Latest / Becker’s Healthcare Podcast / Bringing AI from Idea to Impact in Healthcare with Nishith Khandwala

Transcript

Bringing AI from Idea to Impact in Healthcare with Nishith Khandwala · Becker’s Healthcare Podcast

• 0:02
  - There's a small circle of people who truly
• 0:04
  - understand what it takes to lead a health
• 0:05
  - system right now. The decisions are enormous.
• 0:08
  - The variables are constantly shifting, and the stakes
• 0:11
  - of getting a call wrong don't stay contained.
• 0:14
  - That's not a complaint. It's a structural reality.
• 0:17
  - And it's why the most effective c suite
• 0:19
  - leaders are deliberate about where they go to
• 0:21
  - pressure test their thinking.
• 0:22
  - Becker's fourteenth annual CEO and CFO roundtable
• 0:26
  - brings together more than 1,500
• 0:28
  - executive level attendees
• 0:30
  - and nearly 400 speakers for four days built
• 0:32
  - around growth, financial sustainability,
• 0:35
  - workforce strategy, and the leadership questions that don't
• 0:38
  - have easy answers.
• 0:39
  - Join us November second through the fifth in
• 0:41
  - Chicago.
• 0:42
  - For the agenda and event details,
• 0:45
  - visit beckershospitalreview.com
• 0:47
  - and click on the events tab in the
• 0:49
  - upper right.
• 0:50
  - This is Scott Becker with the Becker's Healthcare
• 0:53
  - Podcast.
• 0:54
  - I'm thrilled today to be joined by an
• 0:56
  - absolutely brilliant
• 0:57
  - entrepreneur
• 0:59
  - health care founder.
• 1:00
  - We're joined today by Nish Kandwala.
• 1:03
  - And Nish is the CEO and the cofounder
• 1:05
  - of Bunker Hill Health.
• 1:07
  - We're gonna talk about what Bunker Hill does,
• 1:10
  - about Nish's Stanford
• 1:12
  - background, and and a lot more.
• 1:15
  - Nish, can you take a moment and tell
• 1:16
  - us a little bit about yourself, a little
• 1:18
  - bit about Bunker Hill Health, and how you
• 1:20
  - got to the spot of founding this company,
• 1:23
  - that's been funded by some of the most
• 1:25
  - well known venture capital funds in the universe.
• 1:28
  - Talk about how you got here and about
• 1:30
  - yourself and about Bunker Hill Health.
• 1:32
  - Sounds wonderful.
• 1:34
  - Thanks for having me, Scott. For the audience,
• 1:37
  - my name is Nish. I am the CEO
• 1:39
  - and cofounder of Bunker Hill, as Scott mentioned.
• 1:42
  - I was previously a student at Stanford working
• 1:45
  - in the AI for medicine labs over there,
• 1:48
  - where
• 1:49
  - a lot of my work was actually focused
• 1:51
  - on building different AI algorithms for different use
• 1:54
  - cases in health care. We worked very closely
• 1:56
  - with physicians
• 1:58
  - and the school of medicine, and we would
• 1:59
  - build some just absolutely wonderful algorithms
• 2:02
  - that would drive better patient care.
• 2:05
  - And
• 2:06
  - at some point, we had this realization that
• 2:09
  - we had built so many algorithms
• 2:11
  - that seemed to work,
• 2:13
  - but we hadn't actually gotten any of them
• 2:15
  - deployed.
• 2:16
  - That it seemed like there was a lot
• 2:19
  - of buzz around AI, but when the Jabram
• 2:22
  - hit the road, we didn't actually see many
• 2:24
  - of them eventually get deployed.
• 2:26
  - And so we got really worried about this
• 2:28
  - because the whole point of this is to
• 2:31
  - not just publish papers, it's to actually
• 2:33
  - drive better patient care, drive better operational efficiencies,
• 2:37
  - and we were just not seeing that translation
• 2:39
  - happen.
• 2:40
  - And we also saw
• 2:42
  - that this problem was only gonna get worse
• 2:46
  - with large language models. When I was a
• 2:48
  - student, large language models weren't a thing, so
• 2:50
  - it actually took us three to six months
• 2:52
  - to build a model
• 2:54
  - for a given use case. But now with
• 2:56
  - large language models, you could build those models
• 2:58
  - even faster. You could build those use cases
• 3:00
  - even faster. But that problem of a lot
• 3:03
  - of potential but not so much actually live
• 3:06
  - in production use, that problem only got exacerbated.
• 3:09
  - So we spun out Bunker Hill with the
• 3:11
  - sole goal of bringing
• 3:13
  - AI from
• 3:14
  - people's might ideas and bringing them into a
• 3:17
  - clinical practice and into production use.
• 3:20
  - Since then, we are now live with over
• 3:22
  - 30 different health systems and have raised around
• 3:25
  - $60,000,000
• 3:26
  - from some of the best venture capitalists like
• 3:28
  - Sequoia Capital and Khosla Ventures.
• 3:31
  - Take a second, Don. Give the audience a
• 3:33
  - sense. You're working with 60 different health systems.
• 3:36
  - Mhmm. Give us a sense of what Bunker
• 3:38
  - Hill Health actually does. What do you folks
• 3:40
  - do so people could just get that much
• 3:42
  - of a clearer sense to that? Of course.
• 3:44
  - So we are an AI agent platform, and
• 3:46
  - it's a way for clinical and operational teams
• 3:49
  - within health systems to build AI agents that
• 3:53
  - end up doing the work end to end.
• 3:55
  - So today, you know, you everyone has a
• 3:57
  - lot of ideas around how they might want
• 3:59
  - to use chat g p t, for example,
• 4:01
  - in their day to day work within the
• 4:03
  - health system, whether that's for something like enabling
• 4:06
  - care pathways on the clinical side to something
• 4:09
  - like helping with,
• 4:10
  - triaging patients in a referral queue for patient
• 4:13
  - access to something on the prior auth side
• 4:15
  - to something on the registry data abstraction side.
• 4:17
  - There are so many of these use cases,
• 4:20
  - but we have found that health systems struggle
• 4:22
  - to actually get those ideas about how
• 4:25
  - technologies like chat g p t can enable
• 4:27
  - with those use cases actually live into production
• 4:30
  - use. So what we have built is this
• 4:32
  - platform. We call it Carebricks,
• 4:34
  - and it's a way for clinical and operational
• 4:36
  - teams to work with us or to build
• 4:38
  - them by themselves by taking our platform and
• 4:41
  - building AI agents that end up doing the
• 4:43
  - work end to end. Each of these use
• 4:45
  - cases that I mentioned and many more,
• 4:48
  - you can probably distill them down to three
• 4:50
  - components.
• 4:51
  - There is one component where you need to
• 4:53
  - get access to data from multiple systems of
• 4:55
  - record.
• 4:56
  - Think
• 4:57
  - EHR. Think ERP. Think other sources of records
• 5:00
  - like clinical guidelines, payer policies, things like that.
• 5:04
  - That's one component.
• 5:05
  - Another component is that of actually running the
• 5:08
  - large language models on top of,
• 5:10
  - that data, and then you have a final
• 5:12
  - component of actually taking actions
• 5:15
  - on top of the output of AI. This
• 5:16
  - is things like writing back to the EHR,
• 5:20
  - things like sending patients' messages or physicians' messages.
• 5:23
  - And so this platform
• 5:25
  - allows
• 5:26
  - clinical and operational leaders to arrange
• 5:29
  - and sort of and
• 5:30
  - assemble
• 5:31
  - these components,
• 5:32
  - these bricks as we call them, a bit
• 5:34
  - that's where the name Carebricks comes from, to
• 5:36
  - assemble these bricks
• 5:38
  - in a way that we can help automate
• 5:40
  - many of those clinical and operational use cases.
• 5:44
  - And take a moment, Nish, on how do
• 5:46
  - systems work with you?
• 5:48
  - Who internally
• 5:49
  - works with your platform?
• 5:52
  - How much help do they get from your
• 5:54
  - team?
• 5:55
  - How does this work in practice?
• 5:58
  - Does the team internally have to be all
• 6:00
  - software engineers?
• 6:02
  - Or or or how does this work with
• 6:04
  - the people at the system
• 6:06
  - so to make this usable for the system
• 6:08
  - without them having to all be computer engineers?
• 6:10
  - Or do they?
• 6:11
  - Yeah. No. That's a great question. So we
• 6:13
  - think that this is actually a great time
• 6:16
  - to make an operational change as well. So
• 6:18
  - we partner when we work with the health
• 6:20
  - system, we are typically working with,
• 6:23
  - someone like a chief AI officer or a
• 6:25
  - chief medical officer where we partner very closely
• 6:28
  - with them to have Carebricks become their AI
• 6:30
  - platform.
• 6:31
  - And we create this governance structure when when
• 6:34
  - when there are new use cases
• 6:36
  - that are coming from the grassroots or when
• 6:38
  - we introduce new use cases that we have
• 6:41
  - found, that receive a lot of love at
• 6:43
  - other health systems.
• 6:44
  - The chief AI officer, the chief medical officer
• 6:47
  - is able to do a very quick sniff
• 6:49
  - test and see if that is of interest
• 6:51
  - to their organization.
• 6:52
  - And if that use case does pass that
• 6:54
  - sniff test,
• 6:55
  - then the chief AI officer
• 6:57
  - introduces
• 6:59
  - us to the problem owners, those use case
• 7:01
  - owners at their health system. And then our
• 7:04
  - forward deployed team works very closely
• 7:08
  - with the subject matter experts, those problem owners,
• 7:11
  - to rapidly build AI agents that help them
• 7:13
  - for their particular problem. And so to give
• 7:16
  - you a very concrete example,
• 7:18
  - from the first time we meet with a
• 7:19
  - problem owner or subject matter expert at a
• 7:22
  - hospital
• 7:23
  - to going live, it takes about two to
• 7:25
  - six weeks end to end, for a net
• 7:28
  - new use case. And so we don't expect
• 7:31
  - the health system to have any engineers
• 7:33
  - responsible for this. We just, expect that we
• 7:36
  - will have a collaboration
• 7:38
  - where our team can rapidly build these AI
• 7:40
  - agents on our platform for those use cases.
• 7:43
  - And what does your forward deployed team look
• 7:46
  - like? Yeah. Because that's so critical, that intersection
• 7:49
  - with the with the client hospital. Because we
• 7:52
  - see so many people that have great technology.
• 7:54
  - But if you don't have that deployment team
• 7:56
  - that's really working hand in hand, it it's
• 7:59
  - almost what you talked about originally when you
• 8:01
  - were back in the day, a graduate student
• 8:03
  - at Stanford. Yeah. And people had ideas, but
• 8:05
  - they weren't really getting put to work. Talk
• 8:07
  - about that integration with the forward deployed team
• 8:09
  - to make this
• 8:11
  - as useful as possible.
• 8:12
  - Yeah. I think our forward deployed team is
• 8:15
  - incredibly curious
• 8:16
  - and is very willing to be collaborative.
• 8:19
  - And so to give you an idea, our
• 8:21
  - VP of forward deployed,
• 8:23
  - he has a background in consulting as well
• 8:25
  - as in start ups. He studied applied math
• 8:28
  - at Brown University.
• 8:29
  - And when he was the only forward deployed,
• 8:33
  - lead that we had,
• 8:35
  - he would, you know, fly out to the
• 8:37
  - health systems that we partner with, meet with
• 8:40
  - the stakeholders, the problem owners. And just if
• 8:42
  - they are already doing a process, but just
• 8:45
  - manually,
• 8:46
  - we would observe what they're doing, shadow them,
• 8:48
  - and try to basically mimic that workflow on
• 8:51
  - our platform, obviously, in a more automated way.
• 8:54
  - Or if it's a process
• 8:56
  - that they haven't the health health system hasn't
• 8:58
  - implemented yet but is interested in doing so,
• 9:01
  - then it's a long form conversation that we
• 9:03
  - have and then very close iteration and collaboration
• 9:06
  - on it as we go live with that
• 9:08
  - use case.
• 9:09
  - How much is the use of AI accelerating
• 9:12
  - in health systems today? What do you sort
• 9:14
  - of see now,
• 9:15
  - use cases? And over the next three to
• 9:17
  - five years, where you sort of seeing the
• 9:19
  - most interest
• 9:20
  - in in in how are things accelerating or
• 9:22
  - cascading?
• 9:24
  - So I think of use cases actually in
• 9:27
  - the form of, like, a Maslow's hierarchy of
• 9:29
  - needs at a health system,
• 9:31
  - where you have your base foundational needs that
• 9:33
  - need to be met. And then as you
• 9:35
  - go towards the top of the pyramid, you
• 9:37
  - hit those enlightened use cases that are very
• 9:40
  - cool, that are very awesome. They make a
• 9:42
  - lot of sense from an ROI perspective,
• 9:44
  - but they are not foundational in the nature.
• 9:47
  - So, like, an example of a foundational use
• 9:49
  - case would be, oh, wow. With the new
• 9:52
  - HR one bill,
• 9:54
  - the definition of who is considered a Medicaid
• 9:57
  - patient is about to change,
• 9:59
  - and our hospital
• 10:01
  - might now have a lower percentage of Medicaid
• 10:03
  - patients,
• 10:05
  - than what is required to be eligible for
• 10:07
  - three forty b discounts.
• 10:09
  - And suddenly, you are at risk as a
• 10:12
  - hospital of losing tens, if not hundreds, of
• 10:14
  - millions of dollars,
• 10:16
  - that are afforded by the three forty b
• 10:18
  - program. And so that is like a foundational
• 10:20
  - need. It is very difficult to care about
• 10:23
  - other things when you have such a foundational
• 10:25
  - need that is unmet. So that will be
• 10:27
  - a great use case to anchor our collaboration
• 10:30
  - on where you could have an AI agent,
• 10:32
  - for example,
• 10:33
  - looking at every patient that is underinsured or
• 10:36
  - uninsured,
• 10:37
  - seeing which of those patients would actually be
• 10:39
  - eligible for Medicaid,
• 10:40
  - and getting them enrolled with the help of
• 10:42
  - some assistance,
• 10:43
  - for example. So I think of these use
• 10:46
  - cases as these are, like, the foundational use
• 10:48
  - cases at the base of the Maslow's hierarchy
• 10:51
  - pyramid, and then you have use cases at
• 10:53
  - the very top that are, like, you know,
• 10:55
  - incredibly frontier care pathways that you have in
• 10:58
  - mind. Like, at the for example, at the
• 11:00
  - Cleveland Clinic, an example of a frontier use
• 11:02
  - case is one where they are doing trials
• 11:05
  - to understand if patients with moderate aortic stenosis
• 11:08
  - would benefit from from earlier intervention, for example.
• 11:12
  - And so they are very interested in a
• 11:14
  - care pathway specifically
• 11:16
  - for a frontier care pathway like that, and
• 11:19
  - that would be an AI agent that helps
• 11:21
  - with patient navigation and follow-up, for example. And
• 11:24
  - so I think of it in terms of
• 11:25
  - when we partner with the health system, we
• 11:28
  - want to get a few use cases that
• 11:30
  - cover the foundation, a few use cases that
• 11:33
  - span across that pyramid going all the way
• 11:35
  - up to the sort of, the top of
• 11:37
  - the the top of the pyramid into the
• 11:39
  - enlightened use cases.
• 11:41
  - Thank you. Absolutely fascinating. They go up the
• 11:43
  - Maslow's hierarchy type of perspective, breaking it down
• 11:46
  - to what's foundational versus what's frontier, what's absolutely
• 11:48
  - needed versus what's nice to have or next
• 11:50
  - level. Exactly. Talk a little bit about what
• 11:53
  - are you most focused on and excited about
• 11:56
  - this next coming year
• 11:58
  - both in in your work with your health
• 12:00
  - systems, your work with your team, your work
• 12:02
  - with your
• 12:03
  - venture capital funds? What what are you most
• 12:05
  - excited about and focused on as you as
• 12:07
  - you work for the next six to twelve
• 12:08
  - months?
• 12:10
  - I think it's two things. The first one
• 12:13
  - is actually something that I hope we can
• 12:15
  - develop as an ecosystem
• 12:16
  - as a whole,
• 12:18
  - and that is how do you actually
• 12:21
  - look at AI as an a as a
• 12:23
  - as an enterprise strategy? What I mean by
• 12:25
  - this is, you know, as we think about
• 12:28
  - how we even want to make a business
• 12:30
  - model around
• 12:31
  - our platform,
• 12:32
  - you know, one approach would have been to
• 12:35
  - sort of charge use case by use case.
• 12:38
  - But I can tell you that it creates
• 12:40
  - a lot of friction when every time there's
• 12:42
  - a new use case in mind,
• 12:44
  - you have to go back to the legal
• 12:46
  - team, negotiate a price point, get that down,
• 12:48
  - and that sort of creates a,
• 12:51
  - you know, loss of momentum over several months
• 12:54
  - as an example. So that's, like, you know,
• 12:55
  - a business model that is not really what
• 12:57
  - we really appreciate because it really reduces
• 13:00
  - the the sort of how much you can
• 13:02
  - use AI for different use cases. So where
• 13:04
  - we are converging towards and we were where
• 13:06
  - we would love to develop this ecosystem towards
• 13:09
  - is thinking of AI similar to how you
• 13:12
  - think of electricity.
• 13:13
  - Just like when you use electricity,
• 13:16
  - you pay for watts,
• 13:18
  - for every unit of consumption, in a similar
• 13:20
  - way, we hope that
• 13:22
  - the ecosystem start starts pricing AI similar to
• 13:25
  - electricity
• 13:26
  - where you pay for tokens of consumption
• 13:28
  - instead of,
• 13:30
  - instead of on a per use case basis.
• 13:32
  - What that will result in is this rapid
• 13:35
  - explosion of usage of AI within a health
• 13:38
  - system where you can go from, you know,
• 13:40
  - zero to twenty, thirty plus use cases without
• 13:43
  - having to constantly think about, oh, wow. We
• 13:46
  - need to go back to the legal team
• 13:47
  - again and again. So that's the first thing
• 13:49
  - that I'm really focused on, which is how
• 13:51
  - can we shape this ecosystem to view AI
• 13:54
  - similar to electricity as opposed to a host
• 13:57
  - of point solutions.
• 13:59
  - That's number one. And number two is actually
• 14:02
  - around this the speed at which AI itself
• 14:05
  - is improving and what capabilities it affords actually
• 14:08
  - unlocks a whole host of use cases that
• 14:11
  - previously were just not possible. So what I
• 14:13
  - am focused on internally within the company is
• 14:16
  - over the course of the last six months,
• 14:18
  - AI has progressed so much with new technologies
• 14:21
  - or new capabilities like computer use. How do
• 14:24
  - we capture those advanced capabilities in unlocking even
• 14:27
  - more use cases
• 14:28
  - with our for our partners and unlocking even
• 14:31
  - more value creation with them. And so those
• 14:33
  - are the two things that I'm really focused
• 14:35
  - on. One is viewing AI's electricity,
• 14:38
  - and two is how do we keep up
• 14:40
  - with the pace of the broader,
• 14:42
  - inherent AI development that's going on.
• 14:45
  - You've had a chance over the last few
• 14:48
  - years to work with some of the highest
• 14:50
  - end venture capital names in the world,
• 14:53
  - Sequoia,
• 14:54
  - Khosla Ventures, and so forth.
• 14:56
  - Give people just some quick insight as to
• 14:59
  - what it's like to work with those
• 15:02
  - funding sources, those people. What is it like?
• 15:05
  - I mean, you're dealing with really bright people.
• 15:07
  - Yes. How do you connect with them?
• 15:10
  - Any thoughts or advice about
• 15:12
  - people working with funds like that? Yeah. No.
• 15:15
  - No. That's a that's a great point. So
• 15:17
  - even at Sequoia, we work with Alfred Lin,
• 15:20
  - who is the head of the Sequoia
• 15:22
  - team. And at and at Khosla, we work
• 15:24
  - with Vinod Khosla himself. He is,
• 15:27
  - obviously,
• 15:28
  - an entrepreneur of of our generation.
• 15:31
  - And
• 15:33
  - so what I'm the reason I shared this
• 15:35
  - is because
• 15:36
  - those two guys, they have nothing left to
• 15:38
  - prove.
• 15:39
  - So when they think about investing in the
• 15:41
  - future, investing in companies,
• 15:43
  - they are not trying to think of, oh,
• 15:45
  - how can I make a quick buck? They
• 15:47
  - are thinking about this in the sense of
• 15:49
  - how can we invest in ideas, in companies,
• 15:51
  - in founders, in teams
• 15:53
  - that can really change the shape of a
• 15:55
  - ginormous industry.
• 15:57
  - And that's how they are looking at us.
• 15:58
  - So when they invested in us, when we
• 16:01
  - were chatting with them and we continue to
• 16:03
  - work with them,
• 16:04
  - where they were coming from is, look, today,
• 16:07
  - health care is usually criticized to be the
• 16:09
  - last industry to adopt new technologies,
• 16:12
  - and we have always criticized health care to
• 16:15
  - be one where things are not nimble, things
• 16:17
  - take a lot of time. And how they
• 16:19
  - see Bunker Hill, how they see our platform
• 16:21
  - Carebricks,
• 16:22
  - is actually a way to make health care
• 16:24
  - more iterable.
• 16:25
  - How can you make it such that the
• 16:27
  - cost of iteration
• 16:29
  - reduces down close to zero as possible? How
• 16:31
  - can you make it such that the speed
• 16:33
  - of iteration,
• 16:34
  - in health care increases as much as possible
• 16:37
  - so that when you have a new idea,
• 16:40
  - it does not take half a decade or
• 16:42
  - a decade to get that into widespread use,
• 16:44
  - that you can make it a lot more,
• 16:47
  - sooner and you can do it in a
• 16:48
  - way that feels very nimble and that feels
• 16:51
  - very,
• 16:52
  - very fast and safe. And so they think
• 16:55
  - of that as if Bunker Hill can do
• 16:56
  - that, it unlocks a whole host of value
• 16:59
  - creation for the entire world. And so that's
• 17:01
  - why they started working with us and we
• 17:04
  - continue to work very closely with them.
• 17:07
  - And what you know, unlike
• 17:09
  - potentially other investors that that we that we
• 17:11
  - know,
• 17:12
  - where they encourage us to to do is
• 17:15
  - what they encourage us to do is to
• 17:16
  - actually swing big, which is don't make compromises
• 17:20
  - just so that you can have a a
• 17:23
  - quick win or, you know, a short term
• 17:25
  - win as opposed to compromising for the larger
• 17:27
  - vision. Take bigger risks. Go bigger.
• 17:30
  - We are here, you know, we are here
• 17:32
  - to support your grand vision. And so I
• 17:34
  - think I feel very privileged in the sense
• 17:36
  - that these investors are these partners we know
• 17:40
  - than Alfred are really able to push us
• 17:41
  - to think big. And, of course, you know,
• 17:44
  - they just come with tons of experience about
• 17:46
  - how to build
• 17:48
  - enormous companies. Like, Sequoia obviously invested in Apple,
• 17:51
  - Google, Airbnb, DoorDash, and Khosla was one of
• 17:54
  - the first investors in OpenAI.
• 17:56
  - They have seen how big companies get built,
• 17:59
  - and and they they really help us make
• 18:00
  - sure that we can learn from how those
• 18:03
  - companies got built.
• 18:05
  - That's fascinating. I mean, it's it's so many
• 18:07
  - people at the end of the day, I
• 18:08
  - count myself as one of them, are more
• 18:10
  - incrementalist
• 18:11
  - than thinking big people. Mhmm. How does that
• 18:14
  - translate from I mean, you're you're obviously
• 18:17
  - brilliant person,
• 18:18
  - straight a's at Stanford in a graduate program,
• 18:21
  - you know, double Stanford graduate.
• 18:23
  - How do you train your mind and work
• 18:25
  - with people to think big
• 18:27
  - versus looking at the next thing? How do
• 18:30
  - you how do you sort of combine those
• 18:31
  - two thoughts,
• 18:32
  - short and big, small and long term, big
• 18:35
  - thinking?
• 18:36
  - How how do you sort of do that
• 18:38
  - as a a founder, entrepreneur,
• 18:40
  - CEO, leader?
• 18:42
  - How do you evolve to do that to
• 18:44
  - do that well? Is is that something that
• 18:45
  - you evolve into or that naturally you see,
• 18:48
  - or how do you look at that? That
• 18:49
  - is such a great question.
• 18:51
  - You know, I I think that
• 18:54
  - we see this in our ecosystem too. I
• 18:57
  - think there's a school of thought where it's
• 18:59
  - like AI has
• 19:01
  - enormous potential
• 19:02
  - for health care,
• 19:04
  - but it also comes with a lot of
• 19:06
  - risks.
• 19:07
  - So we gotta get everything right
• 19:10
  - on the first try, and we gotta make
• 19:12
  - it such that the first time we do
• 19:13
  - it, it's correct and we can execute on
• 19:16
  - that. I think that that's a recipe for
• 19:19
  - failure, especially
• 19:21
  - when this is such a new industry and
• 19:23
  - it's such a fast moving industry.
• 19:25
  - Like, none of us here, no one in
• 19:27
  - this world
• 19:28
  - can claim that, oh, we have five years
• 19:31
  - or ten years of experience
• 19:33
  - deploying AI
• 19:35
  - because that AI in its current form did
• 19:37
  - not even exist
• 19:39
  - more than a couple of years ago. And
• 19:40
  - so I think it's important to almost take
• 19:43
  - a step back and be self aware
• 19:46
  - that we are not gonna get everything right
• 19:48
  - as an industry
• 19:49
  - on day one
• 19:50
  - despite
• 19:51
  - the potential for AI being so huge.
• 19:54
  - And when you have that sort of awareness,
• 19:58
  - what that almost translates to is that you've
• 20:01
  - gotta set up yourself in a way
• 20:04
  - where not everything feels like a one way
• 20:06
  - door.
• 20:07
  - Not everything feels like a three to five
• 20:09
  - year commitment when you know it's risky. Not
• 20:12
  - everything feels like a huge gaping hole in
• 20:15
  - your p and l.
• 20:16
  - It should
• 20:18
  - feel like
• 20:19
  - the stakes are low. It should feel
• 20:21
  - like it's the cost of iteration is lower.
• 20:24
  - And so
• 20:25
  - where I like to sort of think about
• 20:27
  - this, not just in terms of our platform,
• 20:29
  - but also internally,
• 20:31
  - is how can we make the cost of
• 20:33
  - making mistakes as little as possible?
• 20:35
  - How can we make the cost of iteration
• 20:37
  - as little as possible? How can we increase
• 20:39
  - the speed of iteration?
• 20:41
  - And when you have that,
• 20:43
  - what that results in is that you make
• 20:46
  - very quick strides
• 20:47
  - towards your ultimate vision.
• 20:49
  - I think you'd you'd dream big, but then
• 20:52
  - you know where you are today as an
• 20:54
  - industry, and then you have a actual path
• 20:57
  - that you can chart from where you currently
• 20:59
  - are to where you want to be and
• 21:01
  - where you see the potential being. And you
• 21:03
  - make it such that you can take small
• 21:05
  - steps quickly,
• 21:07
  - iterate,
• 21:08
  - backtrack if it doesn't work, and continue
• 21:11
  - to move towards
• 21:12
  - that end state. And so I think it
• 21:14
  - ultimately just boils down to how can we
• 21:17
  - make it such that you can iterate as
• 21:18
  - rapidly as possible.
• 21:21
  - No. And I and I love that perspective
• 21:23
  - on thinking. So I'm I'm very much an
• 21:25
  - incrementalist,
• 21:26
  - an iterative type person,
• 21:28
  - and and and, like, I you know, if
• 21:29
  - I'm writing something, it's gonna go through 10
• 21:31
  - drafts till I get it to where I
• 21:32
  - want it to be. And one of the
• 21:34
  - worst things that a hiring manager could do
• 21:36
  - at one of the companies I work with
• 21:38
  - is get scared of hiring because they've not
• 21:40
  - hit it perfectly.
• 21:42
  - Because, like, I don't want people to get
• 21:44
  - gun shy because they missed on a hire,
• 21:45
  - but they don't hire the next hire. That
• 21:48
  - life and almost everything we do is a
• 21:49
  - work in process. I mean, one of the
• 21:51
  - things you point to is the epics of
• 21:52
  - the world, the apples of the world, and
• 21:54
  - one of the things they've gotten very right
• 21:56
  - is its constant improvement. It it is minimum
• 21:59
  - viable and then constant improvement.
• 22:01
  - And what was Apple
• 22:03
  - fifteen years ago is so different than their
• 22:06
  - operating system today. And the same thing with
• 22:08
  - Epic, it was very much an an incremental
• 22:11
  - incremental type of effort going big, but realizing
• 22:14
  - it's a work in process constantly and it
• 22:16
  - never stops. And that seems to mirror some
• 22:18
  - of the thinking that you have.
• 22:20
  - Yeah. You'd look at the first iPhone that
• 22:22
  - was released
• 22:23
  - to the 17 Pro Max right now,
• 22:26
  - and you see that there's just been, you
• 22:28
  - know, decades of innovation that has been,
• 22:31
  - that has been born through through its cycles
• 22:33
  - of iteration.
• 22:35
  - Really fascinating.
• 22:36
  - Nish, I wanna thank you for joining us.
• 22:39
  - Again, Nish Kanwala,
• 22:41
  - CEO, cofounder of Bunker Hill Health,
• 22:44
  - one of the smartest people that I get
• 22:45
  - to visit with and talk to. What a
• 22:48
  - pleasure to visit with you, Nish, and I
• 22:49
  - look forward to visiting with you some more,
• 22:51
  - and and continued good luck in your journey
• 22:53
  - and what you're doing. Just fantastic. Thank you
• 22:55
  - so much, Scott. Really nice.