American Companies Have 36 Months to Go AI-Native or Get Left Behind | Drew Cukor, TWG AI

The same tools that slowed the U. S. military down in Afghanistan (PowerPoint, Excel, email, and Word) are now slowing American businesses down in the AI race. Drew Cukor spent 30 years as a Marine intelligence officer, helped build Project Maven into a battlefield command and control system, served as Chief Data Officer at JP Morgan, and is now leading AI transformation at TWG AI. In this episode, he joins Craig Smith to make a case that most enterprise AI strategies are fundamentally broken, not because the technology isn't there, but because companies are storing their data in Microsoft file folders where it becomes inaccessible to AI, appointing AI officers who block progress rather than enable it, and mistaking chatbot deployments for transformation. Cukor's prescription is specific: take a company's core workflows apart, how it acquires customers, delivers services, handles back office operations, and rebuild them from scratch with AI embedded throughout, protected inside Palantir Foundry, delivered within 36 months, with the CEO owning the outcome rather than delegating it to a CTO or a made-up AI officer role. The stakes, he argues, are not abstract: China is going AI-native from the start without the legacy infrastructure that's slowing American enterprise, token spend is approaching the cost of a human salary making poorly designed AI workflows as expensive as bad hiring decisions, and the window for acting is closing. The most important video he recommends any business leader watch isn't one where the AI wins, it's the footage of Lee Sedol losing to AlphaGo and realizing mid-game that he no longer understands how the game works. That moment, Cukor says, is coming for every legacy business that doesn't move now. Subscribe to Eye on A. I. for weekly conversations with the people building and deploying the future of AI.

• 0:00Speaker 00

  I took my desktop computer to Afghanistan and I proceeded to do my mission as an intelligence officer, primary intelligence officer with 3,500 Marines using Word, PowerPoint, Excel and email. Now, that's probably fine for 2001, but was it fine in 2010? Is it fine in 2015? How about 2020? How about 2026? It is not fine to move little chicklets around a PowerPoint slide. Now, I want to offer the same thing to the business world. Is it fine to do your asset management work in Excel? And then bury all of that inside in an Excel spreadsheet and file it away in a Microsoft file folder system? And put it on SharePoint? Because when you put your data inside of these Microsoft products, you're basically putting a tombstone on it because that's where data goes to die. This is a disaster.

• 0:43Speaker 01

  You think it's important that enterprises throughout the economy adopt AI in order to keep ahead of China? Can you talk about what are the risks?

• 0:52Speaker 00

  We need to face them in a competitive way. They're adopting AI much faster than we are because they're literally going AI native from the start.

• 1:00Speaker 01

  Legacy companies, if they don't make this transition quickly, having to face AI native startups in their industries. Let's start by having you introduce yourself and give that background, your work at JP Morgan and then how you got to TWG. Then we'll go back and sort of run through my Maven questions and then get to TWG, which is not unrelated.

• 1:31Speaker 00

  Yeah, hi, I'm Drew Cukor. And I spent 30 years in the Marines as a young officer, all the way through the rank of Colonel. It was a wonderful time. It was a great career. I wouldn't change a thing about it. And I was an intelligence officer during that time and had the privilege of serving with phenomenal Marines and soldiers and sailors and airmen and great commanders and great Marines that served with me. And, you know, I just, I look at other people and I, I just say, I'm so lucky I got to do that. So, and then I left, I left the Marine Corps because of, you know, retirement, mandatory retirement age.

• 2:18Speaker 00

  And I got, I got a chance to work at JP Morgan and, and get to serve there. I had a chance to work with Jamie Dimon and Lori Beard and all the business CEOs. And, you know, my job was to help with AI transformation. And, you know, I had a, a large team of data scientists that I had the privilege of working with and a large team of engineers and we built some great stuff.

• 2:42Speaker 00

  And then I had another chance to come back to California, which is where I grew up. And I'm actually coincidentally here today in, in the Santa Monica office that we have. And now I work for the Walter Group, TWG AI, and have the privilege of leading another 120 or so engineers and data scientists and working on some of the world's most interesting workflows. Not quite as complex as what we did in government, but definitely amazing workflows around asset management, investment banking, insurance, sports, ticketing. So it's, it's very exciting.

• 3:18Speaker 01

  Yeah, and on the TWG, which we'll get to, it's, it's timely because we're in this scaling up period. And there are a lot of issues in how companies are scaling what they need to do to be successful. I've been talking to a lot of people about that.

• 3:42Speaker 01

  But let me jump back to, to Maven since I'm so fascinated by that. At, you know, I stopped, as I said, I, I had Jack Shanahan on and Bob Work. And then I stopped sort of tracking it back then just because, you know, the world moved in a different direction. But, but it grew significantly since I stopped tracking it in that it initially was a computer vision project for drone surveillance

• 4:22Speaker 01

  to make it easier for analysts to go through hours and hours of footage by, by identifying objects that they could check. But, but it grew into a sort of whole battlefield management system. And now I guess Palantir is involved.

• 4:47Speaker 01

  Or, or I don't quite understand if Palantir is providing software and Maven still is with the military. But can you just talk about how the project developed under your leadership? And, and where you see it today? And is it as core to the military's AI transformation as it, it appeared at the beginning?

• 5:19Speaker 00

  Yes, I'm happy to help with that. And look, a lot of this is in tons of literature.

• 5:26Speaker 00

  Obviously, I left the government in 2022. So it's been a few years. But remember, the revolution that we were going after in government

• 5:38Speaker 00

  was that software is actually a weapon system. And, and by weapon system, I mean it in the sort of acquisition phrasing of it. Most people consider software to be a business thing like Microsoft Office. And so it has a diminished status and is not considered important. In fact, in many circles, software is something you just have a bunch of engineers build. And it's free, like my email's free, a lot of my AI is free. Why, why is software such a pain in the ass? I want to get back to my tank or my fifth generation aircraft or my $5 billion aircraft carrier or my rifle or my piece of hardware.

• 6:22Speaker 00

  And I want to live there because I understand that world. But our revolution, you know, should it have emerged? And I think it is. I mean, this is something that's going to grow and there's going to be many hands. I got to row for a while in the canoe. And it was a great honor and privilege to do that. But like, there's a lot of people that are rowing very hard and working very hard on this. But software is not something that's free. It's not air. It's, it's, it's a real important piece of how wars are planned, considered, fought and won, frankly. It is frankly, the, this, the, the, the enabling capability

• 7:05Speaker 00

  that brings all those very expensive pieces of hardware to what it needs to, what they need to do. And it isn't something that you just dismiss. And, you know, it's something that has to be seriously and thoughtfully considered. And that was what we were pushing for. So in the case of Maven, you know, yes, we were doing computer vision work. But like, that doesn't, that's, that's one component of a larger system of software. The detections and the classification and the reasoning, whether it's with words or numbers or with, you know, video or still images. Those are, those are data streams that are coming in,

• 7:44Speaker 00

  but ultimately has to land in front of a human and interact with the human in critical ways for decision making. And so that's what we were always building. It wasn't just, you know, a computer vision algo. That algo had to go somewhere and it had to deliver detections and it had to deliver knowledge in a way that the human could process it and do something with it.

• 8:08Speaker 01

  Get off the subject, but I, I was in Ukraine two months ago talking to people about sort of the, the state of the art of their drone development or military drone development. And it came through that despite all of the protestations that, that, you know, we need a human in the loop, that it's moved to a human on the loop.

• 8:39Speaker 01

  And that they're very close, a lot of them told me, to fully autonomous weaponized drones that, that can hunt for targets, a military vehicle or a personnel within a geofenced area. Do you think, I mean, and, and that's pretty, that technology is, is, is solved largely? I mean, of course, there's the being able to differentiate between friendly and enemy. But is that, is that a direction you think the, the world is going generally?

• 9:23Speaker 00

  Look, autonomy is not new to the US military.

• 9:29Speaker 01

  Right.

• 9:30Speaker 00

  We have landmines, claymore mines, okay? We, we position those in critical areas. They're designated, they're carefully mapped and they're triggered autonomously. Someone hits a wire, it goes off. We have other systems similar to that. So this is not known or unknown to us. Like we have a long tradition in the US military and in the armed forces of understanding these platforms. And we, we build rules around them so that we can fight under the law of, of war, which is, you know, a Geneva Convention approved, uh, you know, that came out of the First World War and all the violence that that created.

• 10:13Speaker 00

  And, and, and all of humanity came together and said, look, we've got to put some rules on this because never again do we want to see this sort of, you know, this type of conflict. So, look, what Ukraine is doing is certainly right there on the bleeding edge of warfare. And we're all watching and studying. I am, I know we're watching what's going on in Iran as well and other conflict zones where we're seeing, you know, people taking, you know, big moves in terms of like, look, if I geofence an area and I let these things fly, definitely like, we're going to watch that. I have great confidence in our system that's going to put rules around this and it's going to run this carefully. But back to your, uh, point about, you know, command and control system, right? We want to make sure people understand, like, at the end of the day, like there are humans in the loop.

• 10:58Speaker 00

  There are people that are very thoughtful on this. I need people to know that when a commander is on a battlefield, he is responsible for what happens there. And he'll carry whatever occurs on that battlefield with him for the rest of his life. You know, many of us walked off of battlefields and these are sacred spaces, by the way, where human lives are taken and no one takes that unseriously. We all take this like, like this is a real thing and we'll live with this for as long as we're alive. We will dream about this at night. It'll affect us forever. And these commanders know this and they sit and they preside over these, you know, conflicts. And when they're existential, in the case of Ukraine, that's getting hammered by, you know,

• 11:40Speaker 00

  the Russia or wherever the conflict might be, like you will take, you will make decisions and you'll live with those decisions and you'll be subject to the law of war and you'll be subject to the conditions of that. So just don't, you know, we don't want to give the impression that these things just happen willy nilly. They're done for reasons. Maybe there's manpower shortages. Maybe this is, you know, there's a critical moment in the battlefield where it just has to go weapons free. We don't understand and maybe they've cleared the space and they've cleared it so there's no civilians. Could there be implications where a, an adversary could just switch, switch and just go autonomous and run these things in New York City? God, God save us from that, which is why we've got to build the preventative measures, right? So that we can hunt these things down and find these things if they're, you know, suddenly let loose in our,

• 12:26Speaker 00

  you know, wherever it might be. But just to be sure, like, let's emphasize, like, these are human conflicts. There's moral agency here and people live with these, uh, the repercussions of these decisions. And it's not just something that a Ukrainian officer is just flipping as a piece of technology is doing this thoughtfully and he understands the consequences.

• 12:45Speaker 01

  Okay, well, let's move, uh, on to, uh, TWG AI. And, and the scaling of AI across, uh, in, across enterprise and particularly, uh, enterprises where, where the, the, the software stack is critical to its business. And, and can you talk about that? How you carried your, you, what you learned at Maven and then JP Morgan to TWG? And, and what are the central problems you're, you're looking at right now?

• 13:26Speaker 00

  Yeah, um, look, I think we started with this software element.

• 13:33Speaker 00

  Now, what happened in the nineties is we invented this Windows system. And obviously Bill Gates and team led the charge and just did amazing things for the US economy and the world, and the global economy by bringing this software interface that could, could interface with computer hardware and really deliver important results. That was an innovation in the nineties. We all remember Windows 95, we all remember WordPerfect, right? And Corel Draw, we remember all these different versions. And then eventually the market settles on really three critical pieces of software. PowerPoint, Excel and, you know, Word. The problem is, is we've been running our businesses on PowerPoint, Word and Excel for too long.

• 14:18Speaker 00

  Okay? And let me give you an example. I, I, you know, I'm running around Iraq and Afghanistan and I'm constantly scratching my head, wondering why we're fighting wars on PowerPoint?

• 14:31Speaker 00

  Why we're fighting wars on Microsoft email? And why we're fighting wars on Google Earth and, you know, Excel and Word? The medium of intelligence and operations, which is the combination of, you know, all of a nation's power ultimately boils down to a PowerPoint slide deck and an Excel spreadsheet and a Word document. And these things are slung all over the world as attachments on emails. And then they're saved in legions of Microsoft file folders. Just God, the depth of these things is insane. And then we, we put them on SharePoint and Confluence pages.

• 15:15Speaker 00

  And I, I just want you to imagine, like, we took the world's greatest innovations in how to run a business and we took it to war. I remember the day that I stepped on a battlefield in Afghanistan, uh, a, a month and a half after the towers came down. And I literally took my desktop computer off the USS Iwo Jima. No, it was the, the time, sorry. I was on the Iwo Jima, the next deployment, one of the greatest amphib ships, uh, ever, LHD five, a great amphib. And I took my desktop computer out of my workspace on that ship and took it to Afghanistan.

• 15:57Speaker 00

  And I proceeded to do my mission as an intelligence officer, primary intelligence officer with 3,500 Marines, using Word, PowerPoint, Excel and email. Now that's probably fine for 2001. But it was a fine in 2010. Is it fine in 2015? How about 2020? How about 2026? It is not fine to move little chicklets around a PowerPoint slide. Now I want to offer the same thing to the business world. Is it fine to do your asset management work in Excel? And then bury all of that inside in an Excel spreadsheet and file it away in a Microsoft file folder system?

• 16:41Speaker 00

  And put it on SharePoint? Because when you put your data inside of these Microsoft products, you're basically putting a tombstone on it. Because that's where data goes to die. You're never going to retrieve that information unless you can figure out where the hell you saved it and then retrieve it.

• 16:58Speaker 00

  So what we did in the government, and, and this is with Project Maven and even before, when we started putting Palantir across the battlefield in 2011 in Helmand Province, first to, you know, uh, Camp Leatherneck, and then it's expanded from there. Is we said, we are tired of fighting wars in Microsoft Office. And it wasn't that we had anything against Microsoft. It's just, it's not a dynamic, living, breathing place that you can retrieve information. You know, we were losing three to six Marines a week, right? Tens of millions of dollars of capital equipment, you know, hundred or so injured, grievously many. And oftentimes it was because we were fighting in these very rote, you know, uh, you know, Microsoft, uh, products.

• 17:43Speaker 00

  And we were saying like, look, why? We need to be able to retrieve this information, we'd be able to fight better. And that started the campaign on advanced analytics. And then eventually what you understand is Maven today, which is a battlefield command and control system. Now, what are we doing in the business world? So I land at JP Morgan and all I see is email, Excel spreadsheets, PowerPoint slides, PDFs, and all of the brain power of an institution, of an entire firm reduced to PowerPoint slides and Excel spreadsheets and Word documents. And buried in just layers deep of file folders across it.

• 18:22Speaker 00

  And now we're starting to turn AI on, and this is in the 2022 timeframe. No, no one will forget November 30th, 2022, when ChatGPT 3.5 emerges on the scene and all of a sudden this stuff is just stunning. It's a day to remember in computer science. And, you know, obviously, over the progressive three years, it gets better and better and better. And then just more recently with uh 4.7, Claude Opus 4.7 and ChatGPT 5.5. Of course, the same thing with Gemini. These models are getting so powerful that the reasoning capabilities of things are insane. And, you know, they could crawl around your 3,000 Excel spreadsheets and try to figure out things.

• 19:07Speaker 00

  But like, why are we still in Excel? We need to be on platforms where all of our data is there. And our AIs can run across it and create new layers of data that are coming from these long-running inference layers. That is our message. Corporate America, the East is coming for us. They didn't have these 30 years of Microsoft Office. They're going straight to, you know, platforms, they're landing their data. They're not living in Excel spreadsheets. They're landing in platforms where all their data is there, both tabular and unstructured. And they're going to be able to reason across this far faster than we are. We've got to move away from the 1990s

• 19:52Speaker 00

  and start moving to an AI-native world where our data is essentially in a platform. And if you still need Excel, we'll have a little safety draft, a little, little button there for you. You can press it and it'll drop the data to Excel. But God save us if we're still doing that in two or three years. We need to be living in platforms and we need to be creating the correct data sets, both unstructured and structured in a platform with AI running all across the top of it.

• 20:18Speaker 01

  You know, I've spoken to a lot of companies that are addressing different parts of that. There's the, particularly with the Gentik now, there's the orchestration layer that's, that's managing different models and different agents and dividing problems into sub-problems and assigning them and all of that. Uh, and there's the observability layer where uh, you know, some system has to make sure that everything's working the way it's supposed to and they can detect anomalies and and uh, and automatically mitigate them.

• 21:02Speaker 01

  Uh, and, and then there's the, the data platform further down that because companies aren't necessarily centralizing their data. But there are these platforms that have access to all the various silos and, and create a unified uh, layer. So where is TW, TWG operating in that? And are you building solutions? Are you, are you uh, consulting uh, to, to, to help companies uh, decide on the right stack to solve these problems?

• 21:43Speaker 00

  Yes, all of that. So, look, my mission statement, and I carry this from, you know, seven years in the Pentagon and then three years at JP Morgan and now my year and a half here at TWG. I've been refining like what it is, you know, my life's purpose is. You know, obviously have a family, you know, worship my God, um, you know, help people. Those are obviously my most important. Um, but in terms of what I want to do in the world, right? And my number one goal is, you know, just like it was at the Pentagon, to get us off of PowerPoint and fight wars with this, this technology and not cheapen the value of software and appreciate its importance. Now, in this phase of my life, my number one goal is to help businesses

• 22:28Speaker 00

  because I fear what's coming from the East, frankly, dramatically. And this isn't hyperbole, where I want to help legacy companies that have amazing balance sheets, have fantastic intellectual property. And I want to help them basically emerge AI-native in a very defined period of time so that they can compete in what is going to be an insane, insane world with these AIs, you know, basically running across data and across, you know, essentially the entire market. And the reason why this is so critical is because, look, where we're getting, where we're getting with, with AI now is what I call the liquefaction of people's IP.

• 23:10Speaker 00

  So everyone's going to be on the best models, right? So we see right now, you know, Anthropic is really leading, right? They're, they have produced these fantastic models. You know, obviously, notwithstanding mythos and all the language around the cyber, you know, weapon thing, but just the performance of 4.7 and its reasoning abilities. And it's, and it's a long inference time that it can go and just run and access data and do all these things. And, and, and we're all just blown away by it. We're all spending all night with Claude Code. We're all, you know, you know, using the tools and we're just stunned by it. Well, everyone is going to be on these tools to include, to include your customers.

• 23:52Speaker 00

  And when you as a firm go and try to pitch an idea or whatever, your customer is already going to have that because we're all on the same AIs, right? And we're going to see this sort of, you know, everyone is now concentrated on two or three models using two or three major toolsets. And the only differentiator is our data, but even that's going to get liquefied over time where everyone is like, what, what's special about us? What I want to offer is, like, I can make your business special. I will make sure that your IP does not get liquefied because you're, you know, basically putting your data in platforms that the AI labs are providing you. Um, and they liquefy essentially your insights. I will, I will bring your data to these models in a secure way.

• 24:37Speaker 00

  I will protect your IP and I will help you emerge in, in, in a period of less than 36 months, AI-native, where we can take all the best that you are and work with your people and, and essentially start a new company. It's still your old company, but you emerge as a whole new entity. Now let me explain what that looks like. Let's say you're an investment bank. Let's say you're a regional bank. Let's say you are a sports team. I mean, you pick the business sector in our economy and you're probably scratching your head like, I got, I got great revenue, my balance sheet is awesome, I contribute to my community, but I am afraid of what the future of AI looks like specifically for my firm.

• 25:22Speaker 00

  Most people are projecting it's a chatbot because honestly, that's what everyone sees. You put a bunch of data in a window and a little dialogue box, the prompt goes out, the completion returns. You see pretty miraculous stuff. You're playing now with Claude Code, you're able to do some vibe coding, even spin up some complicated workflow work. And you're thinking that is what AI looks like for your firm and you're confused because you don't know how that rings the cash register. And where is the transformation for you? Well, actually, I'm here to tell you that that isn't the best window into what the future looks like for AI and an AI-native firm. What the future looks like is workflows, the things that matter to your business.

• 26:04Speaker 00

  I'm talking specifically about like how you bring money in the door through customer engagement, whether it's sales teams or people coming in. How do you do that? And how you reimagine that with AI? And when it looks differently than it does now, it'll have new forms of engagement, new ways to land customers and do client onboarding. So it's easier and cleaner and faster. And then once you land your new customers, what is the product that you're landing? What does the workflows look like that deliver the value-adding services and the differentiating capabilities for each of those customers? And those are workflows. And right now, a lot of those workflows are email, PowerPoint, Excel and a ton of phone calls. Well, you break those workflows down in business process reengineering and you build them back up.

• 26:49Speaker 00

  You still have the humans in there, but you've got a whole new way of doing things. And as every model changes, you insert that new model in there and it's giving you even more power and more ability to make really creative business solutions for your customers. And then the last piece of the business is the reconciliation work. How do I charge people? How do I get the money after I provided the service? All that backend, back office work. And each one of those has very specific workflows that humans spend countless hours, usually in Excel, usually in Word, usually in email, and we rebuild those. And at the end of the day, what happens is you've got this beautiful data layer like you were talking about.

• 27:30Speaker 00

  You've got AI use cases running along those workflows and you have an AI layer on top that is giving the CEO every morning when he wakes up at 4:30 before he jumps on his treadmill, how his business is doing. You missed this opportunity. These are your inflows and outflows. These clients left you. These are your new clients. And here's how it's going in the back office for, you know, invoicing. All of that just coming to you with lots of AI insight. It's breaking the business down, down to the studs and building it back up AI-native. And that's what we help companies do.

• 28:05Speaker 01

  Yeah, and, and you mentioned a few things there. Uh, but it sounds like the core of what you're doing is uh, is the data layer, is making sure this data is available to whatever agent or uh, model you have uh, that, that needs to access it. Is, is that the, the focus? And, and are you building a platform yourself? Or do you have a platform yourself? Or are you advising people on which platform to pick or how to build their own platform?

• 28:44Speaker 00

  Yeah, data layer and platform. So let's start with the data layer. Uh, I was a chief data officer at JP Morgan.

• 28:52Speaker 00

  And, um, look, I want to make sure people can clearly understand. This is not about data migration. Okay. Many of us have been through this where we counted every single week, how many petabytes we landed up in the cloud in some lake. And what we ended up creating was a bunch of swamps, right? And data has to be very specifically tuned for the things that it's needed for. So we always advocate for a use case in a workflow driven data project. Let me explain. Let's say the job is to land more clients. So it's a sales, uh, use case. What is the data you need

• 29:36Speaker 00

  to make the correct decisions on who to pitch and how to pitch? And how to stay engaged with that customer? And let's say there's 20 data sets that you need for that. 14 of them are third party data that you pay for. And the rest, the other six are a bunch of other data sets that are internal. That's your own internal bookkeeping and your pricing and all that stuff. We're only talking about that data. We're not talking about everything. We're talking about the data to make that successful. So we have to scope things. We have to be very careful that we're not boiling the ocean. We're just doing the data we need. Once we target that data, we then bring that data in through data integrations. We have it now available.

• 30:20Speaker 00

  We build the workflow and the AIs are running on top of it. But we have to be very careful on how we define data because we're not in a massive migration project here. We're just talking about the data to deliver business outcomes. The rest of the data is still out there. You know, I don't know what you've got running on back there and you can keep that data, but let's stay focused on business outcomes. Now let's talk about the platform.

• 30:45Speaker 00

  I use the world's greatest platform. And my engineers build on top of it. And it's called Palantir Foundry and Palantir AIP. I really have a bias in this space. I don't have time for other solutions. They don't work. They don't have the security. They don't have the technology built in. They're falling behind and it's not about a big warehouse of data. I mean, look, rows and columns are great. I can do that in Foundry. So I partnered and I have a joint venture with, uh, Palantir and Alex Karp's team, and I use the world's greatest platform. There's nothing, there's nothing better than that. In fact, I often deal with people in foreign countries that are like, you know, hey, um, we want to have our own system that's native to this country.

• 31:30Speaker 00

  And I'm like, are you building your own version of Microsoft Office? Like, why would, why would you take that on? Like, honestly, you'll never get there. It's going to take forever and you're going to fall short. This is the world's greatest platform that's built for this age. It's fantastic. And what we do is we take that platform that has all these services and capabilities. I land it within the four walls of the company that we're working with. And then my hundred plus engineers and data scientists, we build use cases and workflows on top of it. And we can deliver, you know, solutions in weeks. And final solutions, these are the production ready, general availability solutions in months. There's nothing better.

• 32:09Speaker 01

  I want to talk about those platforms, but, uh, uh, beyond data, observability is, I just had a conversation with someone who's talking about if you, if you, when something does go wrong, uh, if you can't quickly identify the root of the problem, you're gonna waste a lot of time and, and compound problems and et cetera. So what, what do you do about observability?

• 32:44Speaker 00

  Observability is baked into our solutions, just like it is at JP Morgan and every other place. You know, you basically look at your workflows and any business, um, that is regulated knows how to do this. But you take your workflows and you rate them. The most critical ones that you lose money on, that if they go down and start, and start malperforming to the ones that are back office and, you know, I don't know, maybe it's a, it's a simple system that, you know, if it goes down, we can get to it over time. You know, all business, uh, workflows are not critical. They're important, but they're not like critical, critical. And by critical, I mean it will garner regulatory attention.

• 33:29Speaker 00

  It will lose you serious money and it will lose your customers. On those critical workflows, you know, observability is baked in. It's baked in because there's humans in the loop that are ultimately responsible for yes, no, and this is going out the door. To, and even if it's straight through processing, there's still humans that are observing that. So there's humans there looking at it. To the, the automations that we build in, um, where we're constantly testing these to make sure they're landing. And then we build scaffolding around these workflows to make sure they don't malperform. Right? And so they're meant to be very accurate and very, you know, perform in this window. And then everything else around the SLAs,

• 34:10Speaker 00

  the service level agreements to make sure that it's performing, that it, that it, that it works as designed. You know, all of that is built in, honestly. And look, observability in AI is, is always been a conversation point. I know this, especially pre, uh, transformers. Where a lot of people are like, hey, these CNNs or these RNNs or these XG boosts, you know, these gradient boost models or these random force, whatever it might be. Are these things performing? Are they, are they starting to malperform? And by malperform, I mean the data that, that it absorbs and that, that it detects off of, or that it classifies sometimes starts to drift. And so we learned to put tools around the data as it was coming into the model

• 34:54Speaker 00

  to detect if data was drifting because models actually don't tend to drift. The data gets out of parameter. The model is still what it is and it loses its mind because it's like, this is not the data I'm used to seeing. I was told data was going to look like this. And suddenly the data starts moving, but the model will continue to perform and it'll, it'll get outside of its, its performance box. So what we want to make sure is that data isn't changing. And we put, we put sensors there to make sure that's not happening. If it does, then it alerts us and we've got to quickly revise the model. Now on the transformer work, this is all about hallucinations. And what we do is we build mountains of scaffolding because we don't want the generative nature of these models.

• 35:38Speaker 00

  We need deterministic results. In other words, we need a real result. We can't have this thing get too creative. In fact, we don't want any creativity. There are use cases where we want tons of creativity and we permit that within bounds. And we put filters and things like that on it. But in terms of getting the precision that we want, we build scaffolding and that's a lot of what our Palantir platform enables us to do.

• 36:01Speaker 01

  When you say the observability is baked in, it's baked into those platforms where it's, it's reading all the, because this is, you know, in a large system, it's more than a human can absorb.

• 36:16Speaker 00

  Yeah, in some cases we have AI, um, you know, as judge. So we'll have three AIs running on a single, uh, problem. And the, uh, convergence of them suggests that we're in agreement and where they disagree it alerts. In other cases, like on the straight through processing, you know, we have checks in there that is making sure everything is good. And we have little lights that say everything is good. Data's in parameter, everything is fine. But yes, uh, observability is baked, that's, that's, that's common practice. In fact, it's in the NIST, in the NIST AI risk management framework. If, if you remember, that's part of the, the principles that we all in industry agree to is ensuring that there's observability throughout the system.

• 37:01Speaker 01

  And so you guys, um, are, are implement, you're sort of system integrators in effect. Is, is that fair to say?

• 37:11Speaker 00

  I think we're more than that. Again, remember our mission statement. Our mission statement is to help fantastic legacy business. And I don't mean legacy in a diminished way. I mean, kick ass businesses have great balance sheet, great CEOs, great IP. And they're like, how do I get into this AI world? And honestly, let's be clear. The market is currently not well served by providers who can help businesses transition to AI native. We've got the AI labs that are pitching these enterprise solutions, but everyone's a little skittish on turning their data over to an AI lab. There's no zero day retention, uh, policies. No one's quite sure what they're doing.

• 37:55Speaker 00

  And as you know, they're racing, uh, cause they have these trillion dollar packages they're building with their data center builds. And so they're releasing models very quickly and people are concerned. I know there's a lot of attention paid to responsible work, especially from our friends at OpenAI and, um, Anthropic. They're very concerned to make sure they deliver. I'm not diminishing them in any way. But there's a natural tension with business leaders of turning their business over to an AI lab, especially when it's not clear the outcomes of the future. Is Anthropic going to win? Is Gemini going to win? Is, oh, wait a minute, maybe it's best for the market that no one win. And we have three or four, or even a handful more of magnificent, um, model builders

• 38:40Speaker 00

  who can do amazing work. And, and maybe I'd rather have a, just something in between to ensure this, you know, that my, my platform is always protected, that my IP doesn't get liquefied and its DNA spread out. Cause I want to always have a special solution and God forbid the day when my customer gets the same answer on Claude that I'm getting.

• 39:03Speaker 00

  And I'm looking at him in the face, trying to be value differentiating. They're like, dude, I just ran the same prompt and I got the same answer. And you're like, wait a minute, that's my proprietary information. And we're all kind of pissed off. You know what I mean? So, yes, we are more than a system integrator. What we are is we come in and we want to hand hold a business, like taking, you know, an older gentleman who's blind through the supermarket on a Saturday morning in New York City, where there's everybody's crushing in there. We want to hand hold this, this company through the aisles. So you don't get hit by the cart or the small child that's running around, or step on the cantaloupe that fell on the floor and

• 39:44Speaker 00

  slip and die, right? We want to hand hold you through the supermarket and get you to where you need to be and build AI native that fits your business, right? And then work with you to understand, like, let's redo this whole thing. Like, this particular workflow you never were happy with. And it's in Excel and Word and email and 50,000 phone calls. How do you want to take it down? And then we build it and they're like, you know what? I want to go further and farther. So we build it again. And they're like, nope, not there yet. That still looks like the same. I want to go for. And we do this with each business carefully and methodically. So we help with that. And then we bring in a whole compliance regime. Because you need data use councils and you need a little bit of bureaucracy to manage

• 40:28Speaker 00

  this. And you've got to have training videos to get everyone there. And we do this fast, by the way. This isn't a 10 year fricking project, right? This is months of work, fast, in quickly. And then we land Palantir, which is this magnificent spaceship, right? Like a spaceship from outer space. And it comes in and it just changes everyone. Everyone's so excited. And we do it within your four walls. So your data is protected. And then we have access to every single model. Everyone, the latest ones drop in, we build the scaffolding, we build those workflows, and we get to outcomes. So it's a piece of consulting, right? Because there's handholding at the supermarket in New York City, right? So you don't get run over by the cart. There is compliance work with data use councils and AI committees and

• 41:13Speaker 00

  all this stuff. And then there's the building, the engineering and the data science to build these things out. And then there's the human side, which is getting everyone trained and comfortable with this. And then the last piece is the executive oversight. You know, you don't spell AI CTO. You don't spell AI Chief Information Officer. AI is spelled CEO and business leaders. And so it's very important that we understand, and this is a big part of what we did at Maven. We didn't go to the Chief Information Officer and say, we're starting an AI project. We went to the warfighter, the commander, the Secretary of Defense. This

• 41:58Speaker 00

  wasn't going to get regulated down to the blinky lights people. Okay, this was going because this is a new form of labor. Look, our businesses and many businesses that moved carefully into AI native are going to quickly realize that the token spend is going to equal the cost of a, of a, of a human employee, of a staff member. And so when you burn these tokens, you better well do it well, because you're literally burning someone else's salary up. And so this is literally a new form of labor that's being birthed. This is not tech business. Tech implements. Got to have blinky lights going and everybody's got to function

• 42:40Speaker 00

  back there. But this is CEO, this is business leader business. That's how you spell AI. Bringing that whole constellation together is what we do.

• 42:49Speaker 01

  Is this something then that sits with the C suite? I mean, I've spoken to people about decision intelligence platforms that a CEO would work with on a daily basis in his office privately, so he can ask all the dumb questions and reason through a problem before he presents it to his employees or to the board or whatever. Is, is it's, are you, do you have that aspect when you talk about, you know, AI being spelled CEO? Is it,

• 43:24Speaker 01

  is it something that sits on the CEO's desk? And, and then just on your use of, of Palantir Foundry, are you're using that to build specific solutions for whatever company you're working with? Or, or are you using it kind of out of the back, out of the box?

• 43:51Speaker 00

  Yeah. All right. Let's address AI is spelled CEO. So absolutely. You know, I wasn't around, but I've only heard tell that this is sitting on the most senior leaders' desks within the Pentagon. Now, it's not a chatbot. Okay. It is the workflow. It is the world's most complex workflows. I'm talking about a command and control system that essentially commands all air, naval, ground power, space, cyber power, all reflective in an interface

• 44:37Speaker 00

  where AI is all throughout it. So this isn't about an AI in front of a CEO. This is about your business represented to you in the way your business operates. If you're a logistics management, it's probably a map. If it's a sports team, it's probably a whole different set of things. Concessions, how's it going? Am I getting my, my, my, my fans in and out of the stadium? So they're not sitting in traffic for an hour and a half. You know, how am I doing with recruiting, right? Do I got the, you know, what am I paying? You know, all this stuff, like all of the business operation. If I'm in asset management, it's inflows and outflows and landing new money and getting new clients and, you know, onboarding new securities. And how well is that going? And how well, if I'm in, if I'm a insurance

• 45:22Speaker 00

  company, how well are my underwriters doing? Am I losing money? Am I doing good? Am I making good decisions? How about my new products? You know, where am I at in terms of anything related to conversions? And how am I doing in marketing spend? And is I'm spending too much? Is it working? These are the reflection of a ton of AI and data science and building an interface. So the CEO can look at it and know their business in 10 seconds. And they can walk out of the room and they have a mental image. This is the stuff we learned in war. Experience and intuition all coming together so that you can show the CEO and they can just start using their experience and their intuition to make business decisions

• 46:03Speaker 00

  that are well informed. Now, you know, you asked the question, you know, you know, is this is CEO business? And where is the C suite on this? Look, the most destructive thing that's happening in the business right now is the appointment of AI officers. I have never seen anything worse than an AI officer who comes into the firm and well intentioned and great human beings, I'm certain, with great families and American flags outside their house. I'm sure these are some of the greatest human beings on the planet. This is a, a person that sits between the firm and bringing AI in. And this is a disaster. Okay, this person probably never has done anything in terms of fielding a real solution

• 46:49Speaker 00

  for a company ever. Look, the only qualifier that I have for, for a great engineer is have you ever deployed something for real? Tell me how it went and did it go into production? And was it used? That's the single most critical qualifier for a good engineer. And I would ask the same question of an AI officer. Tell me what you fielded. What worked? How'd it go? They don't have any stories because these guys are made up jobs. Okay. And so we're appointing these AI officers because the CEO wants to have something between him and his work. And these guys are literally blocking amazing work happening and they're picking their favorite things, or they have their own intuition. God knows. It's a disaster.

• 47:34Speaker 00

  You don't spell AI with an AI officer. You spell AI with CEO. Now, he can have a little staff because he doesn't have time for all this. He's got to fight his business. Okay. But he have a little staff, but he has got to take ownership of this. Or in three years, he's going to get smoked. Now, let's address the Palantir question. Yes, it's a platform. We land it.

• 47:58Speaker 00

  It's got many pro- many services. It's got a, you know, a whole data layer. It's called ontology, where you can bake in the logic of the business. And then on top of this, we build the workflows. The data lands inside of the ontology with the business logic, and we can literally turn use cases green and get them into GA in weeks. And then on top of that is the AIP layer, which allows us to build scaffolding around the AI so that these use cases and these workflows have AI across, uh, across them.

• 48:29Speaker 01

  You're emphasizing throughout this, the speed at which you guys work. And I've seen you talk about, you know, the competition from China. And of course, we all know about the AI race and certainly the military, uh, competition. But, uh, but, uh, you think it's important that enterprises throughout the economy adopt AI in order to, to keep ahead of China? I mean, can you talk about that? What, what are the risks? Uh, is it sort of economic dominance in the world market? Or what, what do you see as the risk if, if we don't, uh, adopt AI in the enterprise?

• 49:14Speaker 00

  Look, China is a mercantilistic, uh, model. They want to turn us into providers of soybeans and raw ingredients like rocks for cement and critical minerals or whatever. And they then want to churn that into a product and sell us, sell it back to us. This is the UK in the 1840s, right? In 1850s. They see themselves as the center of the world and we're just a supplier of raw materials and a consumer of their goods. I say bullshit. That's not, that's not the America I'm in. Okay. I'm certainly not going to be in their tailpipe, right? Breathing their exhaust. And we're in a real competition because that is what they want the world to become. They want to be the center and we're just supplying raw

• 49:59Speaker 00

  materials and buying their finished goods. And that is not right. Well, look, the only way we're going to prevent that from happening is by building up our own modes. And these can't be, you know, regulatory modes because that destroys, you know, the competitiveness of, of, of an economy. The European Union is going to do that. They're going to throw up all these things and it's going to be a disaster. We need to face them in a competitive way. But look, the way we face them is we, we look ourselves in the mirror and we realize we're falling behind. They're adopting AI much faster than we are because they're literally going AI native from the start. They're going to figure out AI native banking. They're going to figure out AI native payments. They're going to figure out AI native insurance. They're going to build products faster because we are complacent. We've got beautiful cash flow

• 50:44Speaker 00

  coming in. We feel quite comfortable where we're at. You know, we feel like with the regulatory framework, we are protected. And it's only a matter of time before they bring these things to market and people start asking themselves, why don't we have this? And then it's going to be too late. So the competition here is let's look ourselves in the mirror. Let's make those investments. Let's please avoid the interlocutor known as the AI officer. You don't spell AI CTO, you spell it CEO and get a plan together. Thirty six months, the AI is there. Build AI native, get your firm ready.

• 51:21Speaker 01

  Yeah. And on the China competition, I mean, certainly we see the innovation. I mean, I've spent at least half of my adult life in China, but China, its economy remains dominated by the state sector and the state sector is notoriously slow in, in moving. And all of this activity is taking place in the private sector. In a way, aren't we at an advantage to China in that regard?

• 51:51Speaker 00

  Look, I would challenge that on two or three points. Not in AI.

• 51:57Speaker 00

  Yes, they're a central command, uh, economy. They, they geared up for electricity. It's practically free. They have geared up for AI and they're pouring and giving huge regulatory, uh, birth, right? Like they, you've got all this room to operate with very little regulatory interference. Like they're in it to win it. And there, there's nothing that can be done to stop. They're going to get chips. They're going to, look, we are resting on our laurels right now. We are very comfortable that we make the best models and, you know, they're benchmarked and of course they're fantastic. They're doing their, their best they possibly can to reverse engineer these things through all kinds of nefarious techniques and means.

• 52:40Speaker 00

  And for now, we're in a good place, right? Obviously we've got great infrastructure. We've got, you know, we invented the data center, right? We, we've got great things. But look, they're going to be so focused on adoption and building this in, whether it's AI, um, robotics or AI, you know, for business setting, which is more on a desktop. Like they are in this to win. And I feel that we have many advantages. Okay. Clearly we do. Got great talent. It's a great country. This sort of open market approach really does generate amazing technology and thoughts. But I think the main point here is let's not get complacent. They're coming and we're going to get surprised if we're not ready.

• 53:24Speaker 01

  I don't know if you know Jeff Ding. Uh, I, I can't remember which university. I think he's in DC. Um, but, but he's written a lot about, uh, how the US is ahead on the innovation, uh, but China is ahead on the diffusion. He talks about how, you know, spreading it through the economy. But it, it seems to me that's, that's starting to happen pretty quickly here. Uh, I mean, the, you're, you're talking about, uh, taking enterprises and making them AI native in, I can't remember. Did you say 36 months?

• 54:04Speaker 00

  Thirty six months.

• 54:06Speaker 01

  But, um, those that don't are going to face AI native startups in their industries that are going to eat their lunch. I mean, I can see that, uh, in the consumer banking industry already with, uh, companies that are coming on stream. Uh, yeah. How, how do you look at that? Not so much competition with China, but a legacy companies if they don't make this transition quickly, having to face AI native startups in their industries?

• 54:43Speaker 00

  Yeah, that, I think that's what we're talking about here. I, I don't know what's worse, but I'll, I'll offer an opinion on this. An entire firm failing? Or a few people losing their jobs as there's a realignment as they move to AI native?

• 54:59Speaker 00

  The, the worst thing for this economy is we just see thousands of firms go under as they get blown away by AI native. Yeah. And look, I, I have been on the good side of asymmetry.

• 55:12Speaker 00

  Because I've been in the US armed forces where, where frankly, our nation and our taxpayers have made the commitment that we are never going to enter a battlefield unless we have the best technology in the world. And so I've been in a military that had a fifth generation aircraft, amazing cyber capabilities. Obviously, you know, our Maven smart system, fantastic rocket, missile, drone, all of it. And there was never even in our minds, the potential of failing. Now politically, there's a whole other matter there on how these things resolve. But like we have, I have always been from a second lieutenant to a colonel on the privileged

• 55:51Speaker 00

  side of having an asymmetric force of massive power. Now I want to tell you that I have never experienced what's like, what's, what it's like being on the other side. When I don't have that power. And I, all I know is there's a great video. Um, it's called Alpha Go. And it, I'm, I'm sure you remember this. This is something that we used in the early days of Maven to shock the department. But this is the classic Lee Sedong versus Alpha Go, Google AI in the game of Go. And the power of that isn't how the AI wins. That's cool. Right? We love that. What is powerful is watching Lee Sedol lose.

• 56:34Speaker 01

  Yeah.

• 56:35Speaker 00

  And what it's like to be on the other side of literally watching your business. In this case, the way he played Go and he was the world's best. And he suddenly realizes in head, I don't even know how to play this game anymore because this computer with this AI has just reinvented the game of Go and I am not a player anymore.

• 56:58Speaker 01

  Yeah.

• 56:58Speaker 00

  That is the place we don't ever want to be as a nation. And that's our call to arms here is get moving. So that doesn't happen to our great companies.

• 57:06Speaker 01

  Is there anything, Drew, that I haven't touched on that you'd like to talk about?

• 57:10Speaker 00

  Look, um, you know, many of us came out of the wars, the last, uh, 21, 22 years. And we're now, uh, bringing our experience back to the economy. You know, veterans have a lot to contribute. You know, we're not broken toys. Okay. Much maligned as we might be by generous benefactors who want to help us, you know, with any trauma that we had. We are very capable. And I would offer that some of the most capable veterans that can contribute back to the economy are those of us that, that saw, you know, conflict, saw the technology and can bring this to your firms. And we'd really like to, you know, double down on that. Look for your veterans out there. They have huge experiences.

• 57:53Speaker 00

  They've got a lot of background. They've got great technical experience. You know, your company would be well served by bringing these guys into your firms and, and letting them do what they do well, which is, you know, grow and build and help your firm prosper.

• 58:06Speaker 01

  Yeah, Drew, that's, it's been fascinating. I'd, I'd love to have you on again, just to talk through all of the, the different facets of, uh, Maven and, and algorithmic warfare and Ukraine and all that stuff. Uh, so if you'll indulge me, I'll, I'll try and get you again. Uh, but this is, this is, uh, it's been fascinating.