-61ce-4bc3-97d5-407c1cba10b2-0
00:39:02.880 --> 00:39:09.294
Where have you seen generative AI
specifically being applied using large

e04b31ae-61ce-4bc3-97d5-407c1cba10b2-1
00:39:09.294 --> 00:39:14.303
language models for search, summarization,
other aspect,

e04b31ae-61ce-4bc3-97d5-407c1cba10b2-2
00:39:14.303 --> 00:39:19.840
other areas that you'd think we should be
paying attention to?

078ee00e-104f-4613-a42a-1f68f8d51b39-0
00:39:23.400 --> 00:39:25.240
Yeah, I mean, I can start.

a37b150f-dd88-43c1-a99a-4cd910a04674-0
00:39:25.520 --> 00:39:28.320
I think there's numerous applications of
Gen.

b71345a1-ab80-4780-8493-b4059e19588e-0
00:39:28.320 --> 00:39:28.800
AI.

12f23dc0-850e-4ce2-9b84-2812baa8ce87-0
00:39:28.800 --> 00:39:32.216
So for example,
AI scribes are pretty common these days

12f23dc0-850e-4ce2-9b84-2812baa8ce87-1
00:39:32.216 --> 00:39:37.158
where you have an AI basically listening
to the conversation between the patient

12f23dc0-850e-4ce2-9b84-2812baa8ce87-2
00:39:37.158 --> 00:39:40.269
and the physician,
transcribing that conversation,

12f23dc0-850e-4ce2-9b84-2812baa8ce87-3
00:39:40.269 --> 00:39:43.320
then generating a medical note for the
clinician.

6091394b-73e7-4666-a8aa-bf54ff33a1e2-0
00:39:44.320 --> 00:39:48.978
We've seen reductions in human effort by
about 3:00 to 4:00 hours per week for

6091394b-73e7-4666-a8aa-bf54ff33a1e2-1
00:39:48.978 --> 00:39:50.040
family physicians.

d3f36624-c04b-4eb7-bd8f-6286d55e2560-0
00:39:52.000 --> 00:39:55.791
So they can either go home and spend time
with their family or see more patients in

d3f36624-c04b-4eb7-bd8f-6286d55e2560-1
00:39:55.791 --> 00:39:56.920
the emergency department.

8f5407f6-5371-4843-a5cd-8cdfe02b8af6-0
00:39:56.920 --> 00:40:03.198
There's data now from it was a bridge in
the United States as well as an emerging

8f5407f6-5371-4843-a5cd-8cdfe02b8af6-1
00:40:03.198 --> 00:40:08.712
data from Alberta as well as here in
Ontario where if you use it in the

8f5407f6-5371-4843-a5cd-8cdfe02b8af6-2
00:40:08.712 --> 00:40:10.320
emergency department.

01202d45-907a-4d97-9971-28a1af1fed06-0
00:40:11.600 --> 00:40:15.764
They're estimating almost 1/3 of an Ed
physician's time and as an Ed doc,

01202d45-907a-4d97-9971-28a1af1fed06-1
00:40:15.764 --> 00:40:17.960
so he knows this much better than I do.

5692de93-072f-473f-b54f-a6c4f1409233-0
00:40:18.240 --> 00:40:22.749
It's actually spent on the administrative
tasks and it's really cumbersome where

5692de93-072f-473f-b54f-a6c4f1409233-1
00:40:22.749 --> 00:40:27.091
what they're reporting is a 10 to 20%
improvement in patient throughput using

5692de93-072f-473f-b54f-a6c4f1409233-2
00:40:27.091 --> 00:40:27.760
the scribes.

9459a2a9-e143-44e9-96dc-cc2ddec4bab8-0
00:40:28.440 --> 00:40:31.207
We've seen other applications,
for example,

9459a2a9-e143-44e9-96dc-cc2ddec4bab8-1
00:40:31.207 --> 00:40:36.238
in there's a paper that was published in
the New England Journal of Medicine AI

9459a2a9-e143-44e9-96dc-cc2ddec4bab8-2
00:40:36.238 --> 00:40:40.199
earlier this year using chat bots for
mental health treatment.

7ca54af8-12aa-4229-b208-8b567d4e4c46-0
00:40:41.160 --> 00:40:45.050
And there was the reality of rather
randomized trial was the findings for

7ca54af8-12aa-4229-b208-8b567d4e4c46-1
00:40:45.050 --> 00:40:48.941
suggesting that the patients were saying
these are about as good as human

7ca54af8-12aa-4229-b208-8b567d4e4c46-2
00:40:48.941 --> 00:40:49.520
therapists.

ca584e86-e52f-4499-8f29-abacb8a1e284-0
00:40:51.400 --> 00:40:54.578
Now I don't know if it'll actually reach
the level of a highly trained

ca584e86-e52f-4499-8f29-abacb8a1e284-1
00:40:54.578 --> 00:40:55.160
psychiatrist.

b8f7f247-e429-4ed6-8a99-4fe82b95db0b-0
00:40:55.160 --> 00:40:55.760
I doubt that.

4bfaf6a3-4ca1-4b7c-a35a-04e7246a8bef-0
00:40:56.160 --> 00:40:59.882
But when you have millions of people with
poor access to mental health,

4bfaf6a3-4ca1-4b7c-a35a-04e7246a8bef-1
00:40:59.882 --> 00:41:01.640
it's probably better than nothing.

1443eb7e-23ea-4b1a-b2f4-d7ba461e7ee2-0
00:41:01.840 --> 00:41:05.757
The last example I'll give is in China,
there's something called WATO GPT in

1443eb7e-23ea-4b1a-b2f4-d7ba461e7ee2-1
00:41:05.757 --> 00:41:09.267
certain regions where a patient can
actually chat with the chat bot,

1443eb7e-23ea-4b1a-b2f4-d7ba461e7ee2-2
00:41:09.267 --> 00:41:13.387
say here are the symptoms of experiencing
and it'll have an interaction with the

1443eb7e-23ea-4b1a-b2f4-d7ba461e7ee2-3
00:41:13.387 --> 00:41:16.440
patient and say,
sounds like you need to see a neurologist.

24a46354-c84e-4530-9c72-e492f86221c5-0
00:41:16.760 --> 00:41:17.840
Here's the closest one to you.

badc9560-3caf-4730-93d6-bd2b527448b8-0
00:41:17.840 --> 00:41:19.360
Would you like me to make an appointment
for you?

097b562f-dfb7-495d-a4cb-0d5f281981a2-0
00:41:20.200 --> 00:41:21.840
And what they've done and audit on it.

17ccf6b1-3454-44e8-bb50-efef7247b092-0
00:41:21.840 --> 00:41:23.520
They've shown 98% accuracy.

37ff0abd-9107-44f5-ba26-386b87601f7e-0
00:41:24.560 --> 00:41:27.000
Those are some examples of of how Gen.

cd168ee6-cc12-4fb1-9011-21ecbcd77836-0
00:41:27.000 --> 00:41:27.840
AI is being used.

93e3323c-ed44-40c5-b1a5-1d851fe794d3-0
00:41:32.680 --> 00:41:34.128
Evan,
do you want to go on the one move to the

93e3323c-ed44-40c5-b1a5-1d851fe794d3-1
00:41:34.128 --> 00:41:34.560
next question.

c4772af6-e9f0-4948-900d-0a57f8ead8a0-0
00:41:34.920 --> 00:41:35.960
I can jump in really quickly.

a3d39f3a-865a-4bda-808c-7f60888d80c8-0
00:41:35.960 --> 00:41:39.980
Just agree with Muhammad,
a great lit review on the fly there,

a3d39f3a-865a-4bda-808c-7f60888d80c8-1
00:41:39.980 --> 00:41:41.320
which was incredible.

bbfcdc11-120b-4708-b629-1021cf7e79ab-0
00:41:42.120 --> 00:41:45.229
And then, you know,
I think it's a specific project that

bbfcdc11-120b-4708-b629-1021cf7e79ab-1
00:41:45.229 --> 00:41:49.701
we're working on in the emerge is trying
to think through how to safely use these

bbfcdc11-120b-4708-b629-1021cf7e79ab-2
00:41:49.701 --> 00:41:52.429
Jenny I tools to directly interact with
patients,

bbfcdc11-120b-4708-b629-1021cf7e79ab-3
00:41:52.429 --> 00:41:56.629
to expand the quality of that visit,
but also to improve health literacy for

bbfcdc11-120b-4708-b629-1021cf7e79ab-4
00:41:56.629 --> 00:41:57.120
patients.

df8971b8-6127-499e-8114-f08fc2829cd5-0
00:41:58.280 --> 00:42:02.036
And really these projects are around
diving into understanding how do you

df8971b8-6127-499e-8114-f08fc2829cd5-1
00:42:02.036 --> 00:42:06.200
validate these tools before you turn them
on and make them patient facing, right?

9ee306ad-479b-4c99-8d2c-de71ee3f524b-0
00:42:06.480 --> 00:42:10.920
We've got a simple bot that can educate
patients around fever when they arrive.

c02fef47-28cb-4d08-899b-611110546460-0
00:42:11.880 --> 00:42:12.640
Parents are loving it.

f2404369-44ac-4f7b-a37e-b0b622d521c3-0
00:42:12.640 --> 00:42:15.809
They're getting their,
their questions answered around how to

f2404369-44ac-4f7b-a37e-b0b622d521c3-1
00:42:15.809 --> 00:42:19.440
use Tylenol and when is it risky versus
safe to, to, to use, etcetera.

4c195cb3-24ed-4eb5-8dbb-a784f841165f-0
00:42:19.880 --> 00:42:22.918
But how do you make sure the prompt
engineering is right and the,

4c195cb3-24ed-4eb5-8dbb-a784f841165f-1
00:42:22.918 --> 00:42:26.601
the bot is actually going to stick to the
knowledge source that you want it to,

4c195cb3-24ed-4eb5-8dbb-a784f841165f-2
00:42:26.601 --> 00:42:29.869
and it's not going to hallucinate in,
in areas where it's dangerous to

4c195cb3-24ed-4eb5-8dbb-a784f841165f-3
00:42:29.869 --> 00:42:30.560
hallucinate in.

06529e24-3d99-49cc-97ae-79a08b529a69-0
00:42:30.560 --> 00:42:34.139
And and then when do you decide that
you've done enough testing and that it's

06529e24-3d99-49cc-97ae-79a08b529a69-1
00:42:34.139 --> 00:42:34.920
OK to turn it on?

3c3b1684-71c6-45e6-94ae-62422fda0283-0
00:42:35.720 --> 00:42:37.600
Like there's no clear answer to that.

19d7c203-5f62-48cd-987e-be0e05f9e33b-0
00:42:37.960 --> 00:42:41.428
And so the approach we're taking is we're
building some of these tools in the

19d7c203-5f62-48cd-987e-be0e05f9e33b-1
00:42:41.428 --> 00:42:44.897
research environment and we're just
sitting beside patients and using it with

19d7c203-5f62-48cd-987e-be0e05f9e33b-2
00:42:44.897 --> 00:42:45.120
them.

d3f3adba-d135-4e89-ae4a-7694de6ed7f4-0
00:42:45.280 --> 00:42:49.212
And just like learning, you know,
in a structured research way to figure

d3f3adba-d135-4e89-ae4a-7694de6ed7f4-1
00:42:49.212 --> 00:42:52.660
out how do we craft a more rich
understanding of of when we can

d3f3adba-d135-4e89-ae4a-7694de6ed7f4-2
00:42:52.660 --> 00:42:56.000
confidently say, yeah,
this one's ready to go patient facing.

c1ebf158-8646-453f-9abd-55978b1f7bd0-0
00:42:56.000 --> 00:43:00.400
OK, have some quick fire questions now.

9e6a644e-5c7b-4a5a-a1b8-54f799277567-0
00:43:00.840 --> 00:43:02.040
So just go for it.

f359a400-f30b-4988-b32f-8002d79843c4-0
00:43:03.000 --> 00:43:07.991
I'll just call out the name and of one of
you and I'll just just answer it whatever

f359a400-f30b-4988-b32f-8002d79843c4-1
00:43:07.991 --> 00:43:09.120
comes to your mind.

854a7d45-05d2-493f-adcb-e66c4bbf2276-0
00:43:09.560 --> 00:43:12.240
So maybe I'll start with you chance
putting you on the spot here.

9b6af7ac-e40f-4975-b635-2987a44c4b26-0
00:43:12.520 --> 00:43:18.798
What's the single trait that separates an
AI leader and think of a person AI leader

9b6af7ac-e40f-4975-b635-2987a44c4b26-1
00:43:18.798 --> 00:43:19.920
from a laggard?

4ab609a8-8ead-4d59-968e-c29138886799-0
00:43:21.400 --> 00:43:22.120
A single trait.

78218259-8fc7-4827-aa15-e47d9bd7a46b-0
00:43:25.760 --> 00:43:30.200
I would say it's,
it's around comfort with risk, right?

add0dd0b-f8b1-4af8-8284-0e9141a8c51b-0
00:43:30.280 --> 00:43:31.600
These are new technologies.

b85f77a3-9a10-4ee1-b6be-e7fc7043b9f9-0
00:43:31.800 --> 00:43:32.760
There is risk.

852e11a4-1c3c-464c-be82-ac87f66ae795-0
00:43:33.120 --> 00:43:38.035
And so can you find that trade off that
works for for your particular context,

852e11a4-1c3c-464c-be82-ac87f66ae795-1
00:43:38.035 --> 00:43:40.400
be it patient facing or, or otherwise?

7921223e-f290-4466-9304-25a60efeab28-0
00:43:40.400 --> 00:43:44.754
So I think it's understand deep
understanding of risk and ability to to

7921223e-f290-4466-9304-25a60efeab28-1
00:43:44.754 --> 00:43:45.480
manage that.

05f0cd7b-a40a-4f3c-b5aa-156e6edc3798-0
00:43:46.320 --> 00:43:50.051
Mahisha,
if you could fix one thing in healthcare

05f0cd7b-a40a-4f3c-b5aa-156e6edc3798-1
00:43:50.051 --> 00:43:52.440
AI tomorrow, what would that be?

01abe1f9-924c-42d1-a857-785504ff6e3c-0
00:43:54.480 --> 00:43:55.600
I had a magic wand.

be3b97c7-d0ed-4ec4-a81b-bfd827341c7a-0
00:43:58.080 --> 00:43:59.400
Well, that's a challenging question.

4a56d2d4-eae3-452b-8cd1-433d51278beb-0
00:44:01.680 --> 00:44:06.010
I think that if I could fix one thing
that would, I think,

4a56d2d4-eae3-452b-8cd1-433d51278beb-1
00:44:06.010 --> 00:44:12.102
unlock a lot of work that we're doing is
to address some of the challenges we have

4a56d2d4-eae3-452b-8cd1-433d51278beb-2
00:44:12.102 --> 00:44:17.974
with access to data across the country
that is in a standardized and accessible

4a56d2d4-eae3-452b-8cd1-433d51278beb-3
00:44:17.974 --> 00:44:21.277
way,
secure and that would allow us to power

4a56d2d4-eae3-452b-8cd1-433d51278beb-4
00:44:21.277 --> 00:44:23.920
and advance some of this technology.

dca30757-bb28-4d8f-91c6-cf5fdbae5dbf-0
00:44:25.120 --> 00:44:28.680
OK, Mohammed,
this is going to be a tough one, I think.

1e01bbef-dadb-4117-83a1-476beaca26df-0
00:44:29.080 --> 00:44:32.480
And you can see a pass if you think this
is something you want to avoid.

5644907d-8616-4210-a97f-d86195ccbb7f-0
00:44:33.080 --> 00:44:36.160
Who should own AI within a hospital?

d6a7c2eb-22d5-40cd-80a0-5c6adee061ff-0
00:44:36.160 --> 00:44:37.600
Is that the CIO?

6e6f129b-39dc-4e31-994f-0d7287cb7e72-0
00:44:37.600 --> 00:44:39.360
Is that the CMIO?

5072f945-2f20-4003-90e1-4529bc3bb5ac-0
00:44:39.360 --> 00:44:42.680
Is that the chief operating officer Who
should own it?

320a78cb-66cc-466e-9e4e-b954443b8839-0
00:44:44.560 --> 00:44:44.880
Yeah.

f2243814-8345-4439-bab0-c09fadf5eea8-0
00:44:44.920 --> 00:44:47.600
You know, I,
I don't think anyone group should own it.

00fe3531-6807-4368-ad48-f220b12ccf93-0
00:44:47.720 --> 00:44:49.480
I think it should be a collective.

7a3104fa-5793-4e3b-8f1b-06afc12e88ae-0
00:44:49.480 --> 00:44:55.040
Now in terms of the operations of it,
my sense is ideally it'd be situated with,

7a3104fa-5793-4e3b-8f1b-06afc12e88ae-1
00:44:55.040 --> 00:44:58.610
with this,
the digital or the CIO type of of group,

7a3104fa-5793-4e3b-8f1b-06afc12e88ae-2
00:44:58.610 --> 00:45:03.622
just because the reliance on the
technology to operationalize it is just

7a3104fa-5793-4e3b-8f1b-06afc12e88ae-3
00:45:03.622 --> 00:45:04.240
so heavy.

9e2804d7-c186-4a46-b45f-b29fda93f483-0
00:45:05.840 --> 00:45:10.062
But at the very least,
there should be a dotted line or a matrix

9e2804d7-c186-4a46-b45f-b29fda93f483-1
00:45:10.062 --> 00:45:14.480
where clinical teams also own it,
administrative teams also own it.

071ed462-d52d-41a7-a503-c8fa93829efe-0
00:45:14.920 --> 00:45:17.960
But the OP should be with the with the
CIO in my opinion.

57a4b958-a2cb-4018-8539-8595cea831d8-0
00:45:18.640 --> 00:45:18.960
Thanks, Don.

77138fbd-31f0-44b8-af13-1662b44443eb-0
00:45:18.960 --> 00:45:24.884
On the screen, Devin,
which clinical area is most ready for AI

77138fbd-31f0-44b8-af13-1662b44443eb-1
00:45:24.884 --> 00:45:27.800
and which one's most resistant?

61afa109-49e3-4906-818d-aaa79844ee18-0
00:45:29.960 --> 00:45:30.560
It's interesting.

26f5f3a9-0c3a-413e-b1f4-276834087d50-0
00:45:30.560 --> 00:45:33.655
I'm reflecting on Mohammed's answer and
thinking that like,

26f5f3a9-0c3a-413e-b1f4-276834087d50-1
00:45:33.655 --> 00:45:37.576
we just created an entirely new group
that involved everyone from different

26f5f3a9-0c3a-413e-b1f4-276834087d50-2
00:45:37.576 --> 00:45:41.755
sections who is most ready for AII mean,
you know that I'm biased because I work

26f5f3a9-0c3a-413e-b1f4-276834087d50-3
00:45:41.755 --> 00:45:43.200
in the emergency department.

72df4ab5-a996-4c87-b8cf-2594a3010018-0
00:45:43.440 --> 00:45:48.028
But I think that there's something to the
large volume of data that gets generated

72df4ab5-a996-4c87-b8cf-2594a3010018-1
00:45:48.028 --> 00:45:50.240
in the emerge and the rapid turn around.

88666073-a3c8-4b02-a1bf-8f7bfa43f860-0
00:45:50.240 --> 00:45:54.982
Like you can go through these rapid
cycles of improvement in specialties that

88666073-a3c8-4b02-a1bf-8f7bfa43f860-1
00:45:54.982 --> 00:45:56.320
have high turn around.

29616459-69c6-457b-8259-b6dfb4207333-0
00:45:56.680 --> 00:46:01.043
And there's specialties that are just
naturally used to being adaptive and

29616459-69c6-457b-8259-b6dfb4207333-1
00:46:01.043 --> 00:46:03.720
changing and just having to figure stuff
out.

aaabb5de-25b8-4945-918f-4e1a6bc41961-0
00:46:03.720 --> 00:46:07.673
And so those are the specialties that I
think are most ready because AI being

aaabb5de-25b8-4945-918f-4e1a6bc41961-1
00:46:07.673 --> 00:46:10.360
integrated into a workflow isn't that new
necessary.

38705b6a-3370-49aa-81e9-fde492b02789-0
00:46:10.360 --> 00:46:13.240
It's just another change that they figure
out out of necessity.

1d843fbc-986b-4f38-964f-44978fdfbaba-0
00:46:14.200 --> 00:46:18.323
The second part of your question,
I'm going to change a little bit and just

1d843fbc-986b-4f38-964f-44978fdfbaba-1
00:46:18.323 --> 00:46:22.555
to say that where I think the biggest
opportunity as a system, as a province,

1d843fbc-986b-4f38-964f-44978fdfbaba-2
00:46:22.555 --> 00:46:25.920
we need to think about is AI in the
community practice space.

d79a1185-ccb8-44ee-b9e2-89ad6de551a1-0
00:46:26.720 --> 00:46:31.353
This is where we have an opportunity to
dramatically expand capacity in a way

d79a1185-ccb8-44ee-b9e2-89ad6de551a1-1
00:46:31.353 --> 00:46:34.680
that would impact care at the highest
numbers, I think.

fd3fa887-a231-4462-a82e-5a534f1a0e09-0
00:46:34.680 --> 00:46:38.278
So even though I work at the hospital and
I'm in the emergency position,

fd3fa887-a231-4462-a82e-5a534f1a0e09-1
00:46:38.278 --> 00:46:41.680
I see the consequences of the community
practice space under strain.

0768ea94-5b23-4e0c-91a4-22045633ed3e-0
00:46:41.760 --> 00:46:45.000
That's where we need to focus a lot of
policy attention and investment.

d25fdce5-b68f-4082-8605-55d9ffc10f0b-0
00:46:46.680 --> 00:46:53.839
Maybe I'll start with you again, Devin,
which AI trend will shape healthcare most

d25fdce5-b68f-4082-8605-55d9ffc10f0b-1
00:46:53.839 --> 00:46:55.760
in the next few years?

6f1f9e41-bdee-4ef2-8408-ab2d39fec78c-0
00:46:56.280 --> 00:47:00.947
I don't like trend questions,
but I think I'd like to know from you on

6f1f9e41-bdee-4ef2-8408-ab2d39fec78c-1
00:47:00.947 --> 00:47:02.920
what you're hearing out there.

7bd89f63-610b-490e-84b5-03e2f176d56b-0
00:47:03.480 --> 00:47:06.508
No,
I think large language models and as they

7bd89f63-610b-490e-84b5-03e2f176d56b-1
00:47:06.508 --> 00:47:10.920
are maturing are going to be dramatically
transformational, right?

365fa718-dc71-4bb4-a3a0-977519a086ed-0
00:47:10.920 --> 00:47:13.960
Like we were in an era of building set.

1e94c1c2-08f4-4bd3-8bc3-be7476762971-0
00:47:15.040 --> 00:47:18.274
You know, you you build an algorithm,
you trained it on a lot of data and you

1e94c1c2-08f4-4bd3-8bc3-be7476762971-1
00:47:18.274 --> 00:47:20.680
try to figure out how might you be able
to generalize it.

ccc290e3-8687-4f01-8b29-77562a312bd3-0
00:47:20.960 --> 00:47:23.782
These foundation models that are starting
to emerge,

ccc290e3-8687-4f01-8b29-77562a312bd3-1
00:47:23.782 --> 00:47:27.777
I think are going to be the most
disruptive and I think they have the most

ccc290e3-8687-4f01-8b29-77562a312bd3-2
00:47:27.777 --> 00:47:30.920
potential to generalize at a scale we
haven't seen before.

55444c1d-2516-48cc-8353-52f06862326a-0
00:47:33.760 --> 00:47:36.483
Would you have,
would you rather have perfect data or

55444c1d-2516-48cc-8353-52f06862326a-1
00:47:36.483 --> 00:47:37.240
perfect models?

9f37c5ab-0e99-4c53-b8b8-ac87fdc4bd6a-0
00:47:37.240 --> 00:47:41.600
I think the two are dependent on each
other, aren't they?

58126fae-0a09-4ffb-932c-79c81bed19be-0
00:47:43.120 --> 00:47:46.240
The, the data are driven,
the models are driven by the data.

b9ed06a2-afbb-4f49-9a17-60c368e66e5d-0
00:47:46.520 --> 00:47:49.760
So I would say you start with the data.

884846f3-81a0-4632-a73f-9755bb5645c4-0
00:47:50.960 --> 00:47:54.200
They'll the perfect models by definition
would have the perfect data.

9aa77d56-6aad-4ef5-b216-4528a73875a3-0
00:47:54.200 --> 00:47:56.120
So I picked the perfect model.

748828f4-b0a9-4671-b9da-a89273940614-0
00:47:58.040 --> 00:48:06.801
I see Don Marsha,
what's perhaps what what's will AI make

748828f4-b0a9-4671-b9da-a89273940614-1
00:48:06.801 --> 00:48:12.240
clinicians more human or less human?

25a694f0-1a80-4c1a-b822-e26e4e65e024-0
00:48:14.840 --> 00:48:19.902
I think that I think that the way in my
opinion that we need to look at AI is

25a694f0-1a80-4c1a-b822-e26e4e65e024-1
00:48:19.902 --> 00:48:25.158
that it's not going to replace there's
expertise and compassion and empathy that

25a694f0-1a80-4c1a-b822-e26e4e65e024-2
00:48:25.158 --> 00:48:27.560
we have within our healthcare system.

f7493c11-f2ec-49b3-a3b2-02284a6d8e36-0
00:48:27.560 --> 00:48:34.021
I think that I think that what we need to
look at is how AI can augment and enhance

f7493c11-f2ec-49b3-a3b2-02284a6d8e36-1
00:48:34.021 --> 00:48:37.560
the work that we're doing for our
clinicians.

bf0bd251-f1e5-4197-8cde-04b8a9ce1dc3-0
00:48:38.240 --> 00:48:39.400
OK, great.

760b65fd-7a26-45c4-8771-4cefd6688d12-0
00:48:40.000 --> 00:48:45.720
And chance which will move faster
regulation or AI innovation?

51c85475-2b49-46ae-b67a-2d8578302603-0
00:48:48.720 --> 00:48:50.520
We don't even have regulation in Canada.

6752ca06-a17a-4a8f-bc53-6c1374cefbd4-0
00:48:50.520 --> 00:48:54.400
That's you know what the, the,
so obviously regulation loses.

92dabb5c-62a0-428e-83d3-34b292444f05-0
00:48:54.400 --> 00:48:59.411
But the one that I'm interested in maybe
that I could throw a question back to the

92dabb5c-62a0-428e-83d3-34b292444f05-1
00:48:59.411 --> 00:49:03.880
panel is we get a lot of questions of
from physicians will AI replace me?

3bd43730-572b-4126-a856-46a3c4ef6a2c-0
00:49:05.000 --> 00:49:09.963
Our answer to date based on everything
we've seen is no physicians remain

3bd43730-572b-4126-a856-46a3c4ef6a2c-1
00:49:09.963 --> 00:49:11.640
accountable for the care.

7942439e-b4c6-4d30-a19c-8d25fdf48ab1-0
00:49:11.760 --> 00:49:15.473
But I'm kind of interested for the other
panels on that question because we get

7942439e-b4c6-4d30-a19c-8d25fdf48ab1-1
00:49:15.473 --> 00:49:16.680
that question quite a bit.

3a747c21-c37e-4dbd-9bd3-b78f25ba7244-0
00:49:19.040 --> 00:49:19.880
Anyone wants to take that?

8459406e-1374-4c16-bea7-effcda56f097-0
00:49:22.960 --> 00:49:24.280
Thanks for moderating that chance.

57fbe321-7cf3-4fff-8fa7-6c10f6472849-0
00:49:25.160 --> 00:49:28.701
I mean, I,
I think we'll start to see that AI will

57fbe321-7cf3-4fff-8fa7-6c10f6472849-1
00:49:28.701 --> 00:49:30.160
add capacity to care.

bca1f429-0333-4f37-87ae-3ba9f2f6bfe9-0
00:49:30.480 --> 00:49:33.053
And so how you interpret that like for
example,

bca1f429-0333-4f37-87ae-3ba9f2f6bfe9-1
00:49:33.053 --> 00:49:36.644
that automated console pathway to
psychiatry that's in our emerge,

bca1f429-0333-4f37-87ae-3ba9f2f6bfe9-2
00:49:36.644 --> 00:49:38.359
like you could see that as well.

5447dda2-9488-4075-b8c8-a16511733c43-0
00:49:38.360 --> 00:49:42.720
You replaced me from sitting there and
typing in, putting the consult in.

8d735d78-1340-476c-bc3e-1f40531d27d7-0
00:49:43.080 --> 00:49:44.720
I interpret it as amazing.

01dd4a54-9ffd-49c0-ad2a-6f1d6d2d6263-0
00:49:44.720 --> 00:49:49.120
That little task is done for me already
and cares advancing for that patient.

e844c591-8fc5-40f1-b67c-79fe6ffadfc5-0
00:49:49.440 --> 00:49:54.267
So I think as AI starts to add capacity,
the replacement narrative will start to

e844c591-8fc5-40f1-b67c-79fe6ffadfc5-1
00:49:54.267 --> 00:49:55.400
drift a little bit.

2b232c8f-9fdd-4e9e-8292-62a88d76f0d4-0
00:49:55.400 --> 00:49:58.880
People will like want you to replace that
aspect of my work.

2e4f6532-0beb-4be4-ad86-02fb6ce51fed-0
00:49:59.240 --> 00:49:59.480
Yeah.

bf85166d-bb2b-4eca-b287-48b7f27a1a76-0
00:50:02.680 --> 00:50:03.080
Anyone else?

3c01c363-99a6-4b96-afbf-e42dbc1d43fb-0
00:50:04.720 --> 00:50:08.880
Yeah, maybe I'll,
I'll kind of push on that a bit as well.

7396bf65-76fb-4266-85bf-4c8240e88b43-0
00:50:09.840 --> 00:50:11.920
Just putting in a little thing on you.

1d59857e-8d29-40e8-b352-d1d4c3861802-0
00:50:12.040 --> 00:50:15.720
You'll see more and more things like this
replaced by AI.

829f9aa1-3598-43a5-b684-3c5b12caa9fa-0
00:50:16.000 --> 00:50:19.163
Amazon cuts 14,
000 jobs and its biggest ever workforce

829f9aa1-3598-43a5-b684-3c5b12caa9fa-1
00:50:19.163 --> 00:50:21.480
slash for bet on artificial intelligence.

7b4be64c-ac85-4b1c-b584-b3dfce266d6c-0
00:50:22.200 --> 00:50:27.586
You know,
my sense is in the past we would say AI

7b4be64c-ac85-4b1c-b584-b3dfce266d6c-1
00:50:27.586 --> 00:50:30.280
won't replace clinicians.

af58cfed-bc64-47d9-b331-8d668b2dea57-0
00:50:30.280 --> 00:50:33.120
Clinicians who use AI will replace
clinicians who don't use AI.

d3d76e84-cd92-4948-ac80-dbbf82088d43-0
00:50:33.120 --> 00:50:34.480
And I think that's largely going to be
true.

e097698a-3639-485e-8cd7-8591401f6c6a-0
00:50:34.480 --> 00:50:38.494
But I would say for certain jobs,
being a realist, yeah,

e097698a-3639-485e-8cd7-8591401f6c6a-1
00:50:38.494 --> 00:50:40.960
I think it's going to replace them.

7bbeedc5-25b0-453c-94e4-1bd5fe064750-0
00:50:42.400 --> 00:50:47.560
Because right now what we're saying is
for menial tasks, AI is pretty good.

a989f033-8212-4aa0-86f4-1f23903ca45f-0
00:50:48.880 --> 00:50:53.831
And our definition of menial is here,
right as the eye gets better,

a989f033-8212-4aa0-86f4-1f23903ca45f-1
00:50:53.831 --> 00:50:58.345
our definition of medial,
as it starts taking over functions,

a989f033-8212-4aa0-86f4-1f23903ca45f-2
00:50:58.345 --> 00:51:01.840
we'll move to here and then it'll move to
here.

7e10b5b0-2bfd-4af1-814a-7d86e59263fc-0
00:51:02.760 --> 00:51:06.942
So I think it's an evolving thing where
given its capabilities now there,

7e10b5b0-2bfd-4af1-814a-7d86e59263fc-1
00:51:06.942 --> 00:51:10.560
there may be a plateau here where it's
only capable of so much.

a3b37557-8dd8-47a4-930d-10fdb1738663-0
00:51:11.160 --> 00:51:15.435
But certainly we're already starting to
see it replace some human functions that

a3b37557-8dd8-47a4-930d-10fdb1738663-1
00:51:15.435 --> 00:51:16.280
are a bit scary.

17d36ef9-35bc-4f15-b4c2-b1bda190fd1f-0
00:51:17.120 --> 00:51:19.320
So I'm going to say it actually will take
jobs.

7ef06b52-dcec-4626-87b0-c51ad81fc3bd-0
00:51:22.200 --> 00:51:22.680
Great question.

b4097bfc-06fc-467d-85b6-00f7186c364d-0
00:51:22.680 --> 00:51:28.963
There chance maybe for anyone here,
AI in healthcare,

b4097bfc-06fc-467d-85b6-00f7186c364d-1
00:51:28.963 --> 00:51:35.480
Canadian healthcare today overrated or
under delivered?

a8ae1dc3-25fe-43fa-a770-be02773b288f-0
00:51:40.640 --> 00:51:41.360
I'd say both.

c30e7c29-7d0c-4f03-87c1-25d746a33418-0
00:51:42.760 --> 00:51:43.320
It is.

4e32299e-3835-4d5d-a007-b7faf803cb26-0
00:51:43.360 --> 00:51:44.280
The hype is real.

9c858c1c-4f9f-48eb-8daf-cdbee3a10bd7-0
00:51:45.320 --> 00:51:46.520
It's heavily under delivered.

1fe687d2-69a9-4ec8-b253-a068cdb6c5b1-0
00:51:46.520 --> 00:51:49.933
I don't think we're leveraging it as well
as we should and I think there will be

1fe687d2-69a9-4ec8-b253-a068cdb6c5b1-1
00:51:49.933 --> 00:51:51.955
some, some catastrophes actually,
to be honest,

1fe687d2-69a9-4ec8-b253-a068cdb6c5b1-2
00:51:51.955 --> 00:51:54.400
because we won't use it properly and
people will be hurt.

a6e6e172-06ac-4ab3-bcbb-01f3f601d3e5-0
00:51:55.160 --> 00:51:57.280
But hopefully we'll learn from it.

5e61261c-f153-4daa-9c4d-4739b06a1209-0
00:52:01.600 --> 00:52:04.560
Let me look at any other questions from
chat.

e65636a6-20b3-4403-8834-7bd036bf5d32-0
00:52:04.560 --> 00:52:06.080
Please keep them coming along.

4e30e66c-1a2d-4da1-9ae6-33f496d2d403-0
00:52:09.680 --> 00:52:12.640
Maybe on the ethical risk side,
we haven't touched risk too much.

2826c6e2-942c-4343-8789-4e24bae3116f-0
00:52:12.640 --> 00:52:13.960
So thanks for that question.

f43be019-6e0a-4370-a5a6-85a6fa5b6bca-0
00:52:14.120 --> 00:52:18.520
What do you see as as a ethical rescue?

8de2ad1f-ce19-4fc2-b73a-2b3104b41dc5-0
00:52:18.520 --> 00:52:20.920
See AI raising specifically in healthcare.

dc5468de-0196-4354-8c37-a5654d027bd6-0
00:52:21.200 --> 00:52:27.287
I know we've talked about the ethics of
applying that in certain cases where bias

dc5468de-0196-4354-8c37-a5654d027bd6-1
00:52:27.287 --> 00:52:29.440
might exist and other things.

365aefa8-1db5-4f3a-95de-eb18b5c4410a-0
00:52:29.880 --> 00:52:34.369
But just give me a little bit of your
view of as you're building AI solutions,

365aefa8-1db5-4f3a-95de-eb18b5c4410a-1
00:52:34.369 --> 00:52:38.120
how much of that is coming in and what
are you doing to mitigate?

472f7823-e4d7-4942-a545-b59a1133ec8e-0
00:52:40.960 --> 00:52:46.161
One thing that we do really well is we,
we take a design methodologies approach

472f7823-e4d7-4942-a545-b59a1133ec8e-1
00:52:46.161 --> 00:52:50.388
early on from the beginning,
which effectively involves engaging

472f7823-e4d7-4942-a545-b59a1133ec8e-2
00:52:50.388 --> 00:52:54.160
stakeholders really intimately
understanding the problem.

eb0b4740-2722-4fa0-bbe2-04bcdbff0e78-0
00:52:54.160 --> 00:52:58.139
But I think it's also to understand that
there may be a problem with wait times,

eb0b4740-2722-4fa0-bbe2-04bcdbff0e78-1
00:52:58.139 --> 00:53:01.774
let's say in a particular category,
but different subgroups of people may

eb0b4740-2722-4fa0-bbe2-04bcdbff0e78-2
00:53:01.774 --> 00:53:05.754
have that problem in different ways and
may experience that problem in different

eb0b4740-2722-4fa0-bbe2-04bcdbff0e78-3
00:53:05.754 --> 00:53:06.000
ways.

72096469-6c35-43ab-98f3-9a45b9d195c0-0
00:53:06.000 --> 00:53:09.989
And so design methodologies really helps
you flush out a deep and intimate

72096469-6c35-43ab-98f3-9a45b9d195c0-1
00:53:09.989 --> 00:53:14.351
understanding of the, the problem space,
which then as you build the technologies

72096469-6c35-43ab-98f3-9a45b9d195c0-2
00:53:14.351 --> 00:53:17.117
and look to solve for it,
you then understand like,

72096469-6c35-43ab-98f3-9a45b9d195c0-3
00:53:17.117 --> 00:53:21.479
what are the key domains from an ethics
perspective that you're not only going to

72096469-6c35-43ab-98f3-9a45b9d195c0-4
00:53:21.479 --> 00:53:24.139
try to proactively address,
but more importantly,

72096469-6c35-43ab-98f3-9a45b9d195c0-5
00:53:24.139 --> 00:53:26.480
I think you're going to monitor for,
right?

6fb6fd87-2444-4e65-b16b-4cddbd6f11ae-0
00:53:26.480 --> 00:53:28.480
And you're going to be able to actually
lean into.

77bc1721-069a-458d-a9f7-402a81796419-0
00:53:29.320 --> 00:53:30.360
And so maybe I'll leave it.

e27f89e9-fe1a-44f8-abff-48d0dc927290-0
00:53:30.360 --> 00:53:32.320
There's so much we can talk about on this
thing,

e27f89e9-fe1a-44f8-abff-48d0dc927290-1
00:53:32.320 --> 00:53:33.920
but I'll see Mohammed's got his hand up.

d1277b4f-4e27-4261-86a2-3bc042129186-0
00:53:34.120 --> 00:53:35.040
Great way to approach it.

f25125bb-f1a5-4969-bedd-ad535b9a4370-0
00:53:37.080 --> 00:53:40.720
Yeah, they'll probably it's, it's such a,
a rich question.

62ce74e8-9df3-4ad2-9f7d-00b9339328ab-0
00:53:41.880 --> 00:53:47.520
The whole ethics of AI is, is really,
really challenging.

7e249466-a3d7-43f1-bc3a-79a2a0b6e47e-0
00:53:47.560 --> 00:53:49.360
Maybe 2 points that I'll make.

a215f9b4-87c5-43cb-a297-83bb0778d17c-0
00:53:49.680 --> 00:53:53.424
One is we have a solution now that to
this day we haven't deployed because of

a215f9b4-87c5-43cb-a297-83bb0778d17c-1
00:53:53.424 --> 00:53:54.240
ethical concerns.

6e256604-310c-462a-ac06-064e80d02107-0
00:53:56.720 --> 00:53:59.858
There's a there's a group of patients who
are highly,

6e256604-310c-462a-ac06-064e80d02107-1
00:53:59.858 --> 00:54:03.520
highly vulnerable IV drug users with
structural vulnerability.

8e503a1e-13af-4b3d-8579-a4fb56a0f8ed-0
00:54:05.760 --> 00:54:08.600
We have an algorithm that can actually
identify them in the emergency department.

f1831708-762b-40f4-acf2-858b02782bb5-0
00:54:09.800 --> 00:54:15.280
And these are patients that would really
benefit from addictions medicine.

cd09852c-a32d-47af-b746-944b8921473c-0
00:54:16.480 --> 00:54:19.793
They have a 13 to 15 fold higher rate of
mortality relative to the general

cd09852c-a32d-47af-b746-944b8921473c-1
00:54:19.793 --> 00:54:20.280
population.

e41208af-a487-4aba-a8ea-dd597c481d07-0
00:54:21.840 --> 00:54:25.446
But how do we actually deploy an
algorithm when the patients themselves

e41208af-a487-4aba-a8ea-dd597c481d07-1
00:54:25.446 --> 00:54:27.600
say,
I don't want to be identified because

e41208af-a487-4aba-a8ea-dd597c481d07-2
00:54:27.600 --> 00:54:31.608
there's massive stigma issues and you're
going to be doing that with AI against

e41208af-a487-4aba-a8ea-dd597c481d07-3
00:54:31.608 --> 00:54:32.560
without my consent?

e9412088-158b-4572-8da0-5bcc88d9307f-0
00:54:33.840 --> 00:54:34.960
That's a bit concerning.

d32dbc91-f387-4bd6-9e25-1f5a9b603c6b-0
00:54:35.960 --> 00:54:40.138
We've had algorithms that have been
proposed where we actually routinely

d32dbc91-f387-4bd6-9e25-1f5a9b603c6b-1
00:54:40.138 --> 00:54:43.000
screened medical images to do to identify
a race.

2703476b-e503-4d1b-bec1-273b49c10d38-0
00:54:43.440 --> 00:54:48.308
Some algorithms have shown to have over
95% accuracy in identifying a person's

2703476b-e503-4d1b-bec1-273b49c10d38-1
00:54:48.308 --> 00:54:51.760
race in very broad, rough terms,
but still pretty good.

7f626043-b894-4ca7-bd22-405036ba3d4f-0
00:54:52.160 --> 00:54:54.040
Can we do that without patient consent?

892df677-0d75-4b76-abb8-187d15f290d8-0
00:54:55.400 --> 00:54:56.040
Probably not.

50d6f8db-72e4-4201-9a89-1addab13bbef-0
00:54:57.040 --> 00:54:59.160
The other thing I'll say is nothing's
perfect.

7f300c71-ec0e-4033-bb94-874324ae0282-0
00:55:00.360 --> 00:55:03.288
So in reality, in the academic world,
you can say, well,

7f300c71-ec0e-4033-bb94-874324ae0282-1
00:55:03.288 --> 00:55:06.680
you can look at 100 different bias
parameters for your algorithm.

8abae866-e8d5-4bf3-a37c-d7e361ca9f76-0
00:55:07.440 --> 00:55:08.520
In reality, you can't.

c69da970-f170-40ee-9360-3f145b69187e-0
00:55:09.640 --> 00:55:14.036
So for example, with one of our solutions,
which monitors our patients every hour on

c69da970-f170-40ee-9360-3f145b69187e-1
00:55:14.036 --> 00:55:17.916
the hour and predicts if they're going to
die and alerts the medical team,

c69da970-f170-40ee-9360-3f145b69187e-2
00:55:17.916 --> 00:55:21.433
we published a paper in the CMHA last
year showing a 26% reduction,

c69da970-f170-40ee-9360-3f145b69187e-3
00:55:21.433 --> 00:55:22.520
unexpected mortality.

cf793df8-efd2-4788-a769-5a8174eac5e8-0
00:55:23.080 --> 00:55:24.400
These things are helping save lives.

3bec254a-2045-44b6-b5fa-29104c478ced-0
00:55:24.840 --> 00:55:29.593
And when I go to present and I say,
we didn't do a bias assessment on race

3bec254a-2045-44b6-b5fa-29104c478ced-1
00:55:29.593 --> 00:55:33.080
because we don't collect race data,
so we can't do it.

35906395-700d-4faa-8958-e4915f9f05c0-0
00:55:33.920 --> 00:55:36.600
We have issues, people saying,
is that ethical for you to deploy it?

01530a81-9476-4234-bbf6-61007d7e8226-0
00:55:37.040 --> 00:55:40.122
And the question becomes,
we can take the algorithm down,

01530a81-9476-4234-bbf6-61007d7e8226-1
00:55:40.122 --> 00:55:44.480
wait many years till we collect race data
and let 26% more people die every year.

013d6069-38a8-4449-b27c-730f17d9b39f-0
00:55:45.120 --> 00:55:45.880
Is that ethical?

dfb233d3-f7f0-475c-9d31-777e671ba127-0
00:55:47.640 --> 00:55:50.712
So in the real world,
it's not as straightforward as we would

dfb233d3-f7f0-475c-9d31-777e671ba127-1
00:55:50.712 --> 00:55:50.960
like.

619fed33-febb-4ffa-afcd-9a6c1b4b07e2-0
00:55:53.640 --> 00:55:54.520
Thanks for sharing that.

8a1350df-892b-490e-aa80-ce258cdcd78e-0
00:55:55.000 --> 00:55:56.560
Such a true story.

6bde6683-d34a-4a88-99e3-1207f584747c-0
00:55:57.320 --> 00:56:01.480
Chance you want to go next just just on
that piece.

1aad5617-7a5f-4f4e-acd5-eba3b34740fe-0
00:56:01.480 --> 00:56:08.213
The the other part is is that, you know,
physicians are being held to like account

1aad5617-7a5f-4f4e-acd5-eba3b34740fe-1
00:56:08.213 --> 00:56:14.865
for bias or consider bias in AI and it's
totally unclear how to even do that Like

1aad5617-7a5f-4f4e-acd5-eba3b34740fe-2
00:56:14.865 --> 00:56:21.679
to to actually come up with a definition
of what is what is bias is is not trivial.

63efabdd-8c47-47d7-8a69-84ce9e9e26a9-0
00:56:21.680 --> 00:56:24.240
It's actually like a deep societal
question.

17d1c4d1-3a2f-43a9-9fb6-d453202f56f3-0
00:56:24.560 --> 00:56:27.680
You know, we'll have a great,
a great example, right?

5d16f284-d1f1-406a-99b1-13ecc7a050a8-0
00:56:27.680 --> 00:56:30.280
Like what, what,
what do we really mean by that?

c947d440-f1d5-4b2c-ace7-3e4f6172def1-0
00:56:30.280 --> 00:56:34.720
And so I think that'll be an interesting
one for sure where we'll be looking at

c947d440-f1d5-4b2c-ace7-3e4f6172def1-1
00:56:34.720 --> 00:56:38.440
deployments and we'll we'll have to
measure it comes back to risk.

d225d6b3-cd66-445d-ad6b-991302692eb9-0
00:56:38.440 --> 00:56:40.120
So yeah, really interesting.

fabbb5d1-f070-4797-bbfb-409f39b846c6-0
00:56:42.000 --> 00:56:46.476
The question on value keeps coming back
around, OK, well, not sort of value,

fabbb5d1-f070-4797-bbfb-409f39b846c6-1
00:56:46.476 --> 00:56:50.953
but I'd say more around how do we make
sure that the accuracy of an AI model

fabbb5d1-f070-4797-bbfb-409f39b846c6-2
00:56:50.953 --> 00:56:52.640
that we built is high enough?

43a5a4fa-e67a-4c0d-89c9-499852d4d011-0
00:56:52.960 --> 00:56:56.912
And when we,
we can sometimes test these AI models,

43a5a4fa-e67a-4c0d-89c9-499852d4d011-1
00:56:56.912 --> 00:57:01.320
the expectation is, well, it should be 99.
9999% or 100%.

11543b16-fdee-49d8-938d-c0697c88aeba-0
00:57:02.000 --> 00:57:03.400
Well, it's 95%.

2087f8b1-40b4-40d3-96b6-d5f2fa17daca-0
00:57:03.400 --> 00:57:04.320
That's not good enough.

168efb22-e58f-4f9f-83ea-ea0d67e6f15c-0
00:57:04.920 --> 00:57:09.200
How have you thought about accuracy of AI
models?

040467c8-75ec-439d-bc5a-7cd03f971c8d-0
00:57:09.560 --> 00:57:12.774
And especially when you contrast it with,
well,

040467c8-75ec-439d-bc5a-7cd03f971c8d-1
00:57:12.774 --> 00:57:15.520
is it better than a human doing that job?

cc89f98b-6e9c-4fe4-b1c1-dba73a0b65c1-0
00:57:16.680 --> 00:57:19.200
Is it,
what do you benchmark that against Marsha?

7992dfd1-53da-43f1-a320-b5f54d55260a-0
00:57:19.200 --> 00:57:20.400
Do you want to go for that one?

72863cad-390e-4981-a526-0e6b3a01e29c-0
00:57:22.000 --> 00:57:24.503
Yeah,
actually I remember hearing a a

72863cad-390e-4981-a526-0e6b3a01e29c-1
00:57:24.503 --> 00:57:29.115
presentation by you Mohammed,
I think it was the last couple of years

72863cad-390e-4981-a526-0e6b3a01e29c-2
00:57:29.115 --> 00:57:34.649
and it really resonated for me around we,
our system is not perfect and the what we

72863cad-390e-4981-a526-0e6b3a01e29c-3
00:57:34.649 --> 00:57:36.559
deliver is not 100% accurate.

b1ee824b-7950-4a66-9898-f93db2d45c96-0
00:57:36.560 --> 00:57:41.543
And so I think our benchmark has to be
against how we're performing today and

b1ee824b-7950-4a66-9898-f93db2d45c96-1
00:57:41.543 --> 00:57:46.336
that if we can implement something that's
augmenting or enhancing and it's

b1ee824b-7950-4a66-9898-f93db2d45c96-2
00:57:46.336 --> 00:57:51.320
improving an outcome for another 5% or
it's reducing the mortality as example

b1ee824b-7950-4a66-9898-f93db2d45c96-3
00:57:51.320 --> 00:57:54.642
that was given,
I think that's a benefit and that's

b1ee824b-7950-4a66-9898-f93db2d45c96-4
00:57:54.642 --> 00:57:56.240
that's worth progressing.

1816fcd6-695a-4f2a-ab8d-61c99ba5ad69-0
00:57:56.240 --> 00:58:02.342
I don't think that we can hold AI systems
and technology to a standard that's so

1816fcd6-695a-4f2a-ab8d-61c99ba5ad69-1
00:58:02.342 --> 00:58:07.240
high that we can never actually realize
benefits moving forward.

12740b64-1214-43a8-ab40-c4824c4f941a-0
00:58:07.240 --> 00:58:11.499
I think that what we need to have around
it is the appropriate safeguards and

12740b64-1214-43a8-ab40-c4824c4f941a-1
00:58:11.499 --> 00:58:16.031
oversight that we can help mitigate and
make sure that it's still being effective,

12740b64-1214-43a8-ab40-c4824c4f941a-2
00:58:16.031 --> 00:58:20.400
it's still enhancing the the performance
as it was when we first rolled it out.

8d40570e-4fb5-4db8-9694-cacaf224c23e-0
00:58:22.280 --> 00:58:23.120
Thanks a lot.

acfa2320-6cf7-474b-b54c-a7d5424698c8-0
00:58:23.240 --> 00:58:25.360
OK,
so we're now we're now getting to time.

973628da-1d49-4989-8b12-9e2bf1488c8a-0
00:58:25.360 --> 00:58:28.846
So I'm going to ask one final question to,
to everyone and just quickly in one or

973628da-1d49-4989-8b12-9e2bf1488c8a-1
00:58:28.846 --> 00:58:31.440
two sentences, if you could answer,
and then we'll close it.

5dcbe3d2-87da-4c31-b648-9f9dd02d54a2-0
00:58:32.600 --> 00:58:37.812
If every healthcare organization in
Canada took one bold step towards

5dcbe3d2-87da-4c31-b648-9f9dd02d54a2-1
00:58:37.812 --> 00:58:42.280
realizing value from AI,
what should it be in your opinion?

ce1e7b53-4707-4b0e-9efb-ece11cb3813f-0
00:58:42.760 --> 00:58:43.520
Anyone can go.

fd8b8e78-26dd-4355-a8b4-322dd92c6065-0
00:58:53.720 --> 00:58:59.907
I'm going to steal something that chances
said earlier,

fd8b8e78-26dd-4355-a8b4-322dd92c6065-1
00:58:59.907 --> 00:59:07.200
and that is a step from risk avoidance to
risk mitigation, right?

8dcc18cf-a116-4289-8c33-81865b35f7ac-0
00:59:10.280 --> 00:59:11.520
Everyone has to answer this question.

332bd7e7-4d9e-467e-ac13-72fcd37b1993-0
00:59:14.600 --> 00:59:19.560
I was going to say the one step you can
take is have thoughtful courage.

34b9fd23-cbd3-474b-8320-ba9c09b2ae78-0
00:59:21.160 --> 00:59:23.960
We we pontificate, we talk quite a bit.

c47b50fe-883f-4b5a-beca-7b3da56892ef-0
00:59:25.440 --> 00:59:27.820
We use,
let's figure out data governance for

c47b50fe-883f-4b5a-beca-7b3da56892ef-1
00:59:27.820 --> 00:59:30.200
years and years and years as a delay
tactic.

2973f8bb-a2b1-4a2c-8da2-25f807ff8d02-0
00:59:31.080 --> 00:59:36.608
The many governance frameworks, pick one,
spend a few months modifying it,

2973f8bb-a2b1-4a2c-8da2-25f807ff8d02-1
00:59:36.608 --> 00:59:41.400
do just get in there and do have a
thoughtful measured approach.

f319457d-37b2-441d-84fc-ae4728d51486-0
00:59:41.760 --> 00:59:45.080
But you got to jump in at now.

f0c94908-adaa-4692-8009-65fa20d45e52-0
00:59:46.040 --> 00:59:49.320
Yeah, I'll jump in next.

443607db-91d2-4cbb-8772-7807ab879100-0
00:59:49.400 --> 00:59:54.795
I think that we need to do this work in a
way that's human centered and value

443607db-91d2-4cbb-8772-7807ab879100-1
00:59:54.795 --> 00:59:55.280
driven.

896d2d8c-4ea3-4ad7-b202-cf1a3bfe46c3-0
00:59:55.400 --> 01:00:02.109
I think that we need to Orient around the
value for the parties that we are

896d2d8c-4ea3-4ad7-b202-cf1a3bfe46c3-1
01:00:02.109 --> 01:00:07.760
impacted and for at the end of the day
our patients and chance.

f142f06d-e999-4602-b35e-5c4277722b19-0
01:00:09.320 --> 01:00:11.480
I did this all wrong to have the last
word.

9df0e3a5-70dc-4118-a2d6-00d42149664a-0
01:00:12.400 --> 01:00:16.518
I guess the only thing I would add is
going back to one of my earlier points is

9df0e3a5-70dc-4118-a2d6-00d42149664a-1
01:00:16.518 --> 01:00:17.600
improved AI literacy.

7737788e-f20d-405e-9ecf-54f32f878dad-0
01:00:17.600 --> 01:00:18.120
So I think.

65c38a0e-2ff9-465c-9139-54b39ec00c3a-0
01:00:18.440 --> 01:00:21.200
I think people have to use these tools to
understand how they work.

77f2916b-cbcc-4c48-8200-66d905806b72-0
01:00:21.200 --> 01:00:25.013
So I would advocate for low risk tools
like an AI scribe,

77f2916b-cbcc-4c48-8200-66d905806b72-1
01:00:25.013 --> 01:00:29.880
assuming that you've taken care of all
the data, privacy risks, etcetera.

c5b713cd-ddd0-43a4-9b00-d06a0088bc0e-0
01:00:30.800 --> 01:00:33.000
Just just start using them to see how to
see how they work.

35779a2f-e4c7-4262-aee0-b8e364960727-0
01:00:34.720 --> 01:00:37.480
Well, thank you folks for for doing this.

5535a3c1-c048-4d35-a120-c14ec8e063a7-0
01:00:37.480 --> 01:00:40.475
We talked about archetypes and value,
the way literacy,

5535a3c1-c048-4d35-a120-c14ec8e063a7-1
01:00:40.475 --> 01:00:44.914
aligning investment and the patient story
and how the process and the policies and

5535a3c1-c048-4d35-a120-c14ec8e063a7-2
01:00:44.914 --> 01:00:46.840
landing the airplane comes together.

1d154891-db94-4818-859b-26bf1ece4c66-0
01:00:46.840 --> 01:00:50.840
So hopefully the audience got got
something out of this.

f8d25894-6e9f-4313-b722-c226214ef289-0
01:00:51.520 --> 01:00:56.008
I just wanted to do a bit of housekeeping
on then for some announcement of an

f8d25894-6e9f-4313-b722-c226214ef289-1
01:00:56.008 --> 01:00:59.000
upcoming event, November 25th,
12th to 1:00 PM EST.

28e5a7c6-98c0-4a52-99c3-77885dd6f167-0
01:00:59.640 --> 01:01:04.326
We're going to be talking about embedding
integrity by design to sustain healthcare

28e5a7c6-98c0-4a52-99c3-77885dd6f167-1
01:01:04.326 --> 01:01:07.451
transformation,
something that was brought up by Marsha

28e5a7c6-98c0-4a52-99c3-77885dd6f167-2
01:01:07.451 --> 01:01:08.400
just now as well.

f089a182-946b-46c6-8ab0-75dc542c7cd1-0
01:01:08.720 --> 01:01:13.800
So with that, we're going to exit now.

8799c35b-9933-4dd0-a227-e1e1de4ef1f6-0
01:01:13.800 --> 01:01:16.880
So thank you again and have a great day
ahead.