01:06Hello everybody, and welcome back to Palantir's fifth edition of AIPcon
02:12My name is Amit
03:18I've been covering the company for the better part of three years now, so I'm honored and humbled to be invited again to go backstage on how Palantir is deepening their customer relationships via AIP
03:27I'm here today with two wonderful Palantir employees, Natalie Schneider and Jack Dobson
03:31You might remember them from the fourth edition of Palantir's AIPcon
03:34Nat and Jack, it's been three months
03:37What's up? It's been good
03:39It's great to be back
03:41Firstly, fifth edition of AIPcon, but obviously all new
03:43We're starting from scratch all over again, as we always do
03:45I mean, it's kind of funny, every time we do these, we sort of, obviously, you guys..
03:48See the show
03:51Our customers connect
03:54It's an incredible event, and then we sort of pack up and go back to, like, boots on the ground effectively
03:56I've been back with a number of our implementation teams, really focusing on our Builder Movement as well
04:00So, our sort of developer community, getting them building on the platform
04:04We rebuild AIPCon from the ground up every time based on what we're seeing out there, what our customers are embracing, and of course, as a product develops as well
04:11So, it's kind of fun to have a whole set of brand new stories and new ways to tell them as well, right? Absolutely
04:19And for the past three months, what I've had the opportunity to do is really dig in with a lot of our teams
04:23Customers around their journeys, specifically with the builder movement that you mentioned
04:27We've had a lot of product updates that we'll get into a little bit later, but it's been very exciting and very invigorating to be able to craft those narratives with a number of our customers
04:39I have actually been working with a lot of our healthcare providers, which has been a really interesting experience
04:44I mean, healthcare, if you think about it, involves massive amounts of data sets, and really what I keep coming back to is that human element
04:52And that's something that I think we'll see not only today, but continuing through the next few months
04:56I mean, it's something that we continuously build back to, and our healthcare..
05:01Providers are really great to work with
05:06And one of the most interesting things that has come up from one of my conversations is that we've actually transformed how one of our providers thinks about time
05:11It's become a Palantir Unit of Time, or PU, as we can call it
05:16What that translates to is essentially a 40-minute interval from problem identification to problem solution
05:22That entirely exists because Palantir has enabled them to do that
05:27So, the building of a completely new time interval — I think that's pretty impressive
05:30That's incredible
05:33And I mean, I guess, Amit, to have you back here, it's great to see you again
05:37Like, we're talking about kind of our view internally, right? What's it been like for..
05:40You? What's the kind of community side of this? What's the perception change over the past three months? So, the past three months have been a crazy ride
05:48As someone who's been analyzing the company for the past three years, I think what has happened in the past three months has uniquely been deepening the relationship with this organic community that, for some reason, over the past couple of years, cannot get enough of learning more about Palantir
06:00I'm wearing a hat right now from the new Palantir merch store because of Ileano
06:04Shout out to Ileano
06:08He has reinvigorated the community by representing the brand, mission, and values through merchandise, and he's made it his mission to say, "Hey, we are actually going..." To do this and put it out into the world." We've got Chad, who unfortunately was not able to be here today, but he spends his time actually responding to all of our silly questions on Twitter every single day about what the ontology is because he cares about communicating with us as people who are just trying to figure out more about the company
06:31I think that deepening of the relationship has been one of the most special things because it's not easy for a company to have the attention of the masses, and for some reason, the masses truly care about Palantir
06:42I mean, even last week, we had Matt Babbitt, who is Palantir's Head of Energy, come on a Twitter Spaces for an hour out of his busy schedule and talk about Palantir's..
06:49Oil and gas business and how they worked with BP exclusively for 10 years
06:53And then we have this extension of the partnership that was announced this week for another five years with BP
06:56So it's been incredible
06:58It's been exciting
07:03As someone covering the company, you know, three years ago, things have changed
07:08And then the biggest thing last Friday, I was doing a live stream on my YouTube channel, and 4,000 people were with me to find out that the S&P 500 has decided to admit Palantir into the index
07:13And for a company that was thought to fail, to never scale, to be recognized as one of the top 500 companies in America, I think it's just special
07:17It's awesome, 100%
07:21I mean, look, S&P 500 inclusion—you still haven't got one
07:23Of those hats, though
07:25We keep wondering why you get merchandise
07:27I don't know how I'm not getting these
07:29But I do think the funny thing, and you mentioned there, is the work that Chad has been doing on the community side of things
07:36The questions that I ask internally about how to use a certain part of a product or how to solve a problem, whatever it might be, all the messages I get back from Chad are, "Go and ask it in the dev community
07:43Go and get those questions out there." If you've used the community, it's community.paloaltonetworks.com
07:47Basically, it's a way by which you can ask questions and be connected with other users on the platform and Palo Alto Networks as well
07:54Our entire company is on there right now answering questions, all in the public
07:58It's a complete step change in how we solve problems and how we communicate with our users
08:02And if you remember last AI Peacon, that's when we launched that developer community
08:06The amount of growth in such a short period, just three months, has been incredible
08:11The engagement we're seeing, not only externally but internally as well, has been truly impressive
08:17Yeah, people care about Palantir, and I think, as someone who has been outside the company for the past three months, people only care more about Palantir
08:23That's because Palantir is giving them a reason to care
08:27And that's why it's super exciting to be at AIP Con
08:29Some of the biggest themes you talked about, Jack, are..
08:31Builders and developers
08:33Can you talk to us about some new products that we have? Yeah, 100%
08:36I mean, this has been a theme for a while now
08:38Previously at AIPCon, we were talking about OSDK and that builder movement
08:40We're doubling down, if not tripling down, on that approach and mindset
08:44So the developers in the platform have never been more empowered to build on top of the ontology and, most importantly, to develop their own custom applications and workflows
08:49For example, VS Code workspaces
08:53Customers are actually building their own custom mobile applications within the platform
08:58I mean, for a lot of our customers, there are some organizations that are completely changing the way they go to market based..
09:04They are building on Palantir, focusing on the products they serve to their customers, fully architected, devised, and deployed using Foundry and AIP
09:11That's a complete step change, where previously the ontology was somewhat confined within that ecosystem
09:17Now, it’s a foundation on which you can build anywhere
09:23One customer that immediately comes to mind for me is Selkirk
09:30For those who may not be familiar with that company, they are a packaged equipment brand
09:36That might not be the first thing that comes to mind when thinking about innovation for customer experiences, but they have fully embraced and opted into Palantir
09:43Just a couple of weeks ago, they were building with us in New York
09:49They're here on site today at AIPcon, and they have ten times their development time, and they are delivering those experiences to their customers in a new, unique way, fully powered by Palantir
09:55Another customer that comes to mind for me as well is Timefold, and I know that you've worked with them as well
10:00Yes, one hundred percent
10:04So, Timefold provides scheduling optimization algorithms and models
10:09It's incredible work, and it's incredibly powerful work, but with these kinds of models, their value and their power are only as strong as the platform within which you actually deploy them and the workflow that you connect them into, right? If you can't action an insight or an output generated by one of these..
10:15Optimization models
10:21What is the point, right? It's just more insight
10:25It's not actionable
10:30So the way they've literally been deploying these models is within Foundry using the Palantir ecosystem and specifically compute modules as well
10:35This is a new feature that allows you to deploy containerized Docker images on the platform that scale up and down depending on demand dynamically
10:40It's a bit of a technical topic, but honestly, for a developer deploying these models, it's super powerful
10:48It means a lot of cost savings and also the ability to scale as your user base grows
10:53So for Selco and for Timefold, this is a new way in which they're engaging their customers and users of their products
10:57Products and the fact that Palantir is allowing them to rethink what was actually possible in terms of a go-to-market strategy, that's just awesome
11:03So Dr
11:09Karp put out a video a couple of days ago about getting into the S&P 500, and in that video, he said the products that Palantir is building are becoming legitimate platforms for companies
11:15So, piggybacking off what you said, are you seeing companies use Foundry AI Platform to build the rest of their enterprise applications on? 100%
11:24I think one of the interesting things, and we'll get into it, of course, but with the keynotes you're going to see and the stories we're going to hear today at AIPcon is how many customers are thinking about how they can serve their customers better
11:35Better, how can they build products for their follow-on customers, etc.? But I think then the lens turns inward, and they actually realize that their own internal operations, how they coordinate their processes, their operations, can be fundamentally transformed as well
11:47And that is the idea behind Warp Speed, which we announced recently
11:52Fundamentally, a lot of the software that today's modern enterprises, especially in the manufacturing space, have been built on, is constraining
11:56It is not fit for purpose
12:01Ultimately, if you're building a company from scratch, feeling the need to constrain to a template of how your business should operate from the architecture level up — like you're… Immediately constrained, right? You're immediately limited by that ambition
12:11We've seen Tesla, we've seen the likes of SpaceX realize that to truly push the boundaries and achieve their vision, they have to build from scratch
12:18They have to build their own systems
12:24Warp Speed is us playing into that, doing what we do best, which is building from the ground up with a decision-oriented focus, creating systems that mold to the business and the company, rather than the other way around—constraining the company to the software
12:31And I think the final thing I'll say, which is something that's close to my heart in terms of doing our Impul work on the ground with customers, is really about playing into our..
12:35Forward-deploying mindset as well
12:39Absolutely
12:43This isn't a templated product that you just sort of deploy and it's self-serve
12:50That last 10%, it's boots on the ground
12:57It is Impul, it's deep work with our customers, and I mean, a lot of people shy away from that, but I think ultimately that's so important to build a solution that uniquely works for each of our customers rather than forcing them to conform to a template
13:04And that ethos is something that's very key to us and how we've been able to succeed
13:08So, to take that to our customers, starting with manufacturing and industrials, this is a massive step change
13:12And one of my keynote presenters mentioned, while we were preparing his keynote, that the..
13:16The enemy of progress is legacy, and I think that perfectly encapsulates what you're saying about the ethos of Project Warp Speed
13:21100%
13:25I think that really hits home as well because ultimately, this event as a whole, AIPcon, there is now legacy
13:29We have over a year's worth of history doing this, and we continue to re-script it
13:34We continue to re-architect it around what our customers uniquely need at that moment
13:39So the legacy is great to have, the wins are great to have, but we don't let that dictate what we do next, right? The same can be said for Warp Speed, and you mentioned the S&P 500
13:51To see our unique approach to building this business and enabling our customers' success, as you said, people said it wouldn't scale, people..
13:54He said it wouldn't work long term
13:58Naysayers, naysayers
14:02Exactly, and I think ultimately it just further—I didn't know my conviction could build any further, but it further builds my conviction that the playbooks out there don't work and they don't have to be followed
14:06And the ability to do something totally brand new from a software perspective is getting our customers really excited
14:14Yeah, what you guys are really saying is that Palantir has these custom solutions that are actually providing transformational value for these clients, and I'm going to be doing a bunch of interviews today behind the scenes that'll be posted online, talking with more Palantir customers
14:21I got to talk to a couple of them last night
14:27And really, what they're saying is that they can't imagine a world in which these platforms don't exist after they've used them for about six months
14:33And it's like, okay, so if you're not going to do anything else, how can you keep building on top of these platforms? It seems like they're going to keep doing that over the next couple of years, which means we've also got some keynotes to hear from today on how these clients have been using them for the past five years and for the next five years
14:41Nat, what can we expect? We've got a great keynote lineup
14:48I'm really excited about our speakers
14:52I want to highlight TrinityRail as kicking it off first
14:56I think some of the exciting use cases I have been seeing across the board are really promising
15:00For all of our keynote presenters, the focus is on this human impact, as I mentioned at the start
15:04Many of the use cases you'll be seeing today are not just applicable to one specific industry; they address core problems that I believe also relate to other sessions happening across the board today
15:17I'm going to briefly mention Meet the Masters, which is something that viewers on the live stream won't be able to see, but that's precisely why we've got Amit here
15:23That's also why I'm here
15:30It's really exciting because we're pairing experts from different industries around a common topic
15:36One that comes to mind that I’m really looking forward to is Tampa General Health and the Department
15:47Of State
15:58I was like, Department of State, there are so many customers at this point, and their entire conversation is going to be about modernizing health impacts, which is going to be… And again, you know, these are two organizations that you wouldn't necessarily see having the same conversation at a table, but really at the core of what they're doing is building an experience that's applicable across all industries
16:09I think this is a theme that I would maybe like to give a little tip to the audience to look out for in the next few keynotes
16:13You mentioned healthcare
16:16We've got Tampa General, Mount Sinai as well, presenting in the first half of our keynotes
16:20There's sort of a primitive forming, and this kind of idea, I like
16:22To think of it as like an engine primitive
16:25Like, what is the underlying engine behind a lot of these use cases? Right
16:27Like Tampa General, Mount Sinai, I don't want to give too much away, but there is a common foundation, a common engine that we'll talk more about during our little halftime show that is basically the foundation of both of these use cases
16:32Right? Absolutely
16:37Because both of them are healthcare companies, so at its core, there's a foundational element
16:42They can have different use cases, but there's something at the core
16:46Exactly
16:50And again, you've got that customization element, there's a configurability there, but I think that common engine primitive—that's something we're going to see grow, and again..
16:54It's not just healthcare, right? Like the eligibility engine that we'll be seeing a little bit of during those keynotes, which is sort of a core primitive, at the end of the day, isn't purely healthcare
17:01You could take that architecture, you could take that archetype, and apply it to many other similarly structured problems
17:05So, I think you're going to see more of that
17:08That may play into our product narrative very soon
17:12I've seen that kind of approach emerge from these stories, so I'm excited to see how it develops on the stage behind us
17:16And the other thing I’ll say about that is that many of our keynote speakers are also participating in different sessions throughout the day, so they'll be able to dig in and explore
17:26Expand upon the presentations that they'll be showing here
17:30So, when we're talking about that core piece of software that's really enabling these use cases, a lot of our attendees today are going to be able to dig in even more
17:33Exactly
17:37Actually, on that point, a new session for this AI Beacon, screen sharing
17:40I don't want to give too much away, but I guess on the live stream, you guys aren't going to get to see this
17:43You got to give us the juice
17:45Yeah, I know
17:46You can share a little bit
17:48There's too much to cover, honestly
17:51I'm just so excited about it
17:54But screen sharing is great because ultimately, it's getting every single person at this event like hands-on keyboard and actually speaking about the work they do day to day in
17:57The platform, right? So literally, we are having people share their screens and do this live in person
18:00We're not doing this on BTC this time around, but literally going around the table and sharing stories of like, why do I live in this screen? Why is it valuable to me? How did I build it as well? I think the unique thing there is, firstly, we deliberately mix up those tables like crazy
18:15It's such a random chaos of different organizations and different user personas as well
18:20So, I've seen conversations already in some of the prep sessions where a VP or a C-suite persona talks about their workflow, the work that their teams are doing, and then you'll have an engineer, like a pair of boots on the ground, who is like..
18:33Building out these workflows, poking at that flow, poking at that product, and asking questions about how it was built is sort of triggering questions in leadership's minds, like, oh, maybe there is a better way I can build this
18:45Maybe I can learn from that
18:49So I honestly learn an absolute ton from our customers every time we do this
18:53But to have our customers learning from each other—that's the payoff
18:58You are actually going to be doing more on the ground than we have ever had here at AIPcon
19:02I'm really excited for it because every AIPcon I think we level up
19:06So now we're doing a pregame show, we'll be doing a halftime show, I have customer interviews
19:10There's just a lot to dive into that people watching from..
19:16I don't get to see much of this, but I'll be posting a lot of it
19:21Since we have two Palantir Technologies employees here, I'm going to take this opportunity, in front of thousands of people—no pressure—to ask some juicy questions that I’m sure people really care about
19:25Dr
19:28Karp released a video congratulating employees and investors on getting into the S&P 500
19:32He's running through a forest, then stops running, looks at everyone, and says, "This was pretty, pretty cool." What was your initial reaction, Jack, when you saw that video? I mean, my initial reaction was to keep running, right? It kind of aligns with the theme here, and you might notice Dr
19:39Karp's shoes are slightly different from mine, certainly today
19:46But, like, I think we're still both striving, right? We're still heading towards those goals
19:52I think, honestly, no, it's brilliant
19:55Dr
19:58Karp is, I truly personally believe, nobody else could lead what we are doing
20:00Absolutely
20:02And the level of vision—this is something that we at Palantir have such high conviction in
20:04We mentioned the S&P to get that validation
20:06I'll be brutally honest, we didn't need that validation
20:08We had the high conviction that this is the way you architect a company
20:11This is the way you solve problems
20:14And to see Dr
20:16Karp speak to us in that way and ultimately continue the conversation
20:20He actually joined our welcome reception here at AI Expo for the first time ever last night
20:24So it's not a case You're going to see him up on stage, and that's it
20:27He's deeply involved in this event
20:30He wants to learn from those customers as well
20:33You've, I know, spoken to him a number of times
20:37So I think this is truly like everybody is on the same playing field here
20:40Everybody can learn from each other, and the fact that we as Palantir feel that, and our customers feel that as well, is almost like welcoming them into our world a little bit
20:44I completely agree with that
20:49I think that video perfectly encapsulates what it means to work for Palantir
20:53Folks on the live stream can't see this, but we have on the side of the wall here that says "Train, Run, Win." Maybe emphasis on "Run." I could probably work on that a little bit
20:58More
21:04But again, that attitude of, you know, we are not a traditional company, and that is what works for us
21:08And Dr
21:11Karp has been able to lead the charge, and it's, you know, we're behind him
21:14For me, I got really excited
21:17Oh, 100%
21:20Actually, on that point, Amit, what do you think the audience is going to be excited for this time around? We've got a whole set of keynotes
21:22We're going to be back again at halftime, right? The fact that they wanted more of us
21:26We can't get enough
21:30That is baffling to me, to be honest
21:33We can't stop coming back
21:37But yeah, what do you think the audience should look out for over the next, I guess, seven or eight keynotes that we have? So I got to talk to a lot of customers, I got to see how they're approaching
21:40Using AIP, I think that on this fifth edition of AIPcon, if you didn't know how customers are using AIP, you're going to learn
21:48And you're really going to understand why the entire world is starting to talk about Palantir, why everyone wants to interview Dr
21:52Karp, and why this company's brand positioning is becoming one of the most interesting companies on the planet
21:57As Karp mentioned a couple of months ago, this journey is just getting started
23:27We'll see you back at halftime, folks
24:58See you in a bit
26:29Don't go anywhere
27:03Our therapy department has to create a daily spreadsheet, which takes on average about 92 minutes
27:37Since the rollout of the Palantir tool, we have decreased that time to 12 minutes per discipline per day
28:11Our numbers, people, that's about 10 and a half-ish more patients per week
28:23With Palantir, we are able to seamlessly integrate data across manufacturing, sales analysis, claims, product support, and procurement
28:29As we continue to embrace this powerful platform, we are confident that it will play a pivotal role in shaping the future of SIMIAT and HDE
28:37Red Pro is committed to simplifying insurance
28:45Using AIP, we can do that better than ever with ClaimSymbol, our new automated claims QA assistant
28:53Regulatory vets were spending 8 to 10 hours prepping for a race card, combing through databases, trying to uncover information on every horse entered for that day
28:58HysaCheck is going to simplify this process
29:04The Palantir tools have been a godsend
29:10With Boundary, we gained visibility into our production and scheduling systems, which we never had before
29:15I can't say how much I love Palantir
29:19It's made my life so easy at work
29:22The most important part of AIP for me is the build pipeline that uses AIP
29:25To run different price increase scenarios, it would take us a couple of weeks, and now with Foundry, we can have different price scenarios in a couple of days
29:37The real game changer has been the alerting system that is improving my staff's behavior
29:49Thanks to the ontology, exporting data and building up a report will take no longer than a minute
30:01It's life-giving to have a company inject itself into the fold
30:10I'm not certainly on… The easiest, but the hardest problems we're facing right now are wildfire mitigation
30:21Well, thank you, thank you
30:31On all other days, I would talk about our AIP prowess, our Foundry, maybe even our government products
30:41And there are some things I'd like to say there, and welcome, obviously, all of you—111 of you—presenting actual things done in Palantir products
30:58We have the leader of our retail movement here, which is important
31:02But this..
31:06For whatever reason, this event was scheduled right before probably the largest event that has ever happened to our company
31:09And we've had some pretty interesting events
31:12First of all, we survived
31:15Most Silicon Valley companies do not survive
31:18We focused on how
31:21Institutions ought to run and build software products around that, starting with clandestine services, moving into special operations, moving into the battlefield, and moving into commercial applications
31:29These products have materially changed the course of society
31:37We DPOed, which is a fancy way of saying we didn't give all the money to the bankers
31:45Instead, we transferred it to loyal individual investors, which is part of the reason we have so many supporters
31:51Yet, no one thought we would be profitable on a free cash flow basis or that we would be profitable at all
32:01The idea that we could be in this exclusive club of the S&P 500 was something that, believe me, everyone internally was saying, "Okay, Alex, just shut up about this idea
32:06This is too much." No
32:08The way we're going to get this to work
32:11And it worked
32:13And for us, this is just like at Palantir
32:16This is arguably the biggest event that has happened to us because, you know, you're actually a very serious company, and it was done on the back of products that are powerful enough that we could keep our company the same size and increase the number of partners we have and the revenue
32:39So, I would say, while getting on the SPAC, we stayed true to our values
32:46That's kind of obvious and sometimes annoying to our partners because we have to speak about certain truths that maybe aren't popular, but it doesn't make them less true
32:56But we also have the very best partners in the world
33:01Palantir, over the course of 20 years, there have..
33:04There have been two things that have been true
33:06We have the very best people
33:09The Palantir degree now is the most important degree in tech, bar none
33:12People now—if we make sure we get people who are mission-aligned—because we have so many people just wanting to join so they can become famous
33:23And then we've had the best, the very best partners
33:31I hate the term clients, but de facto clients
33:40And obviously, we could go through the government clients we're allowed to acknowledge, and they are the most elite and interesting in the world, but I would say our commercial partners— as witnessed here, and the people presenting—are the most interesting, most dynamic, and in general, the highest integrity commercial partners that you will ever meet
33:48The people running them, who purchased our products, and the people supporting the people running in and doing their own things in IT, engineering, production, and marketing have one thing in common: which is, I need my business to work better, and I'm going to look at the product and not at the steak dinner
34:11And that's how we ended up with the best people, the best customers, and the best use cases
34:16And of course, that is just the massive structural advantage
34:21And that was an advantage before you got to AI
34:26 What is the large language model, by the way? As a kind of non-sequitur, I'm not sure what the person, our partner from the US government, is going to say today because I didn't even know we were allowed to..
34:40Talk about this stuff, but I'm really excited that we can have a glimpse into that
34:43I will see what we're allowed to know
34:47But that's super cool
34:50Then, if you have this revolution, large language models and algorithmic systems in the US government, it's more, in general, more about algorithms than just large language models
34:55It is a weird revolution because, in a normal revolution, you have lots and lots of startups, and one or two of them get all the value
35:01In this revolution, you have..
35:07the large language models and AI need to be processed within enterprises
35:13And what that basically means is that enterprises with high business acumen and high talent can raise their game and become essentially the startups that would have disrupted them
35:19Disrupt the startups
35:25So it's like an incredible time to be you, and arguably not as good a time to be a venture investor
35:30It's like, it's great to be us, it's great to be you because the intermediary of large language models is how do you process them? How do you fix your margins? How do you turn your company into a portfolio? How do you understand where the risk allocation is, the capital allocation? How do you turn your non-engineers into engineers? How do you disrupt the distribution? How do you make up for distribution while distribution has capital costs? The nature of political uncertainty disrupts your products
35:59All these things are solvable with our product, your business acumen, your data, your way of going
36:10To business, and you can accelerate them
36:21And it's that acceleration that is going to drive our partners to what I view as what's going to happen to us, which is to become stronger, bigger, more successful, and actually happier, because you do not want your best workers working on things that they don't enjoy, that there's no point in doing, where they're not productive, where they can't prove the results, where the margin changes aren't actually showing up on the bottom line
36:32And all those things are possible with our product and with you, and it's just an incredibly exciting time
36:37But again, the most exciting thing for me, and I'm very happy to have you guys here, this is probably going to be an AI Pecan that's part work, and Quite frankly, you may see them partying
36:44Surprise, they're even awake
36:50You know, it’s like, you know, college kids don't really get the opportunity to party
36:54High school didn't work
36:58No one would party with us
37:03College..
37:07college, you were busy trying to get into Palantir
37:12And then you have to work and work, and we had a DPO, but no one really thought that was going to work, so it was a surprise that there was no partying
37:15So now, if they're slightly slower, slightly inebriated, forgive them
37:18Enjoy the vibe
37:22It's a rare time for us, and we're going to be back to crushing it with less alcohol and other things
37:26Yeah, we don't—we don't, it's like sex, drugs, and rock and roll, but we don't really have any of them
37:30But we do have a..
37:33Great product
37:36It's a great product
37:38You don't need any of that extra stuff
37:40The product makes up for it
37:45It's the size of the product that really matters
37:50So, with that, welcome to what I think will be a really exciting time
37:57I am exceedingly happy to have you here
38:04I'm generally very happy yesterday, especially regarding the SAP, SNP admission, and other things, and having you and 111 people actually demonstrating working solutions
38:11I hope you have a great time and come back again
38:19Take care
38:26Please welcome from Trinity Industries, Chief Supply Chain Officer Michael Shin, and Vice President of Data and Analytics Steven Ecker
38:31Well, good morning, everyone
38:35Good morning, Palantir, distinguished guests
38:40Once again, my name is Mike Shin and I'm Steven Ecker's co-pilot
38:51We're absolutely honored to follow Dr
38:54Karp
38:56Pretty tough act to follow, but we'll do our best
38:59We're here to share with you a special journey
39:02It's a really cool story about an ontology and how this one DNA strand evolved into three generations of use cases that made a financial difference for Trinity Rail Industries
39:14Trinity Rail has been trading on the New York Stock Exchange for the past 50 years
39:25We design, develop, manufacture rail cars, lease rail cars, provide aftermarket repair, maintenance parts and components, and lease rail cars
39:31So we're in the train business
39:35And I forgot the clicker
39:39Bear with me
39:42So with that, the setup is this
39:45Over the past two years, we have..
39:48Doubled revenue, from $1.5 billion to $3 billion last year
39:54And during this period of rapid growth, we faced the most turbulent, most daunting, and most challenging supply chain constraints our generation has ever seen
40:06And with that, we faced a crisis
40:11How do you deliver cars with all those supply constraints? At our lowest point, we hit 67% on-time delivery
40:23So we were in the pressure cooker
40:29It was my fault, our fault, we're the fall guys, we're in the hot seat
40:34We overcame that, obviously, because we delivered on the revenue
40:40The way you do that is when you're at 70% on time, because we're a steel business, we buy steel, we convert steel, we ship steel in rail cars
40:45Rather than ordering a ton of steel, maybe you should order 1.3 times that amount
40:51Tons
40:58Now, at the time, we faced a $2.5 billion backlog, so we had confidence that we'd need that steel, but not everything matched up
41:04What happened was we had inventory challenges, and our excess inventory swelled
41:09We had a big problem
41:15When you have a big issue, why not partner with the world's best problem solver? In came Palantir
41:21And with Steven and my team, we worked in collaboration, developing a steel substitution algorithm so that we could utilize that excess inventory and buy less
41:33The beauty of that approach, especially in a hyperinflationary environment where steel prices can go up 2 to 3 times within months, is that we're utilizing that excess steel, that inventory on hand, at the lower price
41:44And avoid buying steel at the higher price
41:50That spread contributed to Trinity delivering stellar operating margins in Q1, Q2, and it shows in our stock price
41:56Right out of the gate, the team knocked it out of the park
42:02But we're not done; we've got plenty of problems ahead of us
42:08Now we're on the Gen 2, and my partner, Steven Eckert, will share with us how we collectively, collaboratively solved the design challenge
42:19 Yeah, so recognizing the success of the supply chain steel project, we started to ramp up our internal development of Foundry
42:25Now, rail cars have a surprising amount of customization, and we pride ourselves on being a built-to-order business
42:30But when you have all this customization, it presents..
42:36Complexity in our production processes
42:42Now, this is a challenge because multiple teams are involved and many factors need to be considered, like regulatory, commercial, and material availability when constructing a bill of materials
42:48Over time, this leads to data silos
42:54Now, operations was pushing for transformation, and we recognized Foundry as the tool to help accelerate it
42:59So, we were able to expand that initial proof of concept of the steel supply chain ontology into what we eventually realized was an AI-powered product lifecycle management application
43:11This application was custom-tailored for railcar manufacturing and helps us unify our materials
43:17So, I'm going to walk you through a little bit of..
43:25The application and the evolution of the functionality
43:33So, we start with the customer inquiry from our CRM, and then we created a model that scores the customer requirements against prior BOMs to give us the best starting BOM as a template to work off of
43:41Now that I've selected the most suitable starting point, I can make changes and traverse the BOM, navigating, updating, and editing
43:52Here, I removed a rod right out of the bill of materials
43:56Seems simple
44:00Next step, though, we leveraged AIP to enhance our parts substitution algorithm using GenAI
44:04This helps us identify other parts substitutions that we may not have considered
44:10From our historical BOMs here, we found a locknut that's 20% cheaper
44:14I initiate the..
44:18Substitution
44:22I can then customize the inputs, and in this case, I'm choosing a machine learning model that helps predict the price for stale and outdated concepts
44:26I apply the substitution, ultimately editing the BOM, and this may seem trivial, but there are a lot of locknuts on a roll cart
44:31And then we took it a little further, where we could substitute full packages of materials
44:42For instance, I select an air brake package made up of multiple pieces of components
44:47Again, I use GenAI to identify suitable replacements, and I found a package that's slightly less expensive
44:52And as I make these substitutions, I get a running total of the changes and the dollar impact
44:58And then I have a fully functioning workflow where I..
45:03We could push to SMEs like engineering for review and approval before I push it to production
45:08Now, this might seem like a simple app, but when you open up the application workflow builder, it reveals the underlying complexity
45:13The historical BOM is made up of data from multiple sources and has complex relationships
45:19We took all these systems and unified them into a single ontology
45:24In the end, there were about 200 tables feeding a single BOM
45:29Finally, we had to encode all the analytics and models to help drive decision-making for the end users
45:35When building with Palantir Foundry and Palantir AIP, all of this is in one place, helping us to iterate faster
45:40It's our AI in the backend
45:45So, you see the AI working? It's fully integrated
45:51Without this, we would generally require complex integrations and orchestrations with the operational application
45:57With AIP, it comes out of the box
46:03Another thing our developers really loved was how easy it is to navigate and edit the functionality of the application
46:09You can quickly spin up objects, and they're easily defined
46:14Essentially, you're navigating or traversing the ontology and enhancing and updating your application
46:20And last, I go back to the underlying data lineage
46:26For us, we use this a lot as that control tower, helping us spot issues in our transform pipelines, assess data quality, and access
46:32And as our BOM (Bill of Materials) structure is complex, this was used a lot
46:37So to reiterate, for us, this fully..
46:43Integrated system helps us enable to build better, faster
46:49With Foundry and AIP, we can do everything without a significant need for front-end development
46:54And then, the ontology that we establish works as a foundation for other applications in our steel supply chain, supply chain, and manufacturing functions
46:59Now, for Trinity, this would have historically been a large project, a big endeavor
47:05We are able to do it with a somewhat of a learning curve, two and a half developers, about nine months
47:10With that, I'll hand it back over to Mike
47:14All right, thank you, Steven
47:17Let me tell you where we take it from here
47:21So, where are we going? The third generation of this: one ontology to solve inventory, solve design
47:27We need to solve supply chain
47:33The problem we had was that the suppliers were not confirming that they're going to ship on time
47:39At the height of our supply chain woes, standing in a steam, frantically running around chasing down our suppliers
47:45Are you going to ship on time? Are you going to ship on time? We found out they shipped too late
47:50The buyers are not doing their job
47:56Their job is to order the material at the most optimal cost, get that inventory down, as opposed to spinning our wheels chasing down suppliers
48:02We've got another problem
48:07So in comes Palantir, working with our team again with rapid execution to develop an ASN (Advance Shipping Notice)
48:11It's the best I've ever seen
48:15I've been doing this for 30 years, and the speed of their execution is phenomenal
48:19On the inventory side, it takes three months to get to a functional workflow with a $30 million impact on the bottom line with the ASN
48:30How it works is you find, you will send an AIP agent—I call it a chat agent—via an email
48:36The supplier will respond to that inquiry
48:42The AI will filter, review the email, and then confirm if it’s on track
48:48If not, it will elevate it to management, and with that visibility and transparency, we get attention
48:54They give us what we need on time
49:00Sometimes suppliers don't want to tell us they're going to be late because, as Jack Nicholson said a long time ago, "You can't handle the truth." Yes, we can if you tell us, so we can adjust our production and cope with it
49:10But ASN was, again, another solution to the problem
49:16And with that, I'd like to… Close on this note
49:23Once again, thank you very much for your engagement
49:30A year ago, I had this aha moment when I met with Erez and Lila, and I thought, here's a company that's the world's best at finding bad guys all over the world in remote places we've never heard of, dangerous places
49:37Well, in our P&L, in our balance sheet, and in our operations— all of you business people—there are bad guys in there
49:48Foundry, go help seek, find the bad guys, and neutralize the bad guys
49:52So this is just the beginning
49:56I always say from this one ontology, from the root to the fruit, we have a long way to go
50:00And I'm here not to tout ourselves, but to motivate everyone here in concert
50:06Let's embrace this technology, embrace this company, and give us a..
50:12Tremendous momentum, where it becomes a ubiquitous, obvious solution
50:18With that, more speed, scale, and affordability
50:22Thank you very much for your time
50:25I look forward to socializing with you afterward
50:29Please welcome Christopher Simpson, Director of Data Strategy and Analytics at Lear Corporation
50:36Good morning
50:43 First, I want to again thank Palantir and all the folks here for the opportunity this morning
50:51My name is Chris Simpson from Lear Corporation
50:56We're one of the world's leading Tier 1 automotive suppliers, and today I want to spend a few minutes to walk you through how we're using cutting-edge digital tools to tackle old problems in new ways
51:07To give a little more background, Lear Lear was founded over a hundred years ago
51:12We primarily produce automotive seating, electronics, and electronic components
51:18Featured on nearly 500 vehicle nameplates, chances are everyone in this audience has ridden in a vehicle with one of our components in it
51:23To meet this demand, we have a global team of over 186,000 people working across 265 facilities in 38 countries, all striving to help achieve Lear's vision of making every drive better
51:39But at Lear, our mission is not just about delivering quality components, but also about driving innovation and efficiency
51:45We do this through our 'Idea by Lear' strategy, which revolves around four key pillars: innovative, digital, engineered
51:50and automated
51:56My team primarily focuses on the digital pillar, and that's where our partnership with Palantir comes in
52:01One of the key drivers of our digital transformation is the use of the Foundry platform
52:07At Lear, Foundry is deployed as an enterprise-wide, end-to-end stack, rather than as a tool for a specific use case
52:12The main reason for this is scalability
52:18With an organization the size of Lear, a central team cannot deliver all the necessary use cases for us to be able to serve the business
52:27Because of that, we democratize the platform throughout the organization and empower the users to go through and solve problems on their own, driving digital transformation
52:35Transformation from the ground up
52:42Today, rather than go through one of the more advanced use cases that we have, I'd rather take the time and actually walk through one of those grassroots examples and show you just how easy it is to build a solution on this platform
52:50Like many in the manufacturing industry, you know, critical shop floor information is often scribbled on whiteboards, buried in paper notes, lost in Excel sheets or even text messages
53:04At the end of each shift, often these valuable insights are either erased or filed away, never to be seen again
53:09One of our plant teams asked if there was a better way for us to capture and utilize this information using Foundry and AIP
53:16Let's walk
53:22Through that example, now where a user without any coding background is able to go through and deploy and configure an LLM-powered mobile app to capture and address shop floor issues
53:29It all starts with the Build with AIP module, which lists practical applications of LLMs
53:35This is incredibly helpful for our users because, while many are familiar with consumer applications like ChatGPT, very few are actually familiar with how these models can be applied to real business problems and solutions
53:52So, our users find a capability that fits their needs, and with a couple of clicks, they are able to go through and quickly install this function
53:57Many of these work out of the box without requiring customization
54:03However, in the event That the user wants to learn more, the platform does enable them using AIP Assist
54:08This allows them to explain the code to the user, as well as enabling the user to ask further questions
54:18This is one of my personal favorite applications inside the platform because AIP Assist exists in many of the different Foundry services, and it really helps us accelerate the learning curve for users as well as myself
54:32Now that the function is installed, the user can go through and follow a few steps to configure the ontology and is now ready to deploy it
54:45The user starts with one of those out-of-the-box workshop templates, in this case an… The inbox template pulls in the newly created ontology object that we just created a few steps ago and is now able to go through and start designing within the designer
54:59Recognizing that most folks are not walking around the shop floor with laptops, the user selects the toggle on the bottom left to convert this into a mobile app, and then continues to go through and configure the application
55:05As you can see, the app design interface is really intuitive
55:11It's largely drag-and-drop components, making it very accessible to everyone in the organization
55:17This is one of the most popular features in our organization and on the platform because of this ability to allow people to create fully functional mobile
55:27Apps using ontology functions like the one we just created without going through and writing a single line of code
55:36This is a game changer for us, not just because it enables our engineers, but everyone in the business, to build solutions in minutes and hours rather than weeks and months
55:46Now, for this user, all they have to do is take a picture of an hour-by-hour board on the floor, and the large language model will automatically identify and log that issue
55:56Since the app is part of our broader Lear Foundry platform, they can also assign that issue to other users for action
56:05Those users then receive an email notification alerting them of their assignment, and they..
56:10Able to go through and log the resolution all inside of the same app
56:14But remember that this is not the only app or use case in this plant
56:18The app can now share data with other manufacturing and business systems that we have connected to the platform
56:23This integration means faster issue resolution, better data-driven decisions, and more agile responses to issues on the shop floor
56:34And these benefits are only going to accelerate as the plant teams continue to develop additional ideas around areas like supply chain, quality, or capacity planning
56:39Even a very simple app, when connected to the broader Lear ontology, drives significant value
56:45But this is not just about one plant either
56:48As I mentioned
56:52Earlier, with 265 sites globally, the opportunity for us to scale is immense
56:56The Foundry platform enables us to have thousands of connected use cases and truly allows every user to become a builder, transforming the way that we operate
57:07To illustrate this feature a little further, let's revisit our example
57:13Once an issue is logged, the LLM doesn't just stop at documentation
57:19It can now analyze the problem, recommend a solution, and in the future, autonomously execute the necessary steps to resolve the issue
57:31Imagine the power of having a system that not only captures insights but actively participates in the problem-solving process, helping us to identify, recommend, and resolve issues more efficiently
57:37Even resolve issues faster than ever
57:42With Foundry and AIP, that future starts now
57:48Thank you, everyone, for your time and attention, and I look forward to continuing this conversation today
57:54Thank you
57:59Please welcome the Director of the National Geospatial-Intelligence Agency, Vice Admiral Whitworth
58:03Thanks, team
58:07It’s great to be here
58:11Such an honor to be invited to this forum, especially because I’m representing a group of people I’m really, really proud of — 16,000 strong, of the National Geospatial-Intelligence Agency, which we’ll call, henceforth, NGA
58:28 What do we do? Well, for the next eight minutes, we’re going to go over three tasks, basically
58:34I’m going to introduce you to NGA
58:39The second task is..
58:44I'm going to introduce you to something that helps us with the scale and speed of our work, and that's NGA MAVEN, and by extension, something called MAVEN Smart System
58:50And then third, we're going to talk about the future and the trajectory of AI and ML at NGA
58:56So here we go
58:59Task one, NGA
59:03NGA is in charge and is the center of excellence for geospatial intelligence
59:07And what that means is we task, process, exploit, and disseminate understanding through imagery
59:13Think satellites
59:18We task satellites, and then we take all of that data that comes in and build understanding with accuracy
59:24It takes people, but it also takes the advantages of IT, and that's why it's so exciting to be here with you
59:35Before I go into Maven, let me just describe the scale of the situation we're facing here
59:41We move more ones and zeros; our throughput as an agency is higher than any other agency
59:47Some people might be surprised to hear that, but then again, you shouldn't be surprised, because if you've ever taken a high-resolution photo with your phone and then tried to send it without realizing you did that, and it just crashes, try doing that around the world
59:56Try doing that with a constellation of satellites that is ever-growing
01:00:05It's a good problem to have, so we need an advantage
01:00:14And we turn to the advantage of artificial intelligence and machine learning, combined with the best of breed photo interpreters and imagery
01:00:20Interpreters at our station
01:00:24InterMAVEN, let's talk about MAVEN
01:00:28This is task two
01:00:32NGA MAVEN, and by extension MAVEN Smart System, is now a program of record
01:00:38It began as a project, initially meant to build excitement within this sector, but also within the Department of Defense, to see the importance of using AI and ML in imagery interpretation
01:00:49It is also the first software to be validated by what we call the Joint Requirements Oversight Council, JROC
01:00:57So this is the real deal
01:01:04And sometimes there might be a tension—a healthy tension—between courage, where you want to fail fast, and stewardship, because we're dealing with the taxpayer's money
01:01:12We don't want to pay twice
01:01:16We want to make sure that we're taking the right risks
01:01:19We're doing it responsibly
01:01:23You're seeing now a slide loop of some screen captures inherent to NGA Maven and Maven Smart System, the graphical user interface that we use
01:01:29And it represents one of the things that we do at NGA, which is targeting
01:01:34We also do warning, safety, and navigation; I'll address those in a few minutes
01:01:46But right now, we're going to talk about targeting and the targeting cycle, and the efficiencies, scale, and speed that are brought by NGA Maven and Maven Smart System
01:01:51Targeting
01:01:55You see words like OODA loop; it's effectively a business process
01:02:00It begins with objectives and guidance, then moves into target development
01:02:04And let's just stop there with target development
01:02:10It involves how to discover the right targets and ensure that you're accurate in your characterization of those targets
01:02:16And especially when you're talking about what we call positive identification
01:02:21This is a presidential prerogative; it is the prerogative of the Secretary of Defense, it is the prerogative of a supported commander, a combatant commander, and as a combat support agency, this is immutable for us
01:02:32It is so important that we get it right
01:02:36So, in target discovery, when you have all that data coming in and you're going to have the same number of eyes, ultimately, looking at that to ensure that we derive positive identification, it's time to use AI and ML in this graphical user interface known as Maven
01:02:52Smart system, and that's the power of it
01:02:58So, as the data comes in, we have models that have been developed with the assistance of our best-of-breed experts, who run inference on all that data and then discover proper targets
01:03:09They don't render decisions; they actually render a confidence level, a certainty level, that helps that supported commander make that decision
01:03:15It's really important to understand that this is a tool for decision-making, not a decision-maker
01:03:26So, what's the scale we're talking about here? Why is this so important? A lot of you may have gone through Global Entry, you've gone through TSA, which takes a quick picture of your face, and you're really impressed by the identity that it is able to verify
01:03:31Establish
01:03:35And you proceed
01:03:40I wish it were that easy in our business
01:03:45That's about 80% of the field of view when you're talking about your face in that particular sensor
01:03:51Try this from space
01:03:57Try this now, where you're looking for an object or some sort of behavior that you need to distinguish, which is two one-hundred-thousandths of a percent of the field of view
01:04:02That's the challenge we have here
01:04:08And thank goodness we do have our partnership with the commercial sector made manifest in something like Maven Smart System
01:04:14I don't go into exactly where and how it's being employed in the real world, but I will tell you, you'll find it in operational centers like United States Central Command, United States European Command
01:04:24Command, United States Indo-Pacific Command, United States Space Command, and hopefully many others
01:04:33And that cycle that we talked about, the targeting cycle, also includes things like force application, force execution, and ultimately combat assessment, and it starts over again
01:04:43And the code is so agile that we're actually able to add those aspects of the targeting cycle, not just target discovery
01:04:51And that's very powerful
01:04:59In the last minute that I have, I want to go over just a few thoughts about the future
01:05:06So, what's important about targeting could also be important when we're talking about our other missions, such as warning and the safety of navigation
01:05:12We intend to take this momentum that we've built with MAVE and apply it
01:05:18to a larger construct that is the entirety of warning
01:05:24This is where you establish a baseline for behaviors and equipment across the world, especially in hotspots where there might be irresponsible activities that need to be identified
01:05:36Look for those anomalies, use the same pairing of machine learning as well as AI to ensure that we are triggered as humans to take a second look
01:05:48Secondly, the responsible application of this particular code is something that is very, very important to me and to our agency
01:05:54So, we've already adopted something called GREAT, Geo-Ent Responsible AI Training
01:06:00Anyone who handles the code is going to be certified to understand the principles that we hold dear, including the laws of armed conflict
01:06:11Conflict, things like the principle of distinction, ensuring that we have absolutely derived positive identification, separating combatants from non-combatants, good behaviors from bad behaviors
01:06:22And then we also have now ventured into an accreditation of models
01:06:28So when you're dealing with something as important as positive identification, you want to ensure that the model itself is grounded in best practices and good training data
01:06:39This, for us, is something that we have announced as a pilot called the Accreditation of GeoLint AI Models, A-Game, and it will have gradations—A-Game 1, 2, 3, 4—based on how intense our examination and certification is of the training data, the personnel, and the..
01:06:48Practices
01:06:57So, this is very exciting, but we can't do it alone
01:07:06That's why I'm here, frankly, to address some of the best and brightest of the commercial sector
01:07:15And I appreciate the fact that you're taking the time to listen as well
01:07:20This partnership is tantamount to ensuring that we keep America safe and ready
01:07:27Thanks for listening
01:07:34Please welcome Chief Strategy Officer at Mount Sinai, Jody Cohen
01:07:38Good morning, everyone
01:07:43I'm Jody Cohen from Mount Sinai Health System in New York
01:07:47We're an eight-hospital health system across three boroughs of New York City, with outpatient locations numbering over 200 across the five boroughs of New York City, New Jersey, Westchester, and Long Island
01:07:57Our biggest challenge over the..
01:08:07The course of handling and managing these billions of patient records and patients on a yearly basis is space
01:08:17We needed a way to optimize our space in order to get patients through the system in the most high-quality and patient-friendly way
01:08:28In 2017, we embarked on the creation of a clinical command center, which helped us to centralize all of our beds in one place and be able to oversee where we might have capacity for our patients at one hospital versus another
01:08:45At the same time, we embarked on a hospital-at-home initiative, which is one of the newer "beds without walls" components of healthcare that you might have read about
01:08:57This allows patients to recover at home as though they're in an acute care setting
01:09:01It reduces..
01:09:04Infection and allows patients to be with their family and to maintain comfort
01:09:08The difficulty with getting this program off the ground is evaluating the patient criteria to see which patients should be eligible
01:09:20Palantir came in to help us build a system that could evaluate patient records in real time, looking at blood results, initial diagnoses, radiology tests, et cetera, to determine which patients might be eligible for this kind of care
01:09:37It allowed clinicians in the emergency room to have decision-making capabilities enhanced by this data, forming lists and allowing patients to go to the right place at the right time
01:09:49Our second biggest problem on the financial front has to do with our clinical denials
01:09:55I don't know if I have any insurance companies in the audience, but don't take anything I say personally
01:10:01We have addressed our financial strategy around clinical denials with help from Palantir
01:10:07And when my boss first introduced the concept to me, I almost jumped up and down to have such support with this issue
01:10:18You can see this graphic shows the growing rate of our clinical denials over time
01:10:24This impacts cash flow and makes our accounts receivable grow quite significantly
01:10:30Using the clinical ontology that was built for the Hospital at Home program and the Clinical Command Center, we were able to leverage financial data and incorporate legal guidelines, our managed care contracts, and other clinical documentation components into the..
01:10:47Digital twin, allowing us to act a little bit faster in relation to our denials
01:10:58Our first step was to generate denial letters based on the documentation that the managed care company was questioning
01:11:04It was a manual process that required eight denial writers and millions of dollars in outsourced resources to evaluate the clinical denial, review the medical records, write a denial letter, and then mail the records to the payer
01:11:26In this circumstance, it would take 45 minutes to one hour per denial
01:11:31With the large language model, reading the chart and consolidating the information, it now takes about eight minutes for the letter to be created and reviewed
01:11:43Then all that needs to be done is to download the letter
01:11:47The payer
01:11:51We've seen 100% FTE efficiency over this
01:11:54We've been able to manage all the letters, whereas before some would get lost in the shuffle, creating more denials for timely filing
01:12:00And based on the overturn rate of those denials, because that's where really the rubber meets the road, we have seen a 3% increase in our overturn rate that we expect to result in over $13 million in revenue that we're entitled to
01:12:10So this is a demo of the process
01:12:20We have a patient, Diana Cronin, with COPD
01:12:29The managed care payer said she wasn't clinically eligible for an inpatient stay, and based on our doctor's decision making and the care she received, we had already deemed her eligible for that stay
01:12:37We read through the clinical
01:12:45Documentation guidelines that have been preloaded into Foundry
01:12:52We match those up with the documentation in the patient's chart and then generate a letter that includes both components in our response to the payer
01:13:04Very simply, the template is created, the letter is written, at some point, and then, based on all the clinical documentation and the legal, if any, requirements— for instance, if a patient stays longer than two days in the hospital— Medicare requires that they're covered for clinical necessity
01:13:13These data points are impossible to match up for a human
01:13:23It would take a day to look through all this information and make sure it's correct
01:13:33So, not only is the letter enhanced through this process, it has a lot
01:13:39More backup behind the scenes
01:13:45We print the letter and it goes
01:13:50We're trying to work together more extensively with Palantir
01:14:00I feel the uses are endless
01:14:10The best thing about it is that all the information is in one place, in a giant data lake, where I'm able to evaluate not only whether the criteria were met, but also whether I should get paid and if I will get paid
01:14:20I can probably predict if this payer will pay me based on this denial letter, this denial letter reviewer, which hospital they were in, and which payer they have
01:14:31I'm looking forward to being able to use this on the front end as well
01:14:36We've already started and expanded our scope so that, on the front end, the social workers and utilization reviewers in the..
01:14:45The emergency room can have the data consolidated for them using the same tool that was built for the hospital at home, to see if they can justify the admission in advance to the payer so that there won't be a denial at the end of the day
01:15:08I want to thank you for listening, and if you have any questions, I'll be around
01:15:14Please welcome from Palantir, Sasha Spivak
01:15:20Good start to the day so far
01:15:24Thank you
01:15:28We've got a lot more coming your way
01:15:31We're going to take about a 30-minute break, get some coffee, stand up a little bit
01:15:37This is the most you'll be sitting all day
01:15:41For those of you joining us on the live stream, we've got a special exclusive for you starting in about five minutes
01:15:46We've got five more customers coming up to share their..
01:15:49Stories with you, and then after the break, we'll continue
01:15:51So get some coffee on the livestream, stay with us, and we'll see you shortly
01:15:54 All right, we are back at the backstage pass at Palantir's fifth edition of AIPCon
01:17:38We have a lot to talk about
01:19:21We had some keynote presentations, and we're going to go over that
01:21:05In just a little bit, we're going to hear from some other customers on how they're deepening their relationship with AIP
01:21:09 Jack, what did you get from those keynotes? I mean, I barely had time to even sort of get my thoughts together on that
01:21:18That was such a rapid-fire set of outcomes, stories, and also different approaches to telling those stories, right? You had, say, Trinity, that sort of had that power of three, like
01:21:26$30 million in savings over three years from just three months of development time—that's crazy
01:21:28That's just a brilliant story in itself
01:21:31Obviously, the ontology story there is great
01:21:34But then you go from that end of the scale to, of course, Liz's story, like Chris telling that story deliberately from the ground up
01:21:42How do people actually get started building use cases? How do they connect to that broader ecosystem of ideas? I'm so excited to hear more about him and his journey as part of that process because they're building something huge at LIA
01:21:53And so much of what we've seen throughout AIPcon is really new for our audience
01:21:58Having Vice Admiral Whitworth present Project Maven, I think that was..
01:22:02Such a unique and special moment for everyone on the live stream as well as in person
01:22:06And it's also teeing up new experiences that we're having across the board, including the new format for the Backstage Pass
01:22:10Exactly
01:22:14Yeah, and what's really interesting about Maven as well is that there's a lot of confidential work that Palantir does with the U.S
01:22:18government
01:22:22So, to be able to get a little more insight— I know from the retail community, a lot of people care about the government business
01:22:24So it was interesting to see how that business is evolving
01:22:26At the end of the day, I'm sure Twitter was blowing up
01:22:29Twitter works well with that
01:22:31Yeah, I hope you've been checking, surely
01:22:33I think also, like… Playing into some of the things we talked about in our little pregame show, that sort of engine primitive, right? We saw that up on stage with Mount Sinai, the eligibility engine
01:22:40Just to go one level deeper on that, because I think it's so awesome, that sort of graph view that we saw, like we mentioned on stage, of course, how much work would go into doing that kind of evaluation manually
01:22:51But in reality, that's thousands of data points—unstructured data, patient nodes, as well as structured data, vitals, and other parameters
01:22:56That just one chart lighting up and showing red for bad, green for good
01:23:02There's so much complexity behind it, and yet it's an engine that they can configure and deploy to
01:23:05Multiple use cases
01:23:08You would say it's going beyond chat
01:23:12Oh, it's never; we're way beyond chat at this point
01:23:17I think specifically there, you're turning the logic of the business, the logic of their nurses, their doctors, their boots on the ground into an engine that can run automatically across every single patient
01:23:22I just think I'm so excited to see how they scale that effort because it's such a unique approach
01:23:24Boots on the ground, we've got boots on stage
01:23:27There are boots everywhere
01:23:28Yeah, I'm still wearing the shoes
01:23:30We're still wearing the shoes
01:23:31That's the main thing
01:23:33Speaking of boots on the ground, we've got some boots joining us this time around, haven't we? Yeah, we've got some new people joining us as well
01:23:36Well, at this halftime show, and this is an exclusive for people watching on the live stream
01:23:43So, we're going to get started with some customers and gain a deeper understanding of how they're using Palantir
01:23:47First up, we have our first speaker from Wendy's Quality Supply Chain Co-op, Krista Pullman
01:23:51QSCC provides supply chain functions for over 6,400 restaurants throughout the United States and Canada
01:23:55Just two weeks ago, they announced a furthering of their partnership with Palantir
01:23:58Let's take it over to Nat and Kris
01:24:01Thanks, Amit
01:24:03Krista, welcome
01:24:06We're so excited to have you here
01:24:09I have a vision of you being up on that keynote stage in the near future
01:24:11Obviously, the partnership was just announced, but we'd love..
01:24:17To learn about how being a data-driven organization has transformed how you operate your supply chain
01:24:20Absolutely, well, one, thanks for having me
01:24:23I'm super excited to be here; this has been great
01:24:28The keynotes have been great so far
01:24:32I'm learning a lot about how other people are using the Foundry system, and you know, as far as a data-driven organization, in supply chain, data is paramount
01:24:37We've always been a data-driven organization, but that data has never been really connected in the most meaningful way to drive the right outcomes
01:24:46And as we started our journey, Palantir was really the one company, of several that we talked to, that was able to get us over the hump of worrying about our data not being perfect, and worrying..
01:24:55About, we didn't have all the data that we thought we needed, and it really convinced us that you're running this operation, you're running it pretty well
01:25:04But you have ideas that you can't unlock because your data's not connected in the fashion that drives the right action at the right time
01:25:14And so, in very short order, in a pilot we did last fall, we were able to see how fast the builds were and how much we actually knew about what we wanted because we were also considering maybe we don't know what we need exactly
01:25:28But as we got into it, we were really excited about how we were able to encode the tribal knowledge of folks like myself and a few others who have been around the organization for 10, 20 years to help make our newer teams more effective
01:25:39Employees become more efficient and are able to take action sooner in their training and lifecycle
01:25:49The example I’d like to share is of an intern who joined us for only three months over the summer, with no background in supply chain, yet was able to access our platform and make decisions
01:26:00It’s because we’ve guided the user to the core problem
01:26:06This is similar to what some of our keynote speakers discussed
01:26:11There’s still a person who needs to make a decision, but we’re eliminating the noise—like putting down the shovel for digging for a problem
01:26:14The problem is here; what do you want to do about it? That’s really where we want to focus our training efforts for our associates
01:26:22Absolutely, and as a newer customer as well, you mentioned getting started with the pilot
01:26:26How do you even begin to think about where to start with a partnership with Palantir? Where do you even begin? Well, it was really about how do we become more efficient in what we do
01:26:37We get stuff from point A to point B, right? And we try to get it there, you know, don't run out, don't have too much, and get it there at the most competitive price, right? That's the simplistic way to talk about supply chain
01:26:45But what we've learned over time is that it wasn't where we started, but the team kept telling us, "You're going to find things that you didn't expect," and that's so true
01:26:51And we've found that we're able to drive value to our restaurant operations
01:26:56Operators in a way that we didn't expect
01:27:01And so we believe that things like automating restaurant ordering, automating inventory, variance benchmarking—like how much waste do I have versus what I should—is really going to drive profitability and ultimately help us grow this wonderful brand that Dave Thomas started in 1969
01:27:07That's our end goal
01:27:13Krista, thank you so much
01:27:17I can't wait to see what we do next together
01:27:21And thank you for joining us today
01:27:22Thank you
01:27:24All right
01:27:25Thanks, Nat
01:27:27Thanks, Krista, for making sure these chicken nuggets can make their way into my stomach
01:27:29Next up, we're joined by Chris Simpson, a director at Lear
01:27:34Now, Lear's AIP journey originally began at a boot camp in..
01:27:37Detroit
01:27:41Our very own Jack Dobson was at that boot camp, and now Foundry is rolling out to thousands of users in the company
01:27:44Let's go to Chris and Jack
01:27:47Thank you so much for me
01:27:51I appreciate it, Chris
01:27:54Great to have you back here and — first time on the keynote stage, of course
01:27:57I think, I mean, as I mentioned at the very top, your story is fascinating in the difference in angle you took, right? Rather than showing kind of the end state, of which you guys have a ton of high-scale global workflows, you're talking about that builder story, those boots on the ground
01:28:09How do those ideas kind of start? How do you build on them? How do you end up turning them into a deployable product? I guess I'm just curious to go that one level deeper
01:28:16Deeper, like, what are some of the stories or maybe even some of the personas that you've seen pick this up unexpectedly? People on the factory floor, where else? Yeah, sure
01:28:23I mean, the example we gave was a real example, right? Notional data, all that stuff, but that is a real example where we had somebody on the shop floor who was like, "Surely, there has to be a better way." And a lot of the time, I also try to emphasize that it doesn't have to be a 500 ontology object-backed application
01:28:43It can be as simple as one object, one action, one workshop app to drive value
01:28:47And really, a lot of..
01:28:51Those use cases are the ones that end up being some of the most impactful, because they're the ones that are like, no, this is a problem I have to deal with every day, day in and day out, whether it's on the shop floor, in the back office, or anywhere in the business around the world
01:29:07We have these types of problems that now Foundry allows us to do something beyond Excel, right? And really enables us to have the horsepower to deliver applications that scale across the business without necessarily having all of the infrastructure and weight that sit behind them
01:29:14That's incredible
01:29:22And I mean, along with what you're saying, every idea has to start somewhere, right? So, the thing I love about Lear, and I sort of..
01:29:30We learned more about building the story with you, and the fact that you've sort of abstracted the process of actually identifying these ideas, kind of collating these use case requests from the business, and then you basically manage the approval, the deployment, and the value measurement, all within Palantir's ontology, right? You've actually built that out as a workflow in itself
01:29:39So, tell me more about that process
01:29:49How did that idea come about? Yeah, like you said, we kind of use the ontology to manage the ontology, so it's a little bit meta in that regard
01:29:58But one of the things that we have, again, with all of that scale, is how do we go through and also still balance the governance
01:30:06Aspect of it with being a publicly traded company and all that sort of thing
01:30:09How do we go through and balance that scale with the ability to still allow for people to build in the platform? So we ended up using a workshop app to go through and build a use case submission and approval system that allows us to capture those ideas, give a little more specificity in terms of, okay, what is the function? What's the area of the business? What's the value that we see? Are you going to try and hire somebody to help you with it? Are you going to deliver it yourself, like the one that we saw? All of that type of thing inside the platform itself
01:30:22And what that does is it then allows us to..
01:30:35Take that use case object and actually then tie it to what are the data sources that this is actually helping to support
01:30:48What are the workshop apps? What are all the different functions? And what that allows us to do is, when new users join the platform, they're able to go through and say, okay, what do I have for my plan? What are the existing use cases? What are the data sources? All that sort of thing
01:31:03Having all of this inside the ontology makes it a very smooth process for them to be able to pick it up and run with it
01:31:07That's incredible
01:31:12A great example of turning the ontology inward to look at your own operations, right? Well, Chris, thank you so much for joining us
01:31:14We're excited to hear more later on today
01:31:16You bet
01:31:18Thank you very much
01:31:20Thank you
01:31:22Alright, thank you, Chris and Jack
01:31:24Next up is LifePoint
01:31:27I'm joined here with Christy Leitzel, the Director of Therapy Services and Wound Care at Trios Health LifePoint
01:31:29Christy, you're very unique in that before you went into healthcare leadership, you spent over a decade as a clinical speech-language pathologist
01:31:34Yeah
01:31:38Can you give us a little bit more insight on how AIP is making LifePoint a more data-driven company? Sure, yeah
01:31:42So at LifePoint, we have several values
01:31:45Two of our values are really championing patient care and making a difference together
01:31:48And so, what has come about out of that is we created, through Foundry, a program called Valinor, that really focuses on..
01:31:53Therapy services in the hospital, such as physical therapy, occupational therapy, and speech therapy
01:31:57What started out as a bunch of post-it notes on my desk has developed into a scheduling system, an AI note-taking system, and actually an AIP system that looks at patients as they come in and flags them for needing therapy
01:32:06That's what we've created
01:32:14When did you see those workflows actually making a difference inside the company? Yeah, so we beta-tested it, of course, to begin with
01:32:22But then, when we pushed it into the hospital, we knew within the first week because scheduling all of our patients across those three disciplines was taking anywhere from 92 minutes to three hours
01:32:31Depending on patient caseload, and so what happened is now we run that report and it takes two minutes, 37 seconds, depending on how many patients are on it
01:32:36Okay, so obviously that's a big transformation
01:32:40Right
01:32:45The goal of the healthcare company ultimately is to serve patients
01:32:49So, how important has this been for those patients at LifePoint? So, the rollout of Valinor has been epic in proportion in the sense that we are able to see more patients now, which is really going back to that championing patient care, right? But we're doing it with the same number of FTEs that we were doing it with before
01:33:09So, on average, at Trios and my hospital, since the rollout of Valinor and AIP, we're seeing on average 10 to 12 more patients a day
01:33:16Than we were before, just because of it
01:33:24Do you have any advice for people beginning their journey to become a data-driven healthcare company? Yeah, I think that you look at the problem, right? And what is the problem that we need to solve? And then you really talk to your frontline workers, and that's what made it unique for us, is I had that experience as a frontline worker
01:33:32Where do you see Foundry taking healthcare over the next decade? Oh man, you know, that's something that I sit up and think about at night, and I wonder, what could we do next? And I have some ideas
01:33:44I think one of them could be looking at harm prevention
01:33:49And what that is, is when a patient comes in, we look at their H&P, so their history, and we say, okay, are they at risk for..
01:33:53Wounds, pressure injuries? Are they at risk for aspiration pneumonia? And so, of course, I think of it from a therapeutic mindset
01:34:02From that perspective, could we take that data and then create a deliverable on what to order next? And we're already seeing that happen
01:34:11I would just like to push it further
01:34:19I think we can push the envelope, right? And so, with the predictive simulations you can do based on the aggregate data you're compiling, that could be something truly transformational over the next decade for different use cases
01:34:27Yeah, I think it could be absolutely transformational
01:34:32What we haven't discussed is that, as part of Valinor, we can actually write evaluation notes
01:34:34In it
01:34:37So, we created a checkbox system where therapists go in and they check the boxes, right, of what they're seeing in their evaluation
01:34:40But then what it allows us to do is it allows us to hit a button and the AI writes the note for us
01:34:48That's not taking the provider out of it; it's providing a basis to write their note, too
01:34:53So, they can go back in and add what needs to be added
01:34:55They can take away
01:34:58But that AI gets smarter the more we use it
01:35:01So, what we're seeing is an evaluation that was taking 45 minutes to evaluate and to treat; now we can do an evaluation in 30 minutes, and the evaluation note takes maybe a minute and a half
01:35:04Wow
01:35:07Thank you, Christy
01:35:10That was awesome
01:35:14And before you hop away, we here at Palantir are very proud
01:35:17Of the work that you've been doing with Palantir, and we feel like you're a part of the family
01:35:21All of our internal teams have flags
01:35:23Oh, my gosh
01:35:24Wow, look at this
01:35:26This is amazing
01:35:28Thank you so much
01:35:29Thank you, guys
01:35:30Have a nice day
01:35:31Wow, that's beautiful
01:35:32Thank you, Chris
01:35:33Thank you so much for joining us
01:35:35Thank you
01:35:38Thanks, Christy
01:35:40 All right, next up, we have Eaton, and we're joined by Pete Green, a Director of Enterprise Data Assets at the company
01:35:43Now, Eaton just signed an expanded partnership with Palantir after months, based on the diversity of use cases they saw and that they've been able to do with AIP
01:35:47They were also at the first-ever AIP Bootcamp in Chicago
01:35:51Let's take it over to Nat and Pete to learn more
01:35:55Thanks, Amit, but I don't want to say they were the first part of the AIP Bootcamp, but Pete was actually there, and so was Jack
01:36:03So, we've got a little bit of a reunion happening here at the halftime show
01:36:07Pete, if you could tell us a little bit more about how AI is being brought into production at Eaton and how you're using the Bootcamp model to do it
01:36:10That's a great question
01:36:14First off, start with the Bootcamp model
01:36:17The great thing about the Bootcamp is it really kickstarts new ideas
01:36:20And you can imagine when you introduce artificial intelligence or generative AI into your corporation, you're going to have hundreds of ideas and hundreds of projects that people would like to look at
01:36:28But what the..
01:36:33Bootcamp really allows us to take an idea and get everybody co-located for two or three days and have a hackathon
01:36:37At the end of three days, you pretty much know whether that product's going to work or not
01:36:46Once you decide if it's a good product and it's going to work, then you can expand it to move into production
01:36:50I think the expansion of AI has been really exciting
01:36:55We were partnering with Palantir before AI was integrated into the Foundry platform, and that's where we learned the most during that first bootcamp
01:37:04What I really see as the power of Palantir, and this is what I always tell other customers, is that there are two parts to being good at building these future-looking projects, and one of them is..
01:37:13Getting the data ingested and connecting the data, so that was great
01:37:18Now, introduce generative AI, and now, all inside of the Palantir platform, you have the connected data, which we refer to as the ontology
01:37:28Now, you can layer in multi-tiered large language models that can help you summarize your data and really produce some intuition that human beings just could not reach
01:37:34So, it's really exciting for us
01:37:40And how has your perspective changed, thinking back to where you were with that first AIP bootcamp to where you are today? Yeah, well, I think we learned a lot
01:37:47I think what companies will see is the more you build out your ontology, probably the better off you're going to be
01:37:56You also run into some… Challenges with, again, once you start introducing artificial intelligence, the projects get more exciting, and people become more excited, and the benefits get bigger
01:38:05Then those hundreds of great ideas turn into multi-hundreds of great ideas, and it's all just really a good story
01:38:10I think we have a lot of confidence
01:38:15I think we realize that we're going to need to scale with these projects
01:38:19So, not only our Eaton team, but our partnership with Palantir, it's just really a win-win for everybody
01:38:24I was going to say, I think there's a story there that I find fascinating, around how, as you build your intuition, as you mentioned, around the platform and the capabilities of these underlying models, you sort of..
01:38:33Starting to redefine your ambition almost
01:38:37Like yours personally and also your team's, they start to rethink, oh, things that I thought weren't possible or would be too complex, maybe they are actually possible
01:38:46So there are some examples where things that were never even on Eaton's radar are now ideas that you're perhaps exploring next
01:38:50Yeah, I think one of the strengths of AI and Palantir is that, before, it was really difficult to get these tools into the hands of users
01:38:58Now you can build all of these tools inside of Palantir, then distribute them using the user interface, and now everyone has access to them
01:39:08You can really make a big impact, whereas prior to that, it would be tough to roll out these solutions
01:39:09These larger projects out
01:39:11- Yeah, sort of democratizing that capability
01:39:13- Yeah
01:39:15- Amazing
01:39:17- Pete, thank you so much
01:39:19- It was a pleasure being here
01:39:21This conference is fantastic
01:39:24- Thank you for joining us
01:39:26- Thank you, Pete
01:39:29Finally, we have Jennifer Penrose, who is a Data Science Manager for the National Center for Missing & Exploited Children
01:39:31Now, NECMEC's mission is incredibly important, and it speaks to the depth of technology integrating with a mission-driven culture to create a human-first process
01:39:36They've been working with Palantir for a while, but their work with Palantir's AIP is actually being made public right here, right now, at the Backstage Pass
01:39:45Let's take it over to Jack and Jen for more
01:39:47Thank you, Matt
01:39:49Jen, great to have you here
01:39:52Thank you so much for joining us
01:39:54I think, I mean, this is such a critical use case, right? This is a very sensitive working environment
01:39:58The subject matter is extremely sensitive, and every decision is obviously mission-critical, right? I think we talk a lot about human-AI teaming, but this is probably the paramount example where enabling your people with that tooling can really make a real, tangible difference in many people's lives, many children's lives
01:40:11So tell me more about the work you're doing, the incredible work your teams are doing, and how is it being supported by this tooling? Well, thank you so much for having me
01:40:23NICNUC serves as the nation's Largest child protection organization, and we are a national and global clearinghouse for reports of crimes against children
01:40:28In 2023 alone, we received more than 36 million reports of suspected child sexual exploitation occurring online
01:40:34These reports come from the general public as well as companies, and we receive reports from all over the world
01:40:44So, as you can imagine, with such a high volume of incoming reports, we have to leverage data and technology to keep up with that daily influx
01:40:49And that's really where Foundry and AIP come in
01:40:55They help us to prioritize, triage, and process these reports
01:40:58So we take one very specific example
01:41:01In recent years, we've seen a dramatic increase in reports to our cyber..
01:41:05Tip line for financial sextortion of children, suspected financial sextortion
01:41:10These are cases where an offender is making demands for payment from a child and threatening to publicly release sexual or explicit images of that child
01:41:18These reports are time-sensitive and urgent
01:41:25AIP can actually identify these reports as they come in and then help analysts prioritize them based on the risk factors found in those reports and the specific details
01:41:33That's incredible
01:41:40I imagine, even in the example you're describing, this is a classic case where all the nuances and subjectivity of the data you're working with, and the nature of this use case as a whole, can really be managed effectively
01:41:48Difference, right, in terms of the optimal decision to make and how to actually handle these very sensitive situations
01:41:56So, specifically with the kind of tooling that your people are using, I imagine having that visibility, having traceability as to how information is being passed, and how that's feeding into a decision, that's pretty critical, right? In terms of actually understanding whether we are making the right choice here
01:42:02Absolutely
01:42:08We really think of AIP as essentially an assistant to our analysts
01:42:14It's surfacing the critical details and data insights they need, with that visibility into exactly where in the reports different information is found, and that then frees up analyst time so they can focus on higher-level tasks
01:42:24Can focus on what they do best, on the areas where their human judgment, their expertise, and their insights add that most crucial value
01:42:28 100%, I think you said it best: focus on what they do best
01:42:32 Honestly, the work your team is doing is absolutely incredible
01:42:37 It's an honor, it's a privilege to be able to support that from a technical perspective, and yeah, very exciting to see what comes next
01:42:41 Yeah, absolutely
01:42:43And we're so grateful for this long-standing partnership with Palantir
01:42:46 It really enables this work, and it would not be possible without these tools
01:42:49 Incredible
01:42:51Thank you so much for sharing your story, Jen
01:42:53 Thank you, Jen
01:42:56 Alright, in just a little bit, we will have the second half of the customer discussion
01:42:58Keynotes for today
01:43:00I'm going to do a little preview of that
01:43:03What can we expect? I think honestly it's crazy that we're only halfway through the keynotes, right? I think we're going to continue to see that engine primitive I was mentioning, like some of the stuff that was teased in the first half
01:43:12You're going to see those same primitives come out but used for a very different outcome, solving very different problems
01:43:19So maybe you could start thinking about, okay, if I look at the archetype on show there, maybe even if I'm not in the healthcare world, maybe that's something that could be reapplied
01:43:26Right? So I think that's one piece, and then I think there's another thing, and it's going to be with… Komatsu's demo, which I’m really excited to see
01:43:33They've been putting a ton of work into really leveling up their workflows, and I think it’s a great example of what we at Palantir often refer to as the levels of AI deployment
01:43:41They started with the kind of AI assistant archetype, like maybe a chat-based interface, then it was proactively suggesting, "Hey, maybe you could change this, maybe you could change that." But specifically, what we're going to see a glimpse of in their demo is moving to that agent-oriented workflow
01:43:55We’re now not even waiting for human input; these agents are working autonomously behind the scenes
01:43:59Scenes, running scenarios, testing different parameters as input to their tooling within their logic, and then actually pushing the recommendations to the user to approve or reject
01:44:12And of course, all of that is fed by an ongoing feedback loop to improve those suggestions
01:44:14So, that level of technical proficiency has just increased dramatically
01:44:17They were also one of those first bootcamps
01:44:19Very excited to see what they have in store
01:44:22And on my part, a lot of what we talked about is speed to value
01:44:24And I think Tampa General really encapsulates that, 100%
01:44:28I don't want to spoil too much, especially for folks
01:44:31It's so hard not to when talking about these keynotes because the work they're doing is so exciting
01:44:35I'll let… They spoil it themselves, but the window between when an old solution was being phased out and when they were able to bring in Palantir was so tight, and it had to be perfect
01:44:45If you're thinking about TAP in general, the users at the end of that are patient
01:44:50These are real-life effects on someone's life, and that is critical work that is happening
01:44:54The work we did with them had to be perfect from the start, leaving no room for error
01:45:03So, when you think about pressure scenarios and time to value, that's such a critical component
01:45:08I believe people will be really impressed and excited about the work they've been doing
01:45:11100%
01:45:14And throughout the rest of..
01:45:17Today, I'm going to be doing more customer interviews and attending all these demo sessions
01:45:20I think there's just so much happening
01:45:22As Dr
01:45:25Karp said, there are over a hundred organizations showcasing different use cases for AIP
01:45:27Great
01:45:29Is that all? Is that all? We need to go bigger, surely
01:45:31What is the energy like for you guys? This is, at this point, your fourth AIP
01:45:33You've been to most AIPs
01:45:34Fifth AIP
01:45:36Fifth AIP, for you guys
01:45:38This is my second one
01:45:40What's the energy like being here? I mean, I think the energy is palpable
01:45:42Firstly, just seeing all these conversations happening in this room alone
01:45:45And literally, as I say that, I get told to stop
01:45:47There you go
01:45:49Probably a good shout
01:45:51But the great thing is, those..
01:45:56Conversations are going to continue
01:46:02I think the thing that I really want to highlight in terms of the mixture of energies here at our Demo Expo that’s happening this afternoon is that we have customers who have been working with us for the equivalent of just about 10 minutes—some of our newest customers using the latest products
01:46:07But also, literally one station over, you'll have a customer like BP working with us for 10 years
01:46:11To have both ends of the timeline, and that timeline will continue long into the future, with both at very different stages of partnership, is truly exciting
01:46:18To have them demoing to each other, sharing ideas—it's a fascinating combination
01:46:21Energy
01:46:23And I will say, just to build off that, the customer interactions that happen on-site at AIPcon cannot be replicated in any other setting
01:46:26No, not at all
01:46:30This is the critical place where people are able to get ideas about workflows that are applicable to their business use cases, that they cannot get anywhere else
01:46:39I mean, I think Meet the Masters, screen sharing, every single conversation that's happening is with someone that they may never have spoken to before
01:46:43And that, to me, is incredibly special
01:46:47Yeah, and I feel that myself
01:46:50I met customers that were here for 10 years yesterday
01:46:53I met customers that signed literally 10 months ago
01:46:56And so the diversity of customers that are here is very exciting
01:46:59We've..
01:47:03Got a lot more to go for Palantir's AI Conception
01:47:06That's it from myself, Jack and Natalie, but we have more keynotes coming up and we'll have a lot more behind-the-scenes footage later today
01:47:08 - That's good
01:47:10 - Thanks, everybody
01:47:12 - Thanks, everybody
01:47:21 - Thank you, everybody
01:47:31 Please welcome Senior IT Manager, Business Intelligence and Technology Information at Komatsu, Bryce Leshinsky
01:47:40 - Hi everyone, good morning
01:47:51Very excited to be here
01:47:56My name is Bryce, I'm with Komatsu
01:48:12I feel very privileged and honored to be able to share with you today everything we're doing in partnership with Palantir
01:48:33When we talk about what we have going in terms of dynamic pricing, I just want to share a little bit about what we do at Komatsu
01:49:15Really, at..
01:49:21Komatsu, we're an industry leader in terms of original equipment manufacturing
01:49:36For many different industries, our customers are in mining, construction, forklift, forestry, and industrial machines
01:49:58So we make equipment and service equipment across a wide range of markets
01:50:09We have a couple of terms at Komatsu that I wanted to share with you today
01:50:14One of those terms is very much synonymous with the "boots on the ground" theme of AIPCon, and that term is the Gemba
01:50:19The Gemba is really, it means where the work actually happens
01:50:24So, truly, what we try to do is understand the challenges at the Gemba, what the true challenges on the ground are, and how we can solve for those challenges
01:50:29Another term that we..
01:50:34Use is Dentatsu
01:50:37Dentatsu means unparalleled and unrivaled solutions
01:50:40So everything we do from an innovation perspective and a transformation perspective, we try to keep Dentatsu top of mind
01:50:43And a third one, just to give you some context in terms of the size and the magnitude of specifically our mining equipment, which I'll be talking a bit more about today, we use the term that if it's not grown, it's mined
01:50:57So everything that you look at, everything you use in your everyday life, was either grown from the earth or it was mined out of the earth
01:51:06We make a lot of big, heavy equipment and are very excited to share how we're looking at that equipment, how we're dynamically pricing parts for that equipment as well
01:51:14So I I wanted to talk a little bit about how we're currently doing our dynamic pricing and what we're trying to achieve in terms of objectives for transforming our dynamic pricing models
01:51:23Really, our challenge is threefold here
01:51:28We have pricing analysts and pricing team members who are receiving hundreds of emails per day
01:51:32And those analysts and the pricing team have to sift through a lot of emails, understand these inquiries that are coming in
01:51:37It could be an inquiry for a new price for a specific part, or a new sales order that was created
01:51:47And really, the team is tasked with going through these emails and working back into the system
01:51:50A lot of manual tasks are involved, right? And there are just a lot of bottlenecks in the process
01:51:54Process with a lot of throughput that could be realized
01:51:57So, we wanted to understand that challenge a bit further
01:52:00We also need to coordinate across global regions
01:52:03Our mining business is very much global in nature, and so trying to do this collaboration across global regions creates a lot of challenges as well
01:52:07We also have a critical mission to maintain machine uptime for our customers out in the field
01:52:14In mining operations, if machines are down, the operations are down
01:52:21So, we need to make sure that we're doing everything that we can for our customers and fulfilling our overall promise and purpose in creating value together with our customers to make sure their machines are staying up in the field as much as possible
01:52:27Possible
01:52:34So, the solution we'll be talking about today really is predicated on being able to gather information from many different sources, both structured and unstructured data, to provide team members with enough information to synthesize and make quick decisions
01:52:40The solution we're going to discuss is based on four main capabilities
01:52:45The first is a shared model—a model that global regions can collaborate on, understand exactly what their pricing strategies are, grasp the nuances and differences across regions, and be able to share that model
01:52:58Collectively amongst one another
01:53:07The second is really an alerting workflow
01:53:16So all of these emails that are coming in, how can we transform that into more of an alerting process so that analysts and pricing team members can effectively do their job, focus more on decision-making, and be able to sift through and increase the throughput on all the information they're gathering on a daily basis
01:53:22Third, we have scenario analysis
01:53:28So, scenario analysis—really being able to look at an individual part or a batch grouping of parts and understand how whatever we're decisioning on for those parts could potentially impact our overall objectives
01:53:34How does that potentially impact our customers? And How do we optimize that in the long run? And then lastly, we have batch pricing
01:53:43Batch pricing allows us to actually take action, either to update the specific price of a part or to notify someone who may be in a different region, enabling communication through this alerting and action workflow, and then be able to push those updates back into our source systems
01:53:56So with that, I wanted to show a demo here on our Pricing 360 tool
01:54:02Today, we have all these alerts coming through
01:54:09Price analysts are reviewing these alerts, and really, where we're heading next is how do we utilize AI agents to process these alerts for us
01:54:15AI agents would have the capability to look at all..
01:54:20These alerts are based on simple logic and rules that we could define as analysts and team members
01:54:24The agent actually processes these alerts and then identifies the ones that require further investigation
01:54:28In this case, I have an example part that I need to investigate
01:54:32I can see that this particular part has an outlier price compared to similar parts in terms of description, and it is also missing segment codes
01:54:41So, we'll take a look at this specific part and see what we might need to investigate
01:54:49Going into the material detail screen here, I can review all the information I need on this particular part, including the price, the gross margin, and the validity of the part
01:54:57Also, the division that this part is associated with—I see that one particular issue is with this part
01:55:01It's exceeding a maximum threshold, and this is based on the suggested price that my AI pricing advisor is making to me
01:55:11The problem here is we don't want to necessarily outprice ourselves from the market, and we want to keep our customers satisfied, not increase the pricing margin by too much
01:55:21So, in this example, we'll look at how we can take the suggested price, understand how our AI agent is providing this price, and determine what we might need to do to have a more proactive pricing recommendation made to us
01:55:35So here, I'm going to instruct my agent to go ahead and regenerate a new price suggestion
01:55:40And this is..
01:55:44I'm going to trigger a chain of thought logic so that I can review all of the context and information that the AI agent is actually processing to make better recommendations based on my feedback to the agent itself
01:55:54I can go ahead and look at all of the information that the agent is gathering and processing
01:56:25It's utilizing tools like querying the ontology, and it's also using tools like the calculator to formulate better recommendations for me
01:56:27Given the context I've provided, it recognizes that the price increases are exceeding our threshold
01:56:29As I review all the information the AI agent is processing, including the feedback I've already given and the feedback I will continue to provide
01:56:31It will be maintained and utilized by the model going forward so that it continuously trains and we can have continuous improvement as well
01:56:32So with that, I'll be able to look at everything that the agent is going through
01:56:38It's going to produce a new suggested price that I can evaluate
01:56:43And I'm going to look at that new suggested price, all the considerations that it took, and be able to determine whether I want to keep this new suggested price or if we need to think of another price to come up with
01:56:48So, looking at the results here, I can see that the AI agent has now taken into account everything that I've instructed it to do
01:56:56It's now within the threshold for maximum price increase, and I would like to accept this suggestion
01:57:00So, going..
01:57:04Back to our material detail screen, I can get the context and the new price suggested, and I can go ahead and submit this one for approval
01:57:08Going to submit this for approval, I can review everything in terms of what is being updated for that particular part
01:57:11I can assign an approver to it, and then I can go to my approvals inbox to see who this has been assigned to for approval
01:57:20That approver can review the updates I wish to make, as well as all the additional updates that our AI agent has already processed
01:57:27Once this gets approved, it will ultimately trigger a write-back to our source ERP system, and the part will be reflected with an updated price back in our system
01:57:35We're very excited about this solution and really what's next here
01:57:39We want to be able to scale this out to all the segment code differences we have across our materials
01:57:50We want to make sure we're consistently pricing our parts more dynamically based on market conditions and what our customers are responding to us with
01:57:59We also want to expand this not only across our Mining Business Division but also our Construction Equipment Division and other divisions that can benefit from the same kind of solution
01:58:07With that, I'd like to thank you all very much for your time today
01:58:12Specifically, I'd like to thank Justin and Pedro
01:58:14Pedro is here with me today, and I appreciate their tremendous leadership
01:58:17Business on this effort
01:58:20And also, our Palantir team—Jordan and Marshall, for sure—have been here every step of the way, as well as Brad, who has helped us out a lot
01:58:27So, we really appreciate the partnership with Palantir and are excited to keep moving forward
01:58:29Thank you
01:58:32Please welcome Vice President of Cognitive Services at Aramark, Dan Klein
01:58:34Hello, good morning, everyone
01:58:36Today, we're going to spend a little time giving you a brief history of what Aramark is and what our company does
01:58:38We'll discuss our foundational use case with Palantir and then what we see in the future
01:58:48If you're not familiar with Aramark, we are a Fortune 500 company that provides hospitality services in 15 countries
01:58:54We have over..
01:59:00200,000 employees and 10 business lines
01:59:06Each of those business lines provides a specialized service based on our client needs and consumer preferences
01:59:12And so, we'll talk a little bit about how that creates a variety of data for us to deal with and work through
01:59:24Again, at our core is service, and so our business is focused on service and the services that we provide to our clients and customers
01:59:30When we do that from a hospitality perspective, that means we have a tremendous amount of diversity of systems out in our field, both on our supply side and our sales side
01:59:47And so, we've always had an opportunity to better understand our data and understand how that functionally impacted our business
01:59:51When we think a little bit..
01:59:55More about our data: we serve over 2 billion meals a year
01:59:58There are millions of products that make up those meals, thousands of recipes, tens of thousands of recipes that contribute to the overall product, and we're all trying to understand what the preferences are for a single consumer
02:00:15And so, we've always looked for opportunities to understand this data better and longitudinally make recommendations on what we can do to better serve our clients and consumers
02:00:26This is where our relationship with Palantir started
02:00:32Again, using the power of Foundry and large language models, we created master products
02:00:38And so, if you think about the number of products, and you can see some of these in the demo, that have very… Similar names but identify as our own unique product
02:00:50And so, we were able to use large language models to match products and take things that were notionally different and match them together
02:01:01And you can see that in the demo
02:01:07Our initial matching found that about 30% of our products could match at more than a 99% confidence level, which is a fantastic, fantastic start
02:01:13We used workshops to perform manual matching from there and train the models to further improve future matching
02:01:23As we gain additional confidence in the matching, we're continuing to move more and more down the stack
02:01:29Right now, anything above a 99% confidence level auto-matches, so it has no human interaction
02:01:35That was our initial point
02:01:41Continue to move that down as we see better results from the, uh, from the model and able to validate that
02:01:46When we think about what we do with the products that are now mastered, we use our ontology to link those products together
02:01:56And so, we can see the products that come in from our buy side, link them through recipes in the middle, and then ultimately see what's happening out on our sell side
02:02:07And so, we've created a full lineage of our data from buy side to our finished products
02:02:12This allows us to perform detailed analysis at the atomic level and really understand what's occurring, both from a financial perspective and from a consumer perspective
02:02:22It provides a new level of visibility to our products that never..
02:02:26existed before and allows us to perform analysis with ease
02:02:30We also looked for opportunities to enhance our data
02:02:34And so, we can see in this, this is our Classification Manager
02:02:38And so, there are certain elements of our data that never existed, right, that didn't come through
02:02:44So, certain things like what the meal type was or where it was served, we use the Classification Manager to add that
02:02:57So again, using the power of LLMs, we do auto classifications
02:03:02We supplemented that with a manual process to train the model
02:03:07So, we did a manual process on about 1% of our products, and we're able to automatically provide classifications for the other 99%
02:03:12This is a significant time savings and provides a level of analysis that we never..
02:03:17We had before
02:03:22Again, now we're able to go down to micro and macro levels to understand our data, how it's used, and how trends are changing based on our consumer preferences
02:03:25Our journey is just beginning with Palantir
02:03:29So we did this in about nine months
02:03:32It was something that we thought would be very difficult to do, but with the use of Foundry and our partners at Palantir, we were able to do it really quickly and continue to build out use cases on this foundational data set
02:03:48Obviously, there are always financial benefits that come from data sets like this, but we also look at the user feedback from our analysts and see what they say about it
02:03:57And some direct quotes from them are, "I use this data every day before..." Meetings with internal and external stakeholders to better understand our business
02:04:07We can quickly and efficiently respond to queries with accuracy and trusted data
02:04:11And one of my favorites is, "I use this data every day to help myself learn about our business." We always want to think about where we're going to go and what our future plans show
02:04:21And we have vertical and horizontal growth with us
02:04:25So we're going to continue to add data sources, obviously, but we also plan to add unstructured data to understand our commitments to our customers and voice of the consumer
02:04:35These sources will provide greater coverage and additional data that we can use to make decisions
02:04:40We also are really excited about the idea to..
02:04:45Predict and simulate within our business
02:04:50And so we are on the verge of creating a full digital twin where we can understand what's going to happen in the future before we make operational changes
02:04:55As I wrap up, I'd really like to thank the partners at Palantir that helped make us successful and the Aramark team members who learned a new technology and took us and executed on this old vision
02:05:04Thanks a lot for your time
02:05:07Please welcome Executive Vice President of Information Technology at Associated Materials, Duane Massey
02:05:10Good morning, AIPcon
02:05:12Thank you all for being here
02:05:14Hey, I got one
02:05:15One good morning
02:05:18Thanks
02:05:20That's good
02:05:22I'm Dwayne Massey
02:05:25I'm the EVP of IT for Associated Materials
02:05:27We're a leader in the exterior
02:05:32Building products market
02:05:36Think windows, think doors, aluminum siding, vinyl siding, things like that for your house
02:05:40We cover residential, commercial, remodel, and new construction markets
02:05:46It's honestly an honor to be here today to talk about the journey that we've been on with Palantir
02:05:52This is my second go-around with Palantir in two different companies, and this one is by far the most ambitious and fundamentally ground-breaking for a business that I’ve seen
02:06:10We are an acquired portfolio company for a private equity group, and we're in the middle of transforming our entire organization
02:06:15What that really means is, for us, transformation is how do we get the most out of our business? How do we operate? The best that we can do as a company? How do we drive change as an organization that's been around for 75 years? I've coined the phrase, "We're a 75-year-old startup," and it's true
02:06:38Imagine if you had a company with all this legacy information: legacy systems, AS/400, mainframe, Microsoft Access, Excel, flat files
02:06:42That's us
02:06:46That's what we've got
02:06:49We haven't had significant technology updates in almost 25 years
02:06:52It's been a very long cycle for the company
02:06:56We're making the investment in the company because it has so much potential in the future
02:07:01The growth is phenomenal, and the opportunities in front of us are just amazing
02:07:07If I talk a little more about the company, the drivers that are really impacting us are..
02:07:13Visibility to operations, visibility to how the company actually works
02:07:19If you think about everybody here, your role in the company is, if you're in HR, it's usually a who, a what, a how, a when, or a where
02:07:24Palantir gives us the why
02:07:30Palantir gives us the ability to think about the why
02:07:35And as we go through, I think there are really two challenges that we have in the Palantir market
02:07:40The first, for me, is personal, and it's getting my Palantir techs who come from Palantir to understand my jokes since I'm 52 years old
02:07:45It doesn't work
02:07:51I spend so much time explaining jokes; it's unbelievable
02:07:54The second is, we have old systems
02:07:58Our oldest in-use application was built two years after I was born
02:08:01It wasn't me, I swear to God
02:08:04We didn't plant that so that I could get a job later
02:08:07We are still connecting and operating on mainframe systems compiled in the 70s
02:08:12Our systems have seen the lifecycle of their parent companies go up, hit the growth cycle, and come back down and no longer exist
02:08:17So, think about that from a company perspective, where you are driving change
02:08:22You are looking for something that will actually impact the bottom line
02:08:27You have to act fast
02:08:31Palantir gives us the ability to do that
02:08:34Our biggest driver for Palantir is time to market
02:08:37And I'm going to give a great example real quick
02:08:40One of the biggest failures that we had as a company, we're 75 years old
02:08:42We make millions of windows a year, millions of squares of siding, aluminum trim, all..
02:08:47The different components that go into your house
02:08:52We have no master customer system
02:08:57We have big consulting companies coming in, saying, "Oh, it'll take 12 months and we can do this
02:09:02It'll cost millions of dollars." Or, "We can do this in 24 months
02:09:06It will also cost millions." We solved this in six weeks on the pilot just by connecting to those legacy systems and tying that data together
02:09:11This is real
02:09:15The way I think about it and the way I communicate it to our executive team is that Palantir does what we want to do in our minds
02:09:20Why can't you just take data set A and connect it to data set B? It's easy, just copy and paste it, just make these systems work together
02:09:25It doesn't work
02:09:30It never has
02:09:35When we look at how we grow as a company, scale, Culture, all of this, we are trying to change the way people think
02:09:40If you think about Palantir, you think about the ontology layer
02:09:44You're now looking right to left
02:09:49You're looking at the outcome and then you're looking at the outputs required to get the outcome that you want
02:09:51Palantir enables that naturally
02:09:54It's a different way of thinking
02:09:56The biggest thing that we can do to our culture to help change is to take out words
02:09:58Take out the phrase, "the problem with," nope, I can't
02:10:02Oh, I gotta do this
02:10:06Nope, I gotta do that
02:10:10Take out the negative and focus on the positive because there is nothing that we haven't seen that we can't accomplish with the Foundry platform
02:10:15We have connected the systems, like I said, that are 50 years old
02:10:18Old
02:10:22We've connected the new systems, old systems
02:10:25The biggest thing that we have accomplished, though, is empowering our people to make the right decisions
02:10:29So, when I go through the demo, I'm not going to give the demo as the guy who built it, because that's not me
02:10:38Like I said, my IT career started with the OG of internet companies, back at AOL in the mid-90s
02:10:42But I'm going to focus on this from a business perspective
02:10:46How does an operations leader look at this? And our demo is really focused on supply chain
02:10:49We are an end-user consumer products company
02:10:52We make products that go to your house
02:10:55If you've ever had a remodel done on your house, you know you need products delivered on time; you have contractors waiting
02:11:00Green line at the top of the screen, that's our windows on time in full delivery
02:11:05We're running at 90%
02:11:08That was 40% before you put in Palantir
02:11:12The yellow line is our Vinyl Metal Siding business
02:11:16You can see where we put in Palantir on the blue dots
02:11:19It is not only growing, it's sustaining
02:11:21That's the key with Palantir: you can sustain the change
02:11:24It's not a one-and-done change
02:11:27We have visibility of the whole supply chain
02:11:29We can see a factory, understand what that factory is doing, see if there's anything overdue
02:11:34This is what my workload is; I'm driving the floor now
02:11:39People are no longer having to think, we've taken the phrase "I feel" out of the question
02:11:44It's no longer, "Geez, I feel like I have too much inventory." I feel..
02:11:47Like, I have too much
02:11:50This mic was saying earlier he had 60 on time
02:11:53I wish we would have had 60 on time; we do now
02:11:58Because we can see the data, we can pull the different systems, and we can see the disparities between those different systems
02:12:02We've built these different platforms and these screens to identify what needs to be made, and when it needs to be made
02:12:11Take the challenge away, take the choice away from the user, and utilize AI to identify here's what needs to happen and when it needs to happen
02:12:21As you go through this, we also have the ability to identify here are purchase orders that need to be cut, here are the work orders that are out there, here are the products that need to be ordered, and here is the material that we're behind on
02:12:34We're identifying by warehouse, by location, by group, what is the failure point, what is the constraint, how many days before we're out of product, how many days before I need to reorder
02:12:44Everything that's in an ERP system, we're doing in Palantir today
02:12:48So if you're here and you're an ERP reseller, I'm sorry, your time's over
02:12:51A lot of you guys here, I know, have legacy systems
02:12:54You can do more in Palantir in less time than it's going to take you to implement a new ERP system
02:13:01For us to implement ERP across 11 factories, we're talking almost five years
02:13:08This is less than nine months of work, and we've implemented over 10 business use cases already
02:13:11We are seeing immediate bottom-line results
02:13:14We're driving on-time delivery
02:13:17Our Customers actually like us
02:13:20We have actually interviewed our customers
02:13:23The reports went from, "Oh man, it's like a Ouija board when we order stuff," to, "Holy cow, they actually deliver on time
02:13:26It's amazing." We're driving that kind of change
02:13:31We know what our inventory is
02:13:35We know what our materials are
02:13:38What does this mean for us as a company? The big thing is we can focus on ideation
02:13:41"Man, I have an idea
02:13:44Here's how I can make it better." Think Six Sigma on steroids
02:13:47Continual improvement, growth, change, opportunity
02:13:51Item and order mastering
02:13:54So, in summary, we have a very unique implementation
02:13:58We're touching legacy systems, but we are transforming absolutely everything we do within a manufacturing company that has been here for years
02:14:00Thank you for your time today
02:14:03I'll be around later
02:14:05Have a great day
02:14:08Please welcome from Tampa General Hospital, Chief Medical Officer Dr
02:14:11Peggy Duggan and Chief Quality Officer Dr
02:14:14Laura Hobner
02:14:18Well, good morning, everyone
02:14:21We're super excited to be here and to talk about the work we do with Palantir, specifically in the healthcare space
02:14:25I'm Dr
02:14:28Peggy Duggan, the Chief Medical Officer at Tampa General Hospital
02:14:32With me is Dr
02:14:37Laura Hobner, who will join us a little later in the discussion to show our demonstration
02:14:42So, we're here to talk to you about a sepsis program that we built out in a very short span of time with our partners at Palantir, which is saving lives every day at Tampa General Hospital
02:14:59But before we get there, let's tell you a little bit about where we work
02:15:06Tampa General is a large academic health system in West Central Florida
02:15:10We're the only academic health system in our region, which creates a real burden and importance for us to be there for the patients who need us
02:15:17We're a six-hospital system, a thousand-bed academic hospital, three community hospitals, and two specialty hospitals, along with a full suite of ambulatory services across the state of Florida, from walk-in clinics and urgent cares to primary care and complex outpatient patient care
02:15:28So, everything you can imagine a patient needs, they can receive at Tampa General
02:15:34What I love about the partnership with Palantir is the alignment with our..
02:15:39Shared purpose and our vision
02:15:45So, at Tampa General, we say we heal, we teach, and we innovate, and we care for everyone every day
02:15:50Our vision is that we will be the safest and most innovative academic health system in America
02:15:56Both of those goals align beautifully with the work we're doing with Palantir
02:16:01So, we're going to talk about sepsis, and I’m going to explain a little bit, especially to those who are not healthcare professionals, what sepsis is
02:16:08Sepsis is a severe immune response to a blood-borne infection
02:16:13What that means is when you have a blood infection in your body, your systems go into overdrive to try to protect you from that infection
02:16:23It's a bit of a war inside your body, and everything is being shunted to prevent that
02:16:29Infection from worsening, but it's starving the rest of your body from the things that it needs
02:16:34So, less oxygen to organs, less blood supply, and then the organs start to fail and shut down
02:16:39What that means for a patient is the very thing your body's trying to do can also cause harm and can actually lead to death
02:16:50And a pretty significant number of people die annually, about 350,000, and 1.7 million across the country are affected every year
02:16:55It is the single most preventable cause of death in U.S
02:17:01hospitals, and hence the purpose and why it's so important for us to focus on
02:17:06So, in 2022, TGH launched an early warning system to detect the subtle signs of sepsis, because the other thing about sepsis is when we know it, we know
02:17:10How to treat it
02:17:15What's difficult is identifying sepsis
02:17:19There are many parameters we look at
02:17:23Most of the time in the hospital, physicians and nurses are looking at pieces of data statically and assimilating it throughout the day
02:17:29What we're seeing with our technology is that we're often missing the early signs of sepsis, and with sepsis, early intervention is what saves lives
02:17:34That's what makes the difference
02:17:39The other important point with sepsis is that we know how to treat it, but you have to do it correctly
02:17:44For simplicity's sake, let's say there are 10 things you need to provide to a patient to treat their sepsis — you have to do it in the correct order
02:17:57So you can't do 3, 7, and 10 before you do 1; those mistakes can lead to worsening conditions
02:18:02To poorer outcomes, which lead to organ dysfunction and death, and failure
02:18:06So, what we did was put together a sepsis team led by what's called a Rapid Response Team
02:18:10The Rapid Response Team is a group of critical care nurses who are constantly monitoring our patients across our thousand-bed hospital, looking for early signs of sepsis
02:18:20They have made a tremendous impact on the lives of our patients, and by conservatively estimating, we save about 300 lives a year—people going home to their families who otherwise would not
02:18:27But we had a problem
02:18:35Our original partner, who set up our sepsis program with us, was no longer able to support it, and we had learned how important this was for our team
02:18:42Members, for our physicians and for our patients, this tool that allowed them to make better diagnoses and treat the patients more quickly was quickly going away
02:18:54So we turned to our Palantir partners and said, "Can you help us?" And this is where Speed to Value comes in
02:19:04We knew we couldn't go a day without this; we knew how important it was for our patients
02:19:09So we sat with Palantir and identified the opportunity
02:19:14The entire process, from scoping the initial work to onboarding and deploying the production workflows, was completed within six weeks, with plenty of time to implement prior to the loss of our previous service
02:19:24What we call this is a forward-deployed approach to clinical workflows
02:19:27And you see here, one of..
02:19:31Our rapid response team nurses use the platform in real time
02:19:34The solution, the engineers work side by side with our rapid response team nurses, not just to rebuild what we had, but to look at what we had, spend time with them, examine the workflows, and build improvements based on what they were looking for and what they felt would make their lives better
02:19:54The rapidity really comes from breaking down the barriers caused by leaders getting in the way of what we're trying to achieve and allowing our frontline teams to work directly with Palantir's engineers to build the product for their use
02:20:03That’s what enabled us to be so quick with Foundry, allowing the team to see in real time their suggestions being implemented, and to evaluate if it was effective
02:20:13Working the way they thought it would
02:20:17They can completely and continuously iterate over time to really drive change and improvement
02:20:21And here's Dr
02:20:24Laura Hobner, who’s going to show you how this works day to day
02:20:26Thank you, Peggy
02:20:28So, good morning, everyone, barely
02:20:30And I’m here to give you a demonstration of the product
02:20:32Key information, as you heard, is that what’s important is both identifying and then treating, and doing that in a timely manner to save lives
02:20:40So, this comes from the seat of the rapid response team nurse
02:20:48This is the Sepsis Hub that we utilize, and it is based on sepsis, what we call sepsis criteria
02:20:55Three key areas: the patient’s vital signs, the combination and ranges of those, and their..
02:21:01Abnormalities, a suspected source of infection, and also evidence of organ dysfunction
02:21:07You can see here that we are able to identify just what is triggering that criteria and when it happened, and in addition, the trends, which is really key
02:21:19Take patient Naomi Kim here, who has escalated to the top of the sepsis concern list
02:21:24In addition to those three areas, and in addition to the vitals that I showed you on the previous slide, it also brings in unstructured data
02:21:34That's a game changer for us
02:21:43Prior to this, we relied on structured data, and then our team, a very highly skilled team, had to spend a lot of time hunting and pecking through the chart to look for information that supported the diagnosis, given some of the..
02:21:48Key vital signs or other lab values
02:21:52Behind this, before I go further, this is really all the work that's been done
02:21:57This is our sepsis data pipeline
02:22:03You can see the complexity of this, and that is what's bringing all that information together for our nurses to be able to paint that picture and start making a decision about whether this patient truly has sepsis and needs to quickly get onto the pathway
02:22:18Again, back to their workspace
02:22:33With this, they are now able to start thinking about, well, is this real? What do we need to do involving the providers—the doctor's or patient's physician—and asking, do you agree that a sepsis alert should be placed? They place that in the system, as you just..
02:22:44Saw
02:22:55And with that, now that initiates a treatment pathway and our ability to start monitoring that and get this essential treatment to the patient in a timely manner
02:23:06Let's fast forward an hour, and an hour later, now we see the treatment pathway detailed in a nice timeline
02:23:12This was actually a really special request from our team because previously they relied on alarms and alerts that told them when time was expiring for a specific treatment or they had missed the ideal timeline and had to catch up
02:23:29Now they can anticipate and see what's coming
02:23:35In addition, you can zero in on the red line, which represents the IV fluids, one of the key elements of the treatment
02:23:40They can also make modifications
02:23:46In this case, with this patient, the..
02:23:52Physicians felt we needed to be sure we had certain criteria related to blood pressure and to monitor that more closely to really customize the care for the patient
02:23:57They’re able to identify that physician’s request and also document it in the record to make it understood
02:24:03What’s key about this is that not only are we treating the patient, but we’re also able to take all this information and feedback, tie the alerts and actions to what's happening, and keep feeding the system that learns from it
02:24:20So, in the end, not only are we treating the patient, but we’re also improving our Sepsis Hub and ensuring that we develop more accuracy, fewer false alarms, and make this workflow even more ideal
02:24:27Again, just like in the..
02:24:35Pipeline, you saw for identifying sepsis in the patient, involves a very complex data pipeline also for the treatment algorithm
02:24:42Now, we can go back to the patient and see that the criteria have been met
02:24:48The nurse was able to go to the bedside, initiate the IV fluids in the way that the physician had authorized, make that change, and note that
02:25:05Again, this overcomes the previous limitations of the electronic health record and our care processes, care delivery processes, and helps us better understand the work being done when it needs to be done
02:25:22Please recognize that, again, our previous algorithms did not give us this much insight into the work that we were doing
02:25:28Common condition of sepsis and being the leading cause of death for patients in the hospital, preventable cause of death
02:25:40We're improving the ability to make what is often a very difficult diagnosis, making it more readily identifiable, and then implementing a series of treatments and additional testing more effectively
02:25:54So, all the while that we're doing that, the end user's actions and the patient's journey are being measured against the course of all the patients
02:26:02So, what you really have is an improvement multiplier
02:26:10With every patient we treat, we improve the care of the next patient
02:26:18That's really the power of this, and we've seen great results from it
02:26:21Our mortality from sepsis is half of what's quoted nationally across the country
02:26:25All health systems
02:26:28We're getting patients home faster to their families and back to their lives
02:26:33We've reduced the length of stay, the time they spend in the hospital, by 15% for patients with sepsis
02:26:38And we're confident now that, as we continually learn, we're going to improve those results
02:26:43Those results are going to get only better
02:26:49So, we're now applying this approach to other clinical conditions and helping our team be successful in the care they provide
02:26:54And that reflects their commitment to their work and to the patients in our community they serve
02:27:00Thank you
02:27:05I'm really honored and appreciative of the partnership we have with Palantir, and I look forward to great things to come
02:27:09One more round of..
02:27:13Applause for everybody
02:27:17From our first FoundryCon to now our fifth AIPCon, this day feels like welcoming you all to our home
02:27:33And since our founding, the forward-deployed engineering model, where we're boots on the ground with our customers, that feels like family
02:27:44This community is a family
02:27:55And in that context, we challenge each other, we push each other, we're brutally honest with each other, we share in each other's wins, we feel the pain of each other's losses, and ultimately, day after day, we keep building
02:28:06We have so much more coming for you in the afternoon
02:28:12Thank you for joining us on the livestream
02:28:18Thank you for being here today in person, and we'll see you back over there.