Latest / Becker’s Healthcare Podcast / YOLO vs. Death by a Thousand Pilots: AI Change Management Lessons from Emory Healthcare

Transcript

YOLO vs. Death by a Thousand Pilots: AI Change Management Lessons from Emory Healthcare · Becker’s Healthcare Podcast

• 0:00
  Hi, everyone. This is Kedar Amate, cofounder and
• 0:02
  chief medical officer of Qualified Health. Thank you
• 0:04
  for tuning in to the Becker's Healthcare podcast
• 0:07
  series.
• 0:08
  I'm really excited today to be talking to
• 0:10
  doctor Nabil Safdar, the chief AI officer of
• 0:13
  Emory Healthcare and Emory University.
• 0:15
  Today, we're gonna be talking about change management
• 0:17
  and how to help with AI adoption in
• 0:20
  our health care institutions.
• 0:21
  Naveel, thank you for joining me here today.
• 0:24
  Yeah. Thank you, Kedar. I'm looking forward to
• 0:25
  it. Well, Naveel, you and I have had
• 0:27
  a chance to work together a little bit,
• 0:29
  with between Qualified and Emery, in the partnership
• 0:32
  that we have. But let's start by naming
• 0:35
  an AI deployment that you've been particularly excited
• 0:38
  about, particularly proud of at Emery.
• 0:40
  What's worked about that? What's made it successful?
• 0:42
  And then we can get into what made
• 0:43
  it challenging.
• 0:45
  Yeah. Kedar, great question. Plenty that I'm proud
• 0:47
  of, plenty that I'm ashamed of as well,
• 0:50
  but I'm gonna go back
• 0:52
  to one of my,
• 0:53
  first AI deployments,
• 0:55
  which was about eleven years ago at Emory.
• 0:59
  This is now considered pretty routine, but CT
• 1:02
  perfusion for stroke imaging is is a pretty
• 1:05
  common thing now at most stroke centers.
• 1:08
  The the brain attack teams, the stroke teams,
• 1:11
  neurologists, neurocritical care docs, neuroradiologists,
• 1:13
  you know, like to use this, especially since
• 1:16
  we know, you know, brain is time.
• 1:18
  But eleven years ago, this was kind of
• 1:20
  like, you know,
• 1:21
  two guys in a research project just started
• 1:24
  a company,
• 1:25
  and we were actually implementing
• 1:28
  CT perfusion in research mode only.
• 1:31
  And,
• 1:32
  I remember
• 1:33
  there was a,
• 1:35
  a neurologist,
• 1:37
  who emailed me, and I had a conversation
• 1:39
  when he said, can we please use this
• 1:41
  clinically,
• 1:43
  you know, rather than just in research mode?
• 1:46
  And I thought to and I was new
• 1:47
  to the organization. I thought to myself, like,
• 1:49
  you know, jeez. Why don't we use it
• 1:52
  clinically? And what I've discovered was just there
• 1:54
  was no contract to pay for
• 1:57
  the darn solution.
• 1:59
  You know, it was available. Just no there
• 2:01
  was a divide
• 2:02
  culturally between the physicians
• 2:05
  who wanted to use the tools
• 2:08
  and the administrators who knew how to get
• 2:10
  that contract through legal, through security review, all
• 2:13
  those types of things. Mhmm. And I said,
• 2:15
  well, I could figure this out. Like, I
• 2:17
  started calling around administrators. They were like, yeah.
• 2:19
  That's not even that expensive. Let's just do
• 2:21
  it. So we implemented it officially, and we
• 2:24
  all all that changed was
• 2:26
  there was no longer
• 2:28
  the little for research use only,
• 2:32
  you know, label that goes on to,
• 2:35
  the AI. But I I learned a lot
• 2:37
  of things from that. One is we hamstring
• 2:40
  ourselves
• 2:41
  unnecessarily,
• 2:43
  by barriers to adoption that don't really exist.
• 2:47
  It's just that we're not willing to, you
• 2:49
  know, take a little bit risk. I was
• 2:51
  new to the organization, so I could ask
• 2:52
  stupid questions like, who pays for this thing?
• 2:55
  Who's our lawyer? You know? And that helped.
• 2:58
  And then out of that, what happened was,
• 3:01
  you know, it was it was useful for
• 3:03
  the stroke team. It was useful for patients.
• 3:06
  I I think the evidence bears that out,
• 3:08
  but we also found that there were many
• 3:11
  ways
• 3:12
  that it was not useful.
• 3:14
  And, you know,
• 3:15
  some of our
• 3:17
  our our team actually published in a paper
• 3:19
  led by Chung on all the ways the
• 3:21
  CT perfusion
• 3:22
  can be misused, and it could go wrong,
• 3:24
  the artifacts, all the way it goes wrong.
• 3:26
  And then, you know, ten years later,
• 3:29
  in that same group, there's a a whole
• 3:31
  paper on AI governance in that department. You
• 3:34
  know?
• 3:35
  So these things tend
• 3:38
  to have a life of their own. There
• 3:40
  tends to be a flywheel effect, like little
• 3:43
  breaking down of barriers
• 3:45
  compound
• 3:46
  over time so that
• 3:47
  now that seems like a pretty routine thing.
• 3:49
  So I I was look. Was it a
• 3:51
  major technical breakthrough? No. Absolutely not. But for
• 3:54
  me, it was, like, the first time I
• 3:55
  was like, I'm gonna ask some dumb questions,
• 3:57
  and let's figure out if we can do
• 3:59
  it. And I was I'm very proud of
• 4:00
  that. May maybe good guidance for how to
• 4:02
  how to approach,
• 4:03
  AI implementations at scale in almost any organization.
• 4:06
  But tell me tell me a little bit
• 4:07
  about,
• 4:08
  what has been a, you know, big surprise
• 4:11
  for you as you've been now tracking many
• 4:13
  projects as they've gone through your system.
• 4:15
  You know, what surprised you about receptiveness
• 4:17
  of of staff member team members to artificial
• 4:21
  intelligence technologies?
• 4:22
  We hear a lot about health systems being,
• 4:24
  you know, or health care workers being particularly
• 4:27
  nervous about it, worried about job security, worried
• 4:29
  about, you know, trusting the,
• 4:31
  accuracy of these technologies. So, you know, kinda
• 4:33
  how have staff,
• 4:35
  responded to
• 4:36
  the technologies that you and your team are
• 4:38
  trying to deploy at Emery?
• 4:40
  Yeah. Great question. I'll tell you what surprised
• 4:42
  me.
• 4:43
  But this shouldn't have surprised me, but, you
• 4:45
  know, I think it speaks to the circles
• 4:48
  that we all travel in, and it has
• 4:49
  to do with culture. You know, I just
• 4:51
  kind of assumed
• 4:52
  that, hey. Hey. Everyone's talking about AI. This
• 4:55
  is the latest, greatest thing ever. You know,
• 4:56
  once I started using it, I think I
• 4:58
  lost, like, forty eight hours of my life.
• 5:00
  I was like, this is amazing, the generative
• 5:02
  AI stuff. You know? It reminded me
• 5:06
  of when I had my first Apple two
• 5:08
  e, and I had to, you know, program
• 5:10
  things in basic or or load up those
• 5:13
  five inch floppy disks. I felt, like, so
• 5:15
  smart because I was like, oh, it's not
• 5:17
  like the movies. You have to give it
• 5:18
  very specific instructions
• 5:20
  in order for it to do it only
• 5:22
  knows what you tell it.
• 5:24
  And and, I don't know, fifty years later,
• 5:26
  forty years later, you know, now I was
• 5:28
  like, oh my god. What I imagined as
• 5:30
  a child,
• 5:31
  computers would do in science fiction, they're able
• 5:33
  to do.
• 5:34
  Mhmm. So I had a very, I think,
• 5:36
  intuitive understanding, maybe built on some experience
• 5:40
  of what Gen AI tools can do and
• 5:42
  how to interact with them. I mean, you
• 5:44
  gotta take that with a grain of salt
• 5:45
  because everybody's new to this game, but, you
• 5:47
  know, we were early in the games. And
• 5:48
  the circles
• 5:49
  that I traveled in were similar. Right? The
• 5:52
  people that I spoke to were similar. Even
• 5:53
  the folks who were not, you know,
• 5:55
  extensively tech, you know, nerds were were using
• 5:58
  the tools. They were curious.
• 6:00
  But when we released,
• 6:03
  you know, one kind of rag architecture chatbot
• 6:06
  that looked at policies,
• 6:08
  to
• 6:10
  nurses at an entire hospital,
• 6:12
  you know, one of the first things that
• 6:14
  our team recognized was when,
• 6:17
  we were when the users were asking
• 6:20
  the AI chatbot
• 6:22
  questions,
• 6:23
  they were interacting with it like it was
• 6:25
  Google. They were entering keywords rather than prompts.
• 6:28
  And that just never occurred to me that
• 6:30
  you would not know that somebody might not
• 6:33
  know,
• 6:34
  that this isn't, a search engine. This is
• 6:37
  a prompt engine. You it's a chatbot.
• 6:39
  Did you have to educate them differently? Did
• 6:40
  you have to change your approach? Absolutely. We
• 6:42
  had to we had to go through and
• 6:44
  start to give an you know, by analogy.
• 6:45
  Hey. This is like you're talking to a
• 6:47
  person. Give it a specific question, you know,
• 6:49
  talking a little bit about prompt strategies, some
• 6:51
  upskilling, some education.
• 6:53
  I'm not blaming the the nurses, by the
• 6:56
  way. That was total, like, you know, blind
• 6:58
  spot on my team's part and my part.
• 7:00
  You know what I mean? When we we
• 7:01
  pivoted, we started giving them much more basic
• 7:03
  information.
• 7:04
  Why would a busy nurse I mean, unless
• 7:07
  they're in that 20%
• 7:09
  of, you know, early adopters, bleeding edge folks
• 7:12
  who are trying this at home, you know,
• 7:14
  they're busy. Right? They've got busy lives. They
• 7:16
  got busy jobs, and this is one more
• 7:18
  thing we're asking to do. Of course, they're
• 7:20
  gonna use the common framework, a common paradigm.
• 7:23
  It reminds me,
• 7:25
  Zainab Tufekci gave a talk recently,
• 7:27
  at Emory,
• 7:28
  and she talked
• 7:30
  about when, horse,
• 7:32
  carriages were first being replaced. One of the
• 7:34
  early models of cars
• 7:37
  actually had a
• 7:39
  a faux
• 7:40
  horse head on top of it so that
• 7:42
  it would be a familiar,
• 7:44
  you know, paradigm for the drivers and also
• 7:46
  for other horses, which was the majority transportation
• 7:49
  mode at that time.
• 7:50
  And I felt like that's that's what these
• 7:52
  users were doing. It was my fault for
• 7:55
  not recognizing that not every this is a
• 7:57
  journey. Right? And everybody's a different place. You
• 7:58
  have to meet people where they are. Yeah.
• 8:01
  That's an interesting an interesting point. You know,
• 8:03
  giving in this idea
• 8:05
  of making the AI familiar, you know, this
• 8:07
  is the horse, you know, put replacing the
• 8:09
  horse kind of idea.
• 8:10
  Do you do you think we have to
• 8:11
  actually
• 8:12
  meter the dose of AI into our health
• 8:15
  systems? Let let me let me explain what
• 8:16
  I mean by that. You know, like, not
• 8:19
  do the the most radical, you know, Rosie
• 8:22
  from the Jetsons kind of,
• 8:24
  solution right away, but instead
• 8:26
  build our confidence slowly in the technology through,
• 8:30
  you know, what I'm calling metered dosing, like,
• 8:32
  small micro almost microdosing
• 8:34
  kind of the artificial intelligence into our systems,
• 8:38
  especially with clinicians kind of building their trust
• 8:40
  and their confidence as they see, you know,
• 8:42
  ambient
• 8:43
  documentation successes, stacking against, you know, helping with
• 8:47
  prior authorization, stacking against, you know, etcetera, etcetera.
• 8:50
  And then moving towards more complex and more
• 8:53
  wholesale use cases. I mean, do do you
• 8:54
  think this is part of the strategy for
• 8:56
  winning,
• 8:57
  clinicians over and building trust in the technology?
• 9:01
  Yeah, Gurk. Man. Well, so much to talk
• 9:04
  about here. I don't know where to start.
• 9:05
  So, you know, we have done a couple
• 9:07
  of projects, what we call at Emory, and
• 9:09
  and I'll credit, my my colleague and friend,
• 9:12
  Rachel Silverman, with this, like, with the YOLO
• 9:14
  approach. Just big bang, turn it on for
• 9:17
  everybody,
• 9:18
  very little training. We'll see what sticks. And
• 9:20
  in other cases, we have done, like, you
• 9:22
  know, death by a thousand pilots where it's
• 9:24
  like, okay. You only get five users, and
• 9:26
  now there's another ten, fifteen users. Yeah. And,
• 9:30
  I'll tell you what I've seen. You
• 9:33
  know, Kate, are you you know, we've done
• 9:35
  stuff like this before. You know I like
• 9:37
  car and travel analogies. Right? So, you know,
• 9:39
  there's a certain capacity that a highway has.
• 9:42
  Right?
• 9:43
  And
• 9:44
  if you
• 9:46
  overwhelm the capacity of that highway with new
• 9:49
  traffic,
• 9:50
  it's just gonna reject the system, and people
• 9:53
  are gonna go around it. Right? Yeah.
• 9:55
  So that's the YOLO approach, and that's what
• 9:57
  we found. You know what I mean? We
• 9:59
  we said, turn this tool on for everybody.
• 10:02
  Everybody use it. We even did some kinda
• 10:05
  rah rah leadership moments,
• 10:07
  and we plateaued around, like, 30% utilization.
• 10:10
  You know what I mean? Now we're gonna
• 10:11
  go back and and try to do more.
• 10:13
  It was a cheap 30% adoption, meaning, like,
• 10:16
  it didn't cost us a lot to just
• 10:18
  turn it on and and say go for
• 10:20
  it, and you get a certain amount of
• 10:21
  adoption because it's a helpful technology.
• 10:23
  I do think metering it in,
• 10:27
  is valuable, and it depends. Right? So, you
• 10:30
  know, for for the most resistant users,
• 10:33
  those who are feeling maybe the most threatened
• 10:36
  because they a high percentage of their job
• 10:38
  can be automated
• 10:39
  as tasks that are are automated. Not the
• 10:42
  entirety of their job, but a certain portion
• 10:43
  of it.
• 10:44
  Those people that have had bad experiences
• 10:47
  with previous iterations of AI, let's call it.
• 10:50
  Maybe it was,
• 10:51
  you know, NLP two generations ago or something
• 10:54
  like that. You know? Yep. For for them,
• 10:56
  I think metering it in is the only
• 10:58
  way
• 10:59
  that you avoid a total immune rejection
• 11:03
  of this new technology. Right?
• 11:06
  And you gotta do it. It's almost like
• 11:07
  allergy shots. You know? But if something is
• 11:10
  so compelling and so useful that it speaks
• 11:13
  for itself, you know what I mean, like,
• 11:15
  people will see it and they'll say, why
• 11:17
  don't I have it? Then I think you
• 11:19
  start YOLO. You know, you you only live
• 11:21
  once. Let's give it all to them, and
• 11:23
  we'll see who adopts it. Then we'll come
• 11:24
  and get the laggards. You know? So, you
• 11:27
  know, I'd I'd like to commit to one
• 11:28
  or the other. So it's it's a bit
• 11:30
  of toggling between different strategies depending on the
• 11:32
  nature of the of the environment.
• 11:34
  It makes total makes total sense, and I
• 11:36
  I love our blending of car metaphors and
• 11:38
  biology,
• 11:39
  what we do together on some level. But
• 11:41
  maybe, as I come close to the end
• 11:43
  of our time here,
• 11:44
  you know, I I think you and I
• 11:46
  agree that there's gonna be
• 11:48
  lots of different solutions necessary for our health
• 11:51
  system. It's not gonna be one thing. You
• 11:53
  know, AI, there's so many different potential applications,
• 11:56
  maybe at least dozens, if not hundreds of
• 11:58
  potential,
• 11:59
  application areas and things we're gonna need to
• 12:02
  implement.
• 12:02
  So, how do you create a, you know,
• 12:05
  change ready culture in an organization? This is
• 12:08
  this is not gonna be one time with
• 12:10
  one big thing or even a couple of
• 12:12
  small things. This is gonna be a constant
• 12:14
  for us, for the next, you know, at
• 12:17
  least five to ten years of many, many
• 12:19
  things coming. How do you how do you
• 12:21
  prepare an organization like Emory for this constant
• 12:23
  pace of change,
• 12:25
  with these technologies coming online?
• 12:27
  Yeah. You know,
• 12:29
  I do think that we're seeing an acceleration
• 12:33
  of availability
• 12:34
  of new technology driven by AI
• 12:38
  that
• 12:39
  really is unlike what I've seen in in,
• 12:42
  you know, previous parts of my career, and
• 12:43
  others share that. Like, the the availability
• 12:46
  of new technology rapidly, new solutions,
• 12:49
  you know, the velocity
• 12:51
  of it, you know, the volume of it,
• 12:53
  kind of, like, the big data aspect of,
• 12:55
  you know, how much there is is is
• 12:57
  something really new. I think,
• 12:59
  you know, we're gonna have to figure out
• 13:02
  how we manage this chain. I like
• 13:04
  the the persuasion,
• 13:06
  you know, kind of literature, especially Cialdini. And
• 13:09
  Cialdini talks about a couple of levers of
• 13:12
  influence.
• 13:13
  One of them is, authority and social proof,
• 13:16
  or there's authority from top down leadership.
• 13:18
  There's social proof from people who have been
• 13:20
  through it before, who have, validity,
• 13:24
  with their peers,
• 13:25
  but particularly
• 13:26
  commitment and consistency.
• 13:29
  You know, I'm thinking about downtimes. If you
• 13:31
  asked me, hey. Does Emory have a great
• 13:34
  approach to how it manages expected and unexpected
• 13:37
  downtimes
• 13:38
  five or six years ago? I would be
• 13:39
  like, I don't even know that we have
• 13:41
  an approach. But now it's a totally different
• 13:44
  ballgame. And I'm thinking about how you know,
• 13:46
  we have had
• 13:47
  expected downtimes from major upgrades, and we've had
• 13:50
  unexpected downtimes like the, you know, CrowdStrike or
• 13:53
  some of these other things that have come.
• 13:55
  And I'll tell you, it's a different world
• 13:57
  now. We have
• 13:58
  a team of people
• 14:00
  who may be analysts who are consistent. We
• 14:03
  have administrators
• 14:04
  who are consistent. We have stand up, you
• 14:07
  know, incident command center,
• 14:09
  you know, policies and procedures where which are
• 14:11
  very consistent.
• 14:13
  And so we have established culture when it
• 14:16
  comes to downtime. People know what their roles
• 14:18
  are. They know what the expectations are,
• 14:21
  and mostly that's driven from consistency in health
• 14:24
  care. We all get a little bit cynical,
• 14:26
  and we start to say, let's is this
• 14:28
  the latest fad, the latest trend? Is it
• 14:30
  gonna blow over in another, you know, two
• 14:32
  months? Let me wait and see. But if
• 14:34
  it's there consistently and then you also,
• 14:37
  you know, leverage all your leavers, reciprocity,
• 14:41
  scarcity, liking,
• 14:42
  you know, social proof, authority. But that consistency
• 14:46
  and that commitment is what changes culture over
• 14:49
  time. Because, look, we're dealing with literally tens
• 14:52
  of thousands of people from all different walks
• 14:54
  of life, different levels of education, different backgrounds,
• 14:57
  cultural backgrounds,
• 14:59
  even languages.
• 15:00
  So how do you how do you, you
• 15:01
  know, create a consistent common culture?
• 15:05
  It's what you tolerate and what you expect
• 15:07
  in behavior and that people only
• 15:10
  learn that by seeing that over and over
• 15:12
  again. And so that's what we gotta do
• 15:14
  with AI. We've gotta we've gotta have
• 15:16
  consistent
• 15:17
  rollouts, consistent
• 15:19
  communication,
• 15:21
  consistent support from top down leadership,
• 15:24
  consistent engagement of the front lines, and then,
• 15:27
  you know, it will slowly build that culture
• 15:28
  over time. I like that advice. Consistency and
• 15:31
  commitment, that makes a ton of sense. Maybe
• 15:33
  I can ask you a final question just
• 15:35
  in rapid fire here as we get to
• 15:36
  the end, Nabil. If you could give one
• 15:38
  piece of advice to any other health system
• 15:40
  that's trying to embark on this journey,
• 15:42
  on,
• 15:43
  adopting generative AI in their environments,
• 15:45
  What's the one quick fire,
• 15:47
  piece of advice you'd give?
• 15:49
  Yeah. Don't my quick fire advice is do
• 15:51
  not relegate your AI strategy to be a
• 15:55
  service that only responds to tickets.
• 15:59
  And and, you know, if if it's really
• 16:01
  becomes,
• 16:02
  hey. I've got something broken. I need you
• 16:04
  to fix it. Or it it becomes
• 16:07
  a thousand requests from various departments and business
• 16:10
  units. You know, look. We need to be
• 16:12
  responsive to the front lines. Absolutely. We should
• 16:15
  do that, but there needs to be capacity
• 16:17
  reserved
• 16:18
  or top down transformational
• 16:21
  AI strategy,
• 16:22
  and that only comes if you position that
• 16:25
  conversation at the right level
• 16:28
  in the organization.
• 16:29
  So it can't just be
• 16:31
  operational AI governance. It can't just be security
• 16:34
  risk compliance, legal counsel governance.
• 16:37
  It's gotta be true corporate governance when it
• 16:40
  comes to AI, which isn't about
• 16:43
  what we do or how we do it.
• 16:45
  It's about who we are as an organization,
• 16:48
  and that means having AI embedded
• 16:51
  in those corporate discussions, in those small group
• 16:54
  discussions.
• 16:55
  This isn't just self serving. Like, I don't
• 16:57
  have to be personally in that conversation, but
• 16:59
  that conversation has to be happening
• 17:01
  at those top levels.
• 17:03
  Fully agree. This is exactly how we've been
• 17:04
  pursuing AI transformations and health system partners around
• 17:07
  around the country. You know, absolutely fully embedded
• 17:10
  at the top level leadership and with organizations
• 17:13
  that see AI as part of who they
• 17:15
  are, going forward. Makes a ton of sense.
• 17:17
  Last word, Nabil.
• 17:19
  Chief AI officer's new role in health systems,
• 17:21
  you are one of them at a major
• 17:24
  institution.
• 17:25
  What do you think is gonna make your
• 17:27
  fellow CAIOs
• 17:28
  successful
• 17:29
  in this new role?
• 17:31
  Yeah. I'm gonna use another car analogy here
• 17:34
  at Kadar. So with with cars, you and
• 17:36
  and let's talk about a racing team or
• 17:38
  something like that. You know, and with the
• 17:40
  racing team,
• 17:41
  f one or rally, whatever it may be,
• 17:44
  you have your mechanics,
• 17:46
  and those would be akin to the, like,
• 17:48
  really techies, you know, chief AI officers or
• 17:51
  chief data and analytics officers that playing that
• 17:53
  role. Great. They're part of the team. You
• 17:55
  have the the drivers who are actually sitting
• 17:58
  there in the cockpit
• 17:59
  racing their cars or the rally cars or
• 18:02
  motorcycles, whatever it may be. Great. They get
• 18:04
  to say that we, you know, we actually
• 18:06
  made the decisions at the last minute at
• 18:08
  the front line.
• 18:09
  Then you have kind of the safety officials
• 18:13
  sitting on the side, like, with the green
• 18:14
  flags, the checkered flags, the, you know, the
• 18:16
  flags that slow down. And then you have
• 18:19
  the folks who are actually the marketing like,
• 18:21
  you know, all the people selling the advertisements
• 18:24
  on the car. Right?
• 18:25
  At the end of the day, if you're
• 18:27
  a chief a officer and you fill one
• 18:30
  of these
• 18:31
  particular profiles, you're the mechanic, meaning you're very
• 18:34
  technical.
• 18:35
  You're the salesperson, meaning you've got all the
• 18:38
  advertisements and all the connections and you're a
• 18:40
  networker.
• 18:41
  Right? You're the driver, meaning you're the operator.
• 18:43
  You've been in there. You've deployed,
• 18:45
  or you're the safety official. You're a governance
• 18:48
  type. You know? You're a you're a kind
• 18:49
  of ethics and compliance type.
• 18:52
  It doesn't matter. What matters is did you
• 18:55
  get to the finish line fast enough and
• 18:56
  far enough? So we can stop counting, frankly,
• 19:00
  how many
• 19:01
  AI tools we've deployed. I think that's cute.
• 19:03
  I I've I do it myself, so it's
• 19:05
  not a criticism to other systems that tout
• 19:08
  like, hey. We've deployed 200, you know, AIs.
• 19:10
  Everybody's deployed 200 AIs. It's impossible to count
• 19:12
  them. They're embedded. They're like the oxygen we
• 19:15
  breathe. They're like the electricity we use. And
• 19:17
  what matters is, did you get outcomes? And
• 19:19
  so five years from now, we're gonna be
• 19:21
  judged on what outcomes
• 19:24
  measurable or or
• 19:26
  less tangible,
• 19:27
  you know, soft or hard, doesn't matter, whether
• 19:30
  it's green dollars, whether it is time saved,
• 19:32
  whether it's satisfaction,
• 19:34
  did you contribute to? Because nobody really cares
• 19:36
  if you're the best mechanic in the world
• 19:38
  if in the race your car didn't win.
• 19:41
  Totally agree. This is, you're helping us understand
• 19:44
  the outcomes that are gonna matter most, and,
• 19:47
  we're gonna see in five years' time whether
• 19:49
  we'll have another interview, Nabil, five years from
• 19:51
  now. We'll we'll see how you're doing against
• 19:52
  that, performance report card as you're thinking about
• 19:55
  it. But, doctor Nabil Safdar, thanks for joining
• 19:57
  me here on the Deckers podcast.
• 19:59
  Really a pleasure to have you and a
• 20:00
  pleasure to work with you every day. Likewise.
• 20:03
  Thank you, Kedar.