AI Should Give Clinicians Back to Patients

WEBVTT

00:00:01.567 --> 00:00:02.186
All right,

00:00:02.246 --> 00:00:04.288
everybody wants to talk about AI,

00:00:04.448 --> 00:00:05.488
or maybe they don't wanna talk about it,

00:00:05.508 --> 00:00:07.008
but they sure are talking about it.

00:00:07.288 --> 00:00:08.227
And the questions come up.

00:00:08.589 --> 00:00:09.808
Will it replace clinicians?

00:00:10.349 --> 00:00:11.448
Will it write our notes?

00:00:11.888 --> 00:00:12.128
Yes.

00:00:12.489 --> 00:00:14.349
Will it make healthcare more efficient?

00:00:14.869 --> 00:00:16.129
Hmm, remains to be seen.

00:00:16.649 --> 00:00:18.890
Larry's latest article asks a completely

00:00:18.971 --> 00:00:19.870
different question.

00:00:21.030 --> 00:00:23.012
What if AI isn't here to replace

00:00:23.051 --> 00:00:23.632
clinicians?

00:00:24.091 --> 00:00:25.492
What if it's here to give them back

00:00:25.532 --> 00:00:27.292
the part of the job they fell in

00:00:27.452 --> 00:00:28.013
love with,

00:00:28.853 --> 00:00:29.774
in the first place.

00:00:30.254 --> 00:00:32.515
So Larry, let's start there.

00:00:33.176 --> 00:00:35.375
Why do you think everybody's asking the

00:00:35.396 --> 00:00:36.036
wrong question?

00:00:36.076 --> 00:00:37.676
Because you open and you say,

00:00:37.777 --> 00:00:38.898
forty years ago,

00:00:39.078 --> 00:00:41.779
I took an oath to care for people.

00:00:41.939 --> 00:00:44.000
You started with an oath forty years ago,

00:00:44.280 --> 00:00:45.179
not with AI.

00:00:45.439 --> 00:00:47.301
Why did you start with that oath from

00:00:47.341 --> 00:00:47.820
long ago?

00:00:47.841 --> 00:00:49.142
Yeah,

00:00:49.822 --> 00:00:51.902
I think it's because the conversation

00:00:51.963 --> 00:00:55.624
about AI has already gone backwards and I

00:00:55.643 --> 00:00:56.965
wanted to reset

00:00:58.000 --> 00:01:01.323
and level set around why we got into

00:01:01.363 --> 00:01:02.744
this profession.

00:01:04.766 --> 00:01:07.549
Everyone is now entering it asking what

00:01:07.590 --> 00:01:09.271
the technology can do.

00:01:09.391 --> 00:01:11.573
I wanted to start with what I promised

00:01:11.614 --> 00:01:13.835
to do as a caregiver, four years ago,

00:01:16.597 --> 00:01:18.340
to a patient as a caregiver,

00:01:18.359 --> 00:01:19.281
not to a vendor.

00:01:20.066 --> 00:01:22.268
And so the oath sort of frames the

00:01:22.328 --> 00:01:23.868
tool, not the other way around.

00:01:23.909 --> 00:01:24.769
If you start with AI,

00:01:24.790 --> 00:01:26.751
you're already conceded the argument to

00:01:26.772 --> 00:01:27.793
the people selling it.

00:01:29.194 --> 00:01:30.614
I'm just not interested in that

00:01:30.674 --> 00:01:32.736
conversation for the purposes of call to

00:01:32.796 --> 00:01:33.076
care.

00:01:33.096 --> 00:01:35.920
I'm interested in what we owe the person

00:01:35.980 --> 00:01:37.180
on the table, the patient.

00:01:37.762 --> 00:01:39.143
So the oath is the standard,

00:01:39.183 --> 00:01:40.203
the code of ethics,

00:01:40.304 --> 00:01:42.165
the whole professional reasons why we got

00:01:42.245 --> 00:01:42.626
in here.

00:01:42.825 --> 00:01:44.968
AI either measures up to it or doesn't.

00:01:46.462 --> 00:01:46.742
Alright,

00:01:46.762 --> 00:01:48.322
so one of the strongest lines in the

00:01:48.382 --> 00:01:49.862
article, I thought anyway,

00:01:49.903 --> 00:01:52.864
was the only question worth asking is

00:01:52.944 --> 00:01:55.204
whether we have the discipline to use this

00:01:55.245 --> 00:01:58.367
tool for the one thing it's actually good

00:01:58.427 --> 00:02:02.328
for good for giving the human being back

00:02:02.388 --> 00:02:03.227
to the patient.

00:02:03.688 --> 00:02:05.248
What do you mean by giving the human

00:02:05.308 --> 00:02:06.049
being back?

00:02:06.170 --> 00:02:06.969
What does that mean to you?

00:02:08.038 --> 00:02:10.098
Yeah, so if you think historically,

00:02:10.879 --> 00:02:13.281
as your time as a PT and healthcare

00:02:13.320 --> 00:02:13.961
provider,

00:02:14.021 --> 00:02:17.263
what we've really been robbed is with time

00:02:17.282 --> 00:02:18.084
with patients,

00:02:18.704 --> 00:02:20.145
particularly in physical therapy where

00:02:20.164 --> 00:02:21.985
everything has been, you know,

00:02:22.466 --> 00:02:24.146
remediated or, you know,

00:02:24.187 --> 00:02:26.027
gone down to a billable unit.

00:02:27.269 --> 00:02:28.629
It's all about documentation,

00:02:28.669 --> 00:02:31.170
pre-authorization, productivity mandates,

00:02:31.230 --> 00:02:33.293
the nine-minute, eight-minute rule,

00:02:33.353 --> 00:02:35.373
not twelve-minute visit windows.

00:02:35.413 --> 00:02:36.754
Every one of those looks like another

00:02:37.770 --> 00:02:40.432
minute of attention redirected at a

00:02:40.552 --> 00:02:41.252
screen.

00:02:41.793 --> 00:02:46.415
So giving the human being back means

00:02:46.536 --> 00:02:48.856
reversing that theft.

00:02:49.798 --> 00:02:50.838
Eyes on the patient,

00:02:50.959 --> 00:02:53.219
listening for the thing that the chart can

00:02:53.259 --> 00:02:54.039
never capture,

00:02:54.120 --> 00:02:56.961
noticing what the patient isn't saying,

00:02:57.062 --> 00:02:58.883
the verbal and the nonverbal.

00:02:59.342 --> 00:03:01.384
That encounter fully present

00:03:02.239 --> 00:03:05.920
undistracted is what got engineered out of

00:03:06.040 --> 00:03:08.763
workflow and maximizing productivity and

00:03:08.782 --> 00:03:11.424
efficiency we need to engineer it back in

00:03:11.924 --> 00:03:13.485
and that's the whole argument everything

00:03:13.544 --> 00:03:17.046
else is footnotes i started writing an

00:03:17.127 --> 00:03:19.209
article and tim flynn was helping me with

00:03:19.288 --> 00:03:21.110
it and i think i actually sent it

00:03:21.129 --> 00:03:22.931
to you to review a bit and it's

00:03:22.971 --> 00:03:25.491
about the five agents everything in ai now

00:03:25.512 --> 00:03:26.673
is about agentic

00:03:27.252 --> 00:03:28.835
movement and moving to agents.

00:03:28.914 --> 00:03:29.475
I love that.

00:03:30.798 --> 00:03:32.659
And I've identified what I believe are

00:03:32.699 --> 00:03:35.002
five agents that can help us in physical

00:03:35.022 --> 00:03:35.522
therapy.

00:03:36.485 --> 00:03:37.885
because at the end of the day they

00:03:37.925 --> 00:03:41.429
bring humanness back to health care and

00:03:41.468 --> 00:03:42.649
the one we're going to talk about in

00:03:42.669 --> 00:03:44.811
this article or what we did talk about

00:03:44.830 --> 00:03:46.872
in this article and it'll transgress a

00:03:46.912 --> 00:03:48.674
little bit also while i'm reintroducing

00:03:48.693 --> 00:03:51.235
call to care is about the whole idea

00:03:51.256 --> 00:03:54.138
of bringing the human back and using ai

00:03:54.237 --> 00:03:56.658
scribe for the documentation pieces but

00:03:56.699 --> 00:03:58.980
don't get me wrong scribes and agents can

00:03:59.001 --> 00:04:01.282
do many other things they can do pre-cert

00:04:02.203 --> 00:04:03.302
They can do outcomes.

00:04:03.323 --> 00:04:05.304
They can give us feedback loops and a

00:04:05.383 --> 00:04:07.264
daily report of how we conducted.

00:04:08.183 --> 00:04:09.723
They can listen to our notes and give

00:04:09.783 --> 00:04:12.365
us constructive criticism on how we better

00:04:12.844 --> 00:04:14.865
have verbal and nonverbal skills.

00:04:14.905 --> 00:04:16.706
So I'm all for AI,

00:04:17.245 --> 00:04:20.346
but I believe its function is really

00:04:20.826 --> 00:04:23.987
around humanity and allowing us to be

00:04:24.067 --> 00:04:26.487
fully human and present for our patients

00:04:26.668 --> 00:04:28.627
as opposed to be fully engaged in a

00:04:28.687 --> 00:04:29.327
technology.

00:04:30.565 --> 00:04:32.067
So let's bring some humanity into it.

00:04:32.307 --> 00:04:34.129
You brought someone into this particular

00:04:34.209 --> 00:04:34.649
article.

00:04:34.949 --> 00:04:35.790
You named her Sarah.

00:04:35.911 --> 00:04:37.591
You said, you know, painted a picture.

00:04:38.432 --> 00:04:39.533
She's not a bad clinician.

00:04:39.995 --> 00:04:41.536
She's been around for about twelve years.

00:04:41.855 --> 00:04:43.778
She's not burned out because she stopped

00:04:43.838 --> 00:04:44.399
caring.

00:04:44.978 --> 00:04:47.440
She's burned out because the system slowly

00:04:47.521 --> 00:04:48.641
pulled her away

00:04:49.591 --> 00:04:51.351
from the work she loved.

00:04:51.771 --> 00:04:52.833
And I think a lot of people will

00:04:52.853 --> 00:04:54.012
say, well, these kids today,

00:04:54.033 --> 00:04:55.173
it's very easy to fall into it.

00:04:55.194 --> 00:04:56.394
These kids today don't.

00:04:56.954 --> 00:04:58.036
But when you look at it that way,

00:04:58.076 --> 00:04:59.456
because she's burned out because the

00:04:59.516 --> 00:05:01.057
system pulled her away from the thing she

00:05:01.096 --> 00:05:01.276
loved.

00:05:02.177 --> 00:05:04.298
How common do you think a Sarah is?

00:05:05.778 --> 00:05:07.600
um well i've known a hundred of them

00:05:07.680 --> 00:05:09.680
i've employed probably hundreds of them in

00:05:09.701 --> 00:05:11.702
the last forty years when your burnout

00:05:11.762 --> 00:05:14.345
rate runs from forty five to seventy one

00:05:14.384 --> 00:05:16.447
percent just on pts and i cite in

00:05:16.466 --> 00:05:19.108
the article dentists and dental hygienists

00:05:19.129 --> 00:05:21.651
and others um and when you have a

00:05:21.790 --> 00:05:24.392
one year arguably fifteen thousand pts

00:05:24.432 --> 00:05:26.675
walking out the door she's not the outlier

00:05:27.394 --> 00:05:29.798
You know, she's she's the norm.

00:05:29.817 --> 00:05:32.300
And and we have to recall the data,

00:05:32.339 --> 00:05:34.461
which shows that on a full time equivalent

00:05:34.502 --> 00:05:34.942
basis,

00:05:35.023 --> 00:05:36.785
only twenty five percent of all physical

00:05:36.824 --> 00:05:38.567
therapists that are licensed are actually

00:05:38.627 --> 00:05:39.848
practicing clinically.

00:05:40.468 --> 00:05:42.391
But what makes Sarah's story so worth

00:05:42.411 --> 00:05:44.052
telling is that she's a great clinician.

00:05:45.192 --> 00:05:46.353
Her outcomes are strong,

00:05:46.394 --> 00:05:48.035
her patients like her,

00:05:48.074 --> 00:05:49.514
her colleagues respect her,

00:05:49.535 --> 00:05:52.997
and she's counting the years until she can

00:05:53.036 --> 00:05:53.817
afford to leave.

00:05:54.617 --> 00:05:55.678
And that's the tell.

00:05:56.478 --> 00:05:58.540
It's not the weak clinician burning out,

00:05:58.579 --> 00:05:59.661
it's the good ones,

00:06:00.040 --> 00:06:01.942
the ones who came in with passion and

00:06:01.982 --> 00:06:03.182
skill that the system,

00:06:03.562 --> 00:06:04.983
the ones who worked hard in school to

00:06:05.043 --> 00:06:06.863
even get accepted in PT school.

00:06:07.403 --> 00:06:08.785
Now, they didn't lose their ability,

00:06:08.805 --> 00:06:10.966
they left their sense of calling.

00:06:10.985 --> 00:06:14.187
They lost their meaningful purposeful work

00:06:14.880 --> 00:06:17.502
And once that goes, the clock is running.

00:06:17.543 --> 00:06:19.163
It's just a function until they're looking

00:06:19.184 --> 00:06:20.966
to get into a non-clinical position.

00:06:22.442 --> 00:06:24.165
I feel like one of the distinctions that

00:06:24.185 --> 00:06:27.726
you made is two sides to a coin.

00:06:27.848 --> 00:06:29.829
People go back and forth and they argue

00:06:29.869 --> 00:06:30.209
over it.

00:06:30.750 --> 00:06:33.052
That burnout isn't this personal failure.

00:06:33.411 --> 00:06:35.434
It's a system that stopped replenishing

00:06:35.473 --> 00:06:36.113
its people,

00:06:36.214 --> 00:06:37.194
what I just said a minute ago.

00:06:37.696 --> 00:06:39.858
Why is that distinction so important to

00:06:39.958 --> 00:06:43.461
focus on where that fault actually is or

00:06:43.860 --> 00:06:44.922
where the solution could be?

00:06:46.245 --> 00:06:47.607
Yeah, it's like everything, anything else.

00:06:47.867 --> 00:06:49.269
You know, once you have the diagnosis,

00:06:49.309 --> 00:06:50.891
then you could determine the prescription.

00:06:50.971 --> 00:06:51.211
Right.

00:06:51.230 --> 00:06:52.932
And I would also argue the system includes

00:06:52.973 --> 00:06:55.055
our education system because we know

00:06:55.115 --> 00:06:57.778
statistically that by the time new PTs

00:06:57.838 --> 00:07:00.620
land in our clinic for their first full

00:07:00.661 --> 00:07:01.401
time job,

00:07:01.521 --> 00:07:03.163
a good portion of them are already burned

00:07:03.283 --> 00:07:03.444
out.

00:07:04.214 --> 00:07:06.476
Now, we can argue about the term burnout.

00:07:06.497 --> 00:07:08.098
There's some medical things around it.

00:07:08.158 --> 00:07:09.999
I've written pretty extensively about it.

00:07:10.379 --> 00:07:12.060
But let's forget about the word for a

00:07:12.100 --> 00:07:12.380
minute.

00:07:12.420 --> 00:07:14.322
Just think of it as people leaving the

00:07:14.343 --> 00:07:16.064
profession or being disengaged.

00:07:16.184 --> 00:07:17.824
When you're disengaged and working with

00:07:17.884 --> 00:07:18.326
patients,

00:07:18.346 --> 00:07:20.286
you're less likely to have very good

00:07:20.367 --> 00:07:20.827
outcomes.

00:07:22.088 --> 00:07:24.149
um if it's a personal failure then the

00:07:24.189 --> 00:07:26.812
fix would be resilience you know wellness

00:07:27.471 --> 00:07:28.932
you know a lot of those kinds of

00:07:28.973 --> 00:07:30.754
things but we've been running that

00:07:30.793 --> 00:07:33.096
experiment now for a long time you know

00:07:33.196 --> 00:07:35.117
burnout has actually gotten worse even

00:07:35.137 --> 00:07:36.898
though there's been a more sensitivity to

00:07:36.918 --> 00:07:39.519
it it's actually gotten worse so the inner

00:07:39.759 --> 00:07:41.802
intervention is is failed because it's

00:07:41.841 --> 00:07:44.122
treated it's treated the wrong problem

00:07:44.142 --> 00:07:45.843
when you name it as a system failure

00:07:46.305 --> 00:07:48.305
you're forced to redesign the system

00:07:48.346 --> 00:07:49.326
that's much harder

00:07:51.947 --> 00:07:53.930
And no one in leadership really wants to

00:07:53.970 --> 00:07:56.992
talk about because it lands accountability

00:07:57.031 --> 00:07:59.553
where it belongs, the operators.

00:07:59.774 --> 00:08:01.774
And again, it's all back to operation,

00:08:01.795 --> 00:08:03.656
the people who built the workflows that

00:08:03.697 --> 00:08:05.577
the very clinicians were trying to keep.

00:08:05.598 --> 00:08:07.059
So that's the distinction.

00:08:07.459 --> 00:08:12.403
It's not a semantic description, Jimmy.

00:08:12.442 --> 00:08:14.603
It's really what determines whether

00:08:15.064 --> 00:08:17.685
anything actually can change or happen to

00:08:17.725 --> 00:08:18.286
improve this.

00:08:19.771 --> 00:08:20.951
Let's talk about AI scribes.

00:08:21.593 --> 00:08:24.576
I feel like everybody markets them when

00:08:24.596 --> 00:08:25.576
you see an ad for them.

00:08:25.637 --> 00:08:27.178
If you get on a call with someone

00:08:27.199 --> 00:08:28.420
who's wanted to talk about them,

00:08:28.759 --> 00:08:30.641
they lead with efficiency.

00:08:31.122 --> 00:08:32.083
That's their sell.

00:08:32.364 --> 00:08:33.384
It's an efficiency tool.

00:08:33.605 --> 00:08:34.846
And then they'll hit you with a statistic,

00:08:34.927 --> 00:08:36.969
usually in time or money saved.

00:08:37.649 --> 00:08:40.852
You argue that's almost missing the point.

00:08:41.653 --> 00:08:42.815
Where are they missing the point on this?

00:08:43.567 --> 00:08:43.826
Yeah,

00:08:43.866 --> 00:08:47.288
so let's just level set here a bit.

00:08:48.727 --> 00:08:50.708
It took us a long time to get

00:08:50.749 --> 00:08:52.068
adoption to EMR,

00:08:52.089 --> 00:08:53.589
especially in physical therapy.

00:08:54.210 --> 00:08:55.350
The good news, I think,

00:08:55.409 --> 00:08:58.051
is that there's been reasonable adoption

00:08:58.150 --> 00:09:01.051
and quick adoption by AI scribes in PT

00:09:01.110 --> 00:09:02.532
and in healthcare in general.

00:09:02.932 --> 00:09:04.692
I've put dentistry a little bit behind in

00:09:04.711 --> 00:09:05.351
that regards.

00:09:06.312 --> 00:09:08.253
But why has that been the push?

00:09:08.293 --> 00:09:10.033
The push is because we want to make

00:09:10.073 --> 00:09:11.573
the system more efficient.

00:09:12.491 --> 00:09:14.004
and efficiency is the wrong goal.

00:09:14.969 --> 00:09:16.570
If you deploy a scribe and tell a

00:09:16.610 --> 00:09:18.591
therapist that the point is that you can

00:09:18.630 --> 00:09:20.631
now see more patients per day,

00:09:21.293 --> 00:09:22.774
you've simply designed a tool, you know,

00:09:25.554 --> 00:09:27.495
you've designed a tool to restore humanity

00:09:27.535 --> 00:09:29.376
to work and used it to accelerate the

00:09:29.417 --> 00:09:32.239
very thing and the very reason why Sarah

00:09:32.278 --> 00:09:33.019
is burning out.

00:09:33.360 --> 00:09:34.620
You haven't solved the problem,

00:09:34.659 --> 00:09:36.461
you've automated it while leaving

00:09:36.500 --> 00:09:37.721
everything else intact.

00:09:37.802 --> 00:09:39.682
And that's sort of the, in the article,

00:09:39.722 --> 00:09:40.663
I talk about Dr.

00:09:40.703 --> 00:09:43.586
Blaine's calcification with a software

00:09:43.666 --> 00:09:44.066
upgrade.

00:09:45.140 --> 00:09:46.062
And the other thing is,

00:09:46.101 --> 00:09:47.163
if the win is, oh,

00:09:47.182 --> 00:09:48.644
we're doing this for our therapists,

00:09:48.663 --> 00:09:49.745
this is how much we care,

00:09:49.764 --> 00:09:51.125
and this is why we want to attract

00:09:51.186 --> 00:09:53.628
more PTs because we have such shortages,

00:09:54.048 --> 00:09:54.989
because you'll be able to,

00:09:55.370 --> 00:09:56.850
by using a scribe,

00:09:57.250 --> 00:10:00.214
you'll have less time on notes

00:10:00.774 --> 00:10:01.695
you know, in the office,

00:10:01.735 --> 00:10:02.895
so more time at home.

00:10:02.936 --> 00:10:04.037
And, you know, that's real,

00:10:04.076 --> 00:10:05.839
and that's a quality of life improvement,

00:10:05.859 --> 00:10:07.840
but that's just optimizing Sarah's

00:10:07.941 --> 00:10:08.360
evenings.

00:10:08.461 --> 00:10:10.783
I'm not interested in what happens to the

00:10:10.822 --> 00:10:12.544
patient, you know,

00:10:12.965 --> 00:10:14.166
what happened to the patient in the room.

00:10:14.206 --> 00:10:17.168
The scribe's purpose is simply to make

00:10:17.688 --> 00:10:19.671
Sarah faster by that, you know,

00:10:19.770 --> 00:10:21.673
unprofessional management business speak.

00:10:22.693 --> 00:10:23.674
That's not the purpose.

00:10:23.715 --> 00:10:25.937
The purpose is to make her fully present

00:10:25.996 --> 00:10:26.616
for the patient.

00:10:27.421 --> 00:10:29.701
Those are very different goals with very

00:10:29.782 --> 00:10:31.101
different outcomes.

00:10:31.442 --> 00:10:32.682
If it's efficiency,

00:10:32.942 --> 00:10:34.643
it's going to lack effectiveness.

00:10:35.043 --> 00:10:36.263
If it's effectiveness,

00:10:36.503 --> 00:10:38.423
it's going to be got effectiveness.

00:10:39.004 --> 00:10:41.803
Management keeps confusing them and the

00:10:41.865 --> 00:10:42.965
patients pay for it.

00:10:43.304 --> 00:10:45.684
So if you look at most large scale

00:10:46.306 --> 00:10:47.865
operating platforms,

00:10:48.381 --> 00:10:49.062
in healthcare.

00:10:49.101 --> 00:10:52.543
They're getting AI scribes to claim they

00:10:52.582 --> 00:10:54.303
have a better workplace for the therapist

00:10:54.344 --> 00:10:55.724
so they have less notes at home,

00:10:56.424 --> 00:10:58.445
claim the therapist can see more patients,

00:10:58.485 --> 00:11:00.145
but none of them are talking about the

00:11:00.186 --> 00:11:01.706
very thing that it's meant to do,

00:11:02.145 --> 00:11:04.206
bring humanity back to healthcare and

00:11:04.246 --> 00:11:06.067
allow them to be fully present and be

00:11:06.148 --> 00:11:08.467
more effective to produce better outcomes.

00:11:08.888 --> 00:11:10.168
That's the real difference.

00:11:10.729 --> 00:11:11.568
You know,

00:11:11.589 --> 00:11:12.649
the first time I took a look at

00:11:12.669 --> 00:11:14.029
an AI scribe a year and a half,

00:11:14.070 --> 00:11:15.931
two years ago, I, of course,

00:11:16.390 --> 00:11:18.451
default to like an advertising campaign.

00:11:18.490 --> 00:11:18.951
And I said,

00:11:19.351 --> 00:11:21.652
this might bring eye contact back.

00:11:21.751 --> 00:11:22.913
And as simple as that is,

00:11:23.212 --> 00:11:25.693
that felt more humane than the twenty

00:11:25.734 --> 00:11:28.413
eight minutes per note saved.

00:11:28.553 --> 00:11:29.394
That felt better.

00:11:30.735 --> 00:11:34.456
You talk about a cockpit analogy instead

00:11:34.596 --> 00:11:35.777
of automation,

00:11:36.297 --> 00:11:36.677
where

00:11:39.417 --> 00:11:41.580
The autopilot does the mundane.

00:11:41.820 --> 00:11:43.602
It maps, it flies,

00:11:43.682 --> 00:11:44.604
it goes up and down,

00:11:44.703 --> 00:11:46.667
but the human comes in at different

00:11:46.706 --> 00:11:47.086
points.

00:11:47.488 --> 00:11:50.792
Why is automation the wrong mental model

00:11:50.812 --> 00:11:51.352
for healthcare?

00:11:51.393 --> 00:11:52.653
Because behind AI,

00:11:52.754 --> 00:11:54.317
automation has got to be the next most

00:11:54.397 --> 00:11:55.258
used word recently.

00:11:55.477 --> 00:11:57.620
yeah it is it's the wrong word again

00:11:57.639 --> 00:11:59.761
because it implies automation implies

00:11:59.802 --> 00:12:02.524
replacement you know on a factory line a

00:12:02.543 --> 00:12:05.667
machine removes the human in a clinic that

00:12:05.726 --> 00:12:09.529
collapses on contact with reality right a

00:12:09.570 --> 00:12:10.029
therapist

00:12:10.986 --> 00:12:12.928
can't hand clinical judgment to a

00:12:13.009 --> 00:12:14.389
probabilistic system.

00:12:14.750 --> 00:12:16.351
It can't automate that out of it.

00:12:16.772 --> 00:12:19.354
No patient ever got better because a model

00:12:19.835 --> 00:12:22.417
was confident, right?

00:12:22.437 --> 00:12:25.078
So the cockpit is the right mental model

00:12:25.099 --> 00:12:28.001
because the roles stay very, very clear.

00:12:28.842 --> 00:12:32.024
The autopilot flies the plane, altitude,

00:12:32.086 --> 00:12:33.326
course, systems.

00:12:34.027 --> 00:12:35.467
The pilot commands it.

00:12:36.188 --> 00:12:39.413
It takes over at takeoff landing.

00:12:39.832 --> 00:12:42.635
The pilot takes over at landing at

00:12:42.696 --> 00:12:43.336
takeoff.

00:12:43.898 --> 00:12:44.577
In the instance,

00:12:44.597 --> 00:12:46.340
something unexpected happens.

00:12:46.419 --> 00:12:49.663
Nobody says the autopilot replaced the

00:12:49.703 --> 00:12:50.183
pilot.

00:12:51.091 --> 00:12:52.793
His or her interventions are fewer,

00:12:52.854 --> 00:12:54.654
better leveraged, and at higher stakes.

00:12:54.754 --> 00:12:56.456
It allows them to be more present for

00:12:56.496 --> 00:12:58.278
when the other problems that can happen.

00:12:58.798 --> 00:13:00.480
And so that's the architecture for

00:13:00.500 --> 00:13:03.142
healthcare AI that I really, really like.

00:13:03.582 --> 00:13:05.264
It allows the therapist to be on

00:13:05.344 --> 00:13:05.985
autopilot.

00:13:06.424 --> 00:13:08.206
The system manages the workflow.

00:13:08.226 --> 00:13:10.369
The clinician manages the decision.

00:13:10.428 --> 00:13:11.990
It manages the human behavior.

00:13:12.269 --> 00:13:13.171
interactions.

00:13:13.630 --> 00:13:15.673
This idea of very much every time we

00:13:15.692 --> 00:13:16.693
have high technology,

00:13:16.714 --> 00:13:17.995
we have to have high touch.

00:13:18.054 --> 00:13:21.758
And so automation allows the therapist to

00:13:21.798 --> 00:13:26.000
be on autopilot to be more hands-on and

00:13:26.061 --> 00:13:27.081
more high touch.

00:13:28.023 --> 00:13:29.703
The minute you call it automation,

00:13:29.724 --> 00:13:32.025
you've already really started to design

00:13:32.046 --> 00:13:34.408
the wrong outcome, in my opinion.

00:13:35.678 --> 00:13:37.139
One phrase that stuck out was,

00:13:37.240 --> 00:13:38.321
and this feels like it could be on

00:13:38.341 --> 00:13:39.681
a bumper sticker or a T-shirt or an

00:13:39.740 --> 00:13:41.743
ad campaign, was AI drafts,

00:13:42.462 --> 00:13:43.744
clinicians decide.

00:13:44.443 --> 00:13:45.985
Why is that line in the sand so

00:13:46.024 --> 00:13:48.186
important for healthcare leaders designing

00:13:48.745 --> 00:13:52.087
or implementing these AI systems?

00:13:52.447 --> 00:13:55.210
Because generating a note is the easy

00:13:55.289 --> 00:13:56.210
ninety percent.

00:13:56.269 --> 00:13:58.991
The question that earns trust is how you

00:13:59.032 --> 00:14:00.533
confirm that it's true.

00:14:01.567 --> 00:14:01.967
You know,

00:14:01.988 --> 00:14:03.950
a clinician who reviews what the model

00:14:03.990 --> 00:14:06.373
produced and signs it isn't a rubber

00:14:06.432 --> 00:14:06.932
stamp.

00:14:07.394 --> 00:14:09.076
She's an accountable professional

00:14:09.375 --> 00:14:10.658
exercising judgment.

00:14:10.677 --> 00:14:12.679
A clinician who signs without reviewing is

00:14:12.759 --> 00:14:14.481
ratifying, not verifying.

00:14:14.522 --> 00:14:14.942
Right.

00:14:15.823 --> 00:14:18.086
And the liability sits with the clinic

00:14:18.167 --> 00:14:19.769
clinician either way from a legal

00:14:19.788 --> 00:14:20.349
standpoint.

00:14:21.129 --> 00:14:22.350
So AI drafts,

00:14:22.410 --> 00:14:24.432
clinicians decide isn't a slogan,

00:14:24.451 --> 00:14:25.772
it's a design principle.

00:14:26.211 --> 00:14:28.812
It forces accountability into the workflow

00:14:29.274 --> 00:14:32.235
rather than bolting it on after the fact.

00:14:32.725 --> 00:14:32.924
You know,

00:14:33.004 --> 00:14:35.525
any system that doesn't build verification

00:14:35.586 --> 00:14:38.287
as the product, not generation, you know,

00:14:38.326 --> 00:14:40.226
as a malpractice machine, you know,

00:14:40.267 --> 00:14:42.107
dressed up with efficiency software,

00:14:42.168 --> 00:14:42.427
right?

00:14:43.628 --> 00:14:45.129
I won't work with one and I won't

00:14:45.188 --> 00:14:46.428
recommend working with one.

00:14:46.469 --> 00:14:47.808
I mean, that's just the key.

00:14:47.849 --> 00:14:49.789
You have to really let clinicians be

00:14:49.809 --> 00:14:54.711
clinicians and not, you know, defer to AI.

00:14:54.750 --> 00:14:55.672
There's a great saying,

00:14:55.711 --> 00:14:56.812
and it's absolutely true,

00:14:56.831 --> 00:14:57.532
and I've said it,

00:14:57.572 --> 00:14:59.133
and that is that AI doesn't have a

00:14:59.173 --> 00:14:59.712
bad day.

00:15:00.413 --> 00:15:02.076
Well, that's exactly right.

00:15:02.115 --> 00:15:03.298
But clinicians do.

00:15:03.839 --> 00:15:05.442
And when clinicians have a bad day,

00:15:05.481 --> 00:15:06.844
their notes aren't going to all of a

00:15:06.884 --> 00:15:09.128
sudden reflect a good day.

00:15:09.467 --> 00:15:11.591
And so you need verification and you need

00:15:11.652 --> 00:15:13.634
the therapist to actively engage them.

00:15:14.565 --> 00:15:15.504
All right,

00:15:15.585 --> 00:15:17.725
if a healthcare CEO and operator called

00:15:17.765 --> 00:15:18.647
you tomorrow and said, hey,

00:15:18.687 --> 00:15:20.347
we're investing heavily in AI,

00:15:21.067 --> 00:15:23.428
what's the one question you'd ask to have

00:15:23.469 --> 00:15:25.809
them reveal whether or not they're

00:15:25.850 --> 00:15:27.650
building something that restores humanity

00:15:27.730 --> 00:15:30.331
or just another productivity tool?

00:15:32.149 --> 00:15:33.770
Yeah, I would say,

00:15:33.791 --> 00:15:35.871
are you measuring your AI scribe and your

00:15:35.912 --> 00:15:39.493
investment in AI by the patient face time

00:15:39.513 --> 00:15:42.374
with the therapist or by how efficient or

00:15:42.475 --> 00:15:45.135
throughput or how many visits per day?

00:15:45.576 --> 00:15:45.735
You know,

00:15:45.755 --> 00:15:48.537
you're not seeing board decks with patient

00:15:48.576 --> 00:15:49.837
face time and outcomes.

00:15:49.857 --> 00:15:51.839
You're seeing board decks with visits per

00:15:51.879 --> 00:15:52.899
day efficiency.

00:15:52.999 --> 00:15:54.220
And oh, yeah,

00:15:54.240 --> 00:15:55.720
we're a better workplace because our

00:15:55.759 --> 00:15:57.561
therapists don't have to do work at home.

00:15:57.600 --> 00:15:59.682
And so therefore we're trying to solve,

00:16:00.101 --> 00:16:01.682
you know, retention and turnover.

00:16:02.142 --> 00:16:04.042
How about changing the narrative a little

00:16:04.062 --> 00:16:06.263
bit and using AI to help therapists

00:16:06.302 --> 00:16:08.003
produce better outcomes and more effective

00:16:08.062 --> 00:16:10.984
time with their patients as opposed to the

00:16:11.004 --> 00:16:12.043
efficiency trap?

00:16:13.423 --> 00:16:16.205
If an operator gives me a productivity

00:16:16.264 --> 00:16:18.024
number on how much AI helped their

00:16:18.065 --> 00:16:19.845
productivity, the patients per day,

00:16:19.884 --> 00:16:21.046
the visits per hour,

00:16:21.485 --> 00:16:22.966
or notes completed faster,

00:16:23.586 --> 00:16:26.365
they're using AI to accelerate the very,

00:16:26.427 --> 00:16:28.687
very problem that was burning people out.

00:16:30.248 --> 00:16:30.707
In fact,

00:16:31.048 --> 00:16:33.110
I wrote even a couple of years ago

00:16:35.052 --> 00:16:38.654
in the Substack about how my initial fear

00:16:38.674 --> 00:16:40.655
around AI was that it was going to

00:16:40.696 --> 00:16:43.918
be used as the tool to see more

00:16:44.578 --> 00:16:47.061
patients per day or be more productive.

00:16:47.120 --> 00:16:48.442
I didn't realize how prescient that was

00:16:48.462 --> 00:16:50.124
because that's exactly what's happened.

00:16:50.604 --> 00:16:52.225
You're not hearing about how more

00:16:52.265 --> 00:16:54.707
effective a therapist is because people

00:16:54.748 --> 00:16:56.089
aren't getting the heart of the burnout

00:16:56.129 --> 00:16:56.509
program.

00:16:56.528 --> 00:16:58.049
They're trying to control the symptoms

00:16:58.070 --> 00:16:58.831
through efficiency.

00:17:00.386 --> 00:17:01.307
The money question here,

00:17:02.369 --> 00:17:04.690
and we've heard this before on pitches or

00:17:04.710 --> 00:17:05.431
in ad campaigns,

00:17:05.590 --> 00:17:08.333
if AI gives every clinician an extra

00:17:08.413 --> 00:17:09.433
thirty minutes a day,

00:17:10.335 --> 00:17:12.977
what's the worst thing a leader could do

00:17:13.017 --> 00:17:15.157
with those thirty minutes that has been

00:17:15.198 --> 00:17:15.719
given back?

00:17:16.678 --> 00:17:16.837
Yeah,

00:17:16.857 --> 00:17:18.858
the worst thing they can do is tell

00:17:18.898 --> 00:17:20.398
the therapist that they have to see more

00:17:20.459 --> 00:17:20.979
patients,

00:17:21.378 --> 00:17:23.140
that they have to do more compliance

00:17:23.180 --> 00:17:23.640
review,

00:17:23.660 --> 00:17:25.599
that they have to do more documentation,

00:17:25.640 --> 00:17:27.500
that they have to get online and pass

00:17:27.540 --> 00:17:30.260
their HIPAA course for the year.

00:17:30.981 --> 00:17:34.382
Allow the therapist some time to reflect,

00:17:34.823 --> 00:17:36.282
to think about their patient care,

00:17:36.403 --> 00:17:38.202
to mentor, to free them up,

00:17:38.242 --> 00:17:39.503
to do clinical work.

00:17:40.223 --> 00:17:42.503
The thing that drives people to become a

00:17:42.564 --> 00:17:44.464
PT was you want to do clinical work.

00:17:45.182 --> 00:17:46.943
Nobody ever became a physical therapist

00:17:46.963 --> 00:17:48.104
because they wanted to be an expert in

00:17:48.124 --> 00:17:50.605
documentation to ever do be a compliance.

00:17:50.625 --> 00:17:53.164
They didn't know the minute rule existed.

00:17:53.585 --> 00:17:54.746
It's all of those kind of things.

00:17:54.766 --> 00:17:56.306
I almost think we ought to make therapists

00:17:56.326 --> 00:17:58.386
sign prenuptial agreements when they get

00:17:58.426 --> 00:18:00.126
into the profession their first day.

00:18:00.467 --> 00:18:02.567
Right.

00:18:02.688 --> 00:18:03.728
You know, and, you know,

00:18:03.788 --> 00:18:06.308
really expose them to the negatives around

00:18:06.388 --> 00:18:07.449
all of these kind of things.

00:18:07.469 --> 00:18:10.430
These distractors are going to put them in

00:18:10.470 --> 00:18:10.490
a.

00:18:11.430 --> 00:18:12.830
in a period where they're going to really

00:18:12.871 --> 00:18:13.791
be there.

00:18:14.172 --> 00:18:16.053
It's almost like that could be used as

00:18:16.093 --> 00:18:17.713
part of a burnout prevention program.

00:18:17.733 --> 00:18:18.434
All right.

00:18:19.295 --> 00:18:21.036
If the conversation about AI usually

00:18:21.076 --> 00:18:22.037
starts with one question,

00:18:22.136 --> 00:18:23.778
what can technology do?

00:18:24.657 --> 00:18:27.339
Allow Larry to help you think it should

00:18:27.359 --> 00:18:28.059
start somewhere else.

00:18:28.259 --> 00:18:30.122
What should the technology give back?

00:18:30.162 --> 00:18:31.622
Because if AI only makes clinicians

00:18:31.782 --> 00:18:32.442
faster,

00:18:32.462 --> 00:18:34.144
I feel like you think we've missed the

00:18:34.183 --> 00:18:34.544
point.

00:18:34.844 --> 00:18:36.325
If it gives them back their attention,

00:18:36.365 --> 00:18:37.405
presence, humanity,

00:18:37.766 --> 00:18:39.547
then maybe technology becomes the thing

00:18:39.567 --> 00:18:40.587
that helps us care more.

00:18:41.048 --> 00:18:41.428
Again,

00:18:41.689 --> 00:18:43.150
if today's conversation challenged the way

00:18:43.170 --> 00:18:44.171
you think and made you a little

00:18:44.211 --> 00:18:45.230
uncomfortable, good.

00:18:45.392 --> 00:18:46.011
That's the point.

00:18:46.612 --> 00:18:48.413
Subscribe to The Operator on Substack.

00:18:48.713 --> 00:18:50.976
And this is part one of a series

00:18:51.016 --> 00:18:52.396
called Call to Care in the Age of

00:18:52.457 --> 00:18:53.017
AI series,

00:18:53.377 --> 00:18:56.940
which will lead into a re-release of some

00:18:56.980 --> 00:18:57.621
literature as well.

00:18:57.641 --> 00:18:58.701
You want to front tease that?

00:18:59.517 --> 00:19:00.297
Yeah, absolutely.

00:19:00.356 --> 00:19:01.617
So Call to Care was a book I

00:19:01.637 --> 00:19:02.898
wrote in two thousand twenty,

00:19:03.098 --> 00:19:05.299
giving humanity health care back to.

00:19:06.000 --> 00:19:09.082
And it's really how we can take positive

00:19:09.122 --> 00:19:10.942
psychology and the social science,

00:19:10.962 --> 00:19:14.183
social neuroscience and understand that

00:19:14.223 --> 00:19:16.905
the soft skills in health care are the

00:19:16.965 --> 00:19:17.726
hard skills.

00:19:17.786 --> 00:19:19.067
And those are the ones that make a

00:19:19.166 --> 00:19:20.948
huge difference on patient outcomes.

00:19:21.708 --> 00:19:23.788
And so you'll see it's expanded with some

00:19:23.890 --> 00:19:26.931
additional chapters on AI, of all things,

00:19:27.010 --> 00:19:28.051
and virtual health,

00:19:28.092 --> 00:19:29.972
which didn't exist even in two thousand

00:19:30.012 --> 00:19:30.373
twenty,

00:19:30.913 --> 00:19:32.494
along with some new psychological

00:19:32.535 --> 00:19:34.695
principles that have stood the test of

00:19:34.736 --> 00:19:36.336
time from an intervention standpoint,

00:19:36.376 --> 00:19:38.478
including implementation intentions,

00:19:38.838 --> 00:19:40.700
some additional things on mindfulness,

00:19:41.339 --> 00:19:43.000
how to be mindful without feeling like you

00:19:43.020 --> 00:19:43.942
always have to meditate.

00:19:44.602 --> 00:19:46.844
And some other areas that I think people

00:19:46.864 --> 00:19:47.345
will chant.

00:19:47.365 --> 00:19:48.385
So at the end of the day,

00:19:48.405 --> 00:19:49.946
the book's got about twenty five percent

00:19:49.967 --> 00:19:51.749
expanded and updated on material.

00:19:51.769 --> 00:19:52.769
So very excited about it.

00:19:52.789 --> 00:19:53.391
All right.

00:19:53.431 --> 00:19:54.070
More to come with that.

00:19:54.432 --> 00:19:56.153
Until next time, keep operating.

00:19:57.074 --> 00:19:57.714
Absolutely.