Accelerate Mission AI: From Data Strategy to Warfighting

Summary

Despite urgency around AI modernization and executive mandates, 75 percent of public sector organizations report they do not yet have the data readiness required for AI, making a trusted data foundation the prerequisite for mission success.
Leidos uses the Databricks Data and AI platform for its own business transformation and partners with Databricks to deliver those same capabilities to public sector customers worldwide, illustrating a shift from homogeneous to heterogeneous approaches toward open standards vertical integration.
Navy CTO Justin Finelli shares how the Department of the Navy is deploying AI to change warfighting outcomes, accelerating procurement, and building unfair advantages through data-driven decision-making.

Accelerate Mission AI: From Data Strategy to Warfighting

Mission success depends on trusted, scalable AI that operates at enterprise scale. this video showcases how organizations including the US Navy are moving AI from experimentation to production, achieving faster decisions and better outcomes. Learn how a unified data platform enables rapid deployment of advanced models grounded in organizational data, with strong governance and executive controls throughout.
Explore the evolution from homogeneous to heterogeneous approaches toward open standards vertical integration, where organizations get value from unified systems without vendor lock-in. Hear directly from Navy CTO Justin Finelli on real warfighting deployments, divestment strategies, and how to build unfair advantages through data-driven decision-making. Discover why the fastest organizations prioritize outcomes over processes and how they align with industry partners to accelerate capability delivery.
🤝

Chapters

FAQs

How is the US Navy using AI to change warfighting outcomes?

Navy CTO Justin Finelli describes real warfighting deployments where AI supports mission-critical operations and decision-making. The Navy is also pursuing divestment strategies to eliminate redundant legacy systems and building unfair advantages through data-driven approaches that accelerate capability delivery.

What is open standards vertical integration and why does it matter for government AI?

Open standards vertical integration is an approach that allows organizations to benefit from unified, integrated systems without locking themselves into a single vendor's proprietary stack. For government agencies, this means retaining flexibility to adopt new technologies while still achieving the operational efficiency of a tightly integrated platform.

Why do most public sector organizations struggle with AI readiness?

According to this video, 75 percent of public sector organizations do not believe they have the data readiness level required for AI. The root cause is typically fragmented data infrastructure and siloed systems rather than a lack of AI talent or tools.

How does Leidos partner with Databricks to deliver AI capabilities to government customers?

Leidos uses the Databricks Data and AI platform internally for its own business transformation and leverages that experience to deploy the same capabilities for public sector customers worldwide. This customer-and-partner model allows Leidos to bring tested, production-proven data and AI solutions to government agencies.

Full transcript

[00:09] Thank you guys all so much for being here today. Um, I'm super excited. Um, my as I was walking out, my son was carrying my four-year-old was carrying my eight-month-old and he was like, "I love him so much, mom. I want to like hug him and just pop his head off." And I was like, "Whoa, that's like very energetic." But then I got here and I
[00:25] saw everyone and I was like, "Oo, like I kind of get what he was feeling. I'm very excited. I'm not going to pop anyone's head off, but I am excited to be here. So, thank you. Um, so I'm Molly Just. I lead our global public sector go to market here at Data Bricks. Um, and again, I just really appreciate everybody coming. Um, so the last two
[00:42] sessions have been amazing. You've heard about why a clear data strategy is necessary and how critical it is to build a trusted data foundation. So, in this session, we're going to dive a little deeper into our real world AI use cases. You'll hear from the CTO of Lidos
[00:57] about how Lidos is using Data Bricks as a customer for business transformation as well as how Lidos and Data Bricks are partnering to deliver these same capabilities to public sector customers around the world. Um, finally, we're hoping that the Department of the Navy's CTO Justin Finelli will be here. I've
[01:14] heard he's on his way and so you'll hear from him on how the Navy is accelerating mission AI. So before we hear from our amazing guest speakers today, I want to take a few minutes and set the stage from like a high level, no pun intended. Um, you know, I I think right now you all are
[01:32] probably seeing this. Every organization, public sector, private sector, is realizing that they need to be a data and AI company and that data and AI are critical to achieving their actual mission outcomes. You know, I think AI isn't a buzzword anymore. Most public sector
[01:47] organizations are leveraging actively leveraging Genai and are actively investing in it. Um, and many are already seeing significant productivity improvements. At the same time, governments everywhere are leaning in. Leaders are pushing to
[02:03] modernize infrastructure and democratize data with mandates, executive orders focused on AI, modernization, data federation, and innovation. So, there's a real urgency to move forward. But none of that means it's going to be easy and I know this group knows that
[02:19] very very well. Um success isn't guaranteed and in fact still 75% of public sector organizations don't believe they have a data readiness level that's required for AI. So that's lack of data availability, lack of data accessibility, lack of data quality,
[02:37] data governance, or just the absence of a scalable data infrastructure. So it's not surprising then that public sector organizations still don't feel confident putting generative AI into production. So moving from those very fun cool experiments into production is
[02:52] super hard and I know you all know that the first time I saw this slide I literally said ah but the bottom line is here the difficulty of managing AI end to end is no joke at scale secure responsible management of AI is very
[03:07] difficult. You have your fragmented data, proprietary formats, disparate systems are all driving complexity and high costs. So, data bricks, it's designed to help simplify the complexity that we're all seeing. Managing the whole AI life cycle in one place, giving your teams the
[03:24] ability to bring data in, prepare it, manage models, and putting those models into use while keeping security and governance consistent throughout. It also gives the teams tools to monitor use, quality, and cost throughout the process. What public sector
[03:40] organizations want in AI is pretty simple. AI that they can trust, data that's ground uh AI that's grounded in their organizations data, strong governance, faster mission cycles, and the ability to leverage control so leaders feel comfortable scaling it.
[03:56] So now to hear more about how data bicks delivers on those AI requirements, I'd like to have Eric Moore from Lidos come up and talk about how Lidos is using Data Bricks to do just that. Thanks, Molly. Appreciate it. All right.
[04:15] It's uh it's great to be Justin's warm-up act. I didn't know that was on the agenda until just now. So that's great. Uh hopefully he's here and it'll be a a great session. Justin's awesome. So um so as Molly said I I want to talk a little bit about um our view at Lidos of
[04:34] how um the journey and some of the decisions and dilemmas that we have seen and I have personally seen customers and ourselves wrestle with for a long time when making decisions about how to move
[04:49] out uh in making investments around uh around data and analytics. Um so I want to talk a little bit about that dilemma, how I think that dilemma has evolved and really changed uh over the last six to 12 months, what that means uh and how we
[05:06] are responding to that both internally and in how we go to support our customers in the delivery for their missions. I'm really interested and you know hopefully there's some time afterwards to hear from you all as to whether this resonates with you this kind of this theory that I have of how
[05:22] uh things have changed over the last 6 to 12 months whether it it's something that you're seeing in your environment in your organization or with your customers whether you have a different opinion about it um it will be it will be interesting to to hear some feedback and have a conversation about it. So with that let's jump right in. Um so the
[05:41] dilemma that I have seen customers uh that I've been working with and within LIDOS we have waffled back and forth on for 10 20 years is this uh dilemma and idea of homogeneity versus heterogeneity
[05:57] right and you can in it think of this in lots of different domains. So you can think of it as a a network stack or a security stack or a cloud and data center hosting environment. Um, but obviously I'm going to talk about it in the context of a data and analytics ecosystem. And what I mean by this idea
[06:16] of heterogeneity and homogeneity and the dilemma between these two approaches is I've regularly seen customers and again within internally within lidos debate the value of one approach versus the other. Right? So on the homogeneity
[06:31] side, this is the concept that you you know fully adopt a vertically integrated ecosystem from a single vendor and you prioritize maximally extracting the value of the commercial investment in that ecosystem. And I've seen customers take that approach. We in Lidos at
[06:48] various points in time have taken that approach. On the heterogeneity side, the concept is that you're focused on reducing vendor lock in and choosing best-in-class bespoke solutions at individual point capabilities and
[07:03] stitching them together in a way that is very purpose-built for your mission or your particular uh business outcome. And what I learned over the last uh two decades uh helping customers navigate this and and working internally to navigate this is I've come to the
[07:19] conclusion that historically there actually isn't a wrong answer to this question. Um I used to think there there was there was a right approach and a wrong approach. But I think what's more has been more important is um understanding the tradeoffs that you're
[07:35] making when you choose one approach versus the other. being committed to the approach that you're taking and being really thoughtful about optimizing your implementation to maximize the pros of the approach that you're taking and minimizing or mitigating the the
[07:51] downsides of the approach that you're taking. So I think of this as an idea that there's like a local maxima of value on either end of this spectrum. And I think historically that is how I came to think of this problem and I would work with customers to say hey if
[08:06] you want to go one way or you want to go the other we can make either one work but we have to understand the trade-offs and we have to be committed to the approach we're taking. My opinion on this has changed. Um, and I believe now that this idea that there's a local maxima on either side,
[08:22] that you can find an equilibrium point that delivers value for you in the right way on either end of the spectrum is is not true anymore. Um, and I think now the reality is there's an absolute maxima that exists in between the two uh when it comes to data and analytics ecosystems. And I want to
[08:39] explain why I've come to that conclusion. So the the reason for that is that the pace of change that is occurring around technology and capability in the market over the last 6 to 12 months in this space has broken
[08:55] that that pre-existing approach and model. So why do I say that? Well, it's probably worth at first rewinding maybe 5 to 10 years. And I think Omar did this if for anybody that was in the previous session, he kind of did the same mental exercise. But but my thinking is if we
[09:10] rewind 5 to 10 years, many of us would probably be in a room like this at a conference like this and we'd be talking about all the transformation that's occurring and how things are dramatically improving and all the technology and capability that's being brought to market. The reality is with the hindsight of today's uh environment,
[09:28] I think we would all actually look back and say we were actually in a period of significant stability. um actually uh things were not transforming step function changes in capability at the pace at which maybe we felt like was true at the time. And so because of that
[09:44] stability, we actually were in a world where there was a significant degree of uh commoditization and uh par occurring within the ecosystem and the vendors and capabilities that were available. What that meant was that you could pick one of these two approaches and be really
[10:00] thoughtful about its implementation and be successful. So think about the homogeneity side, right? You could pick an ecosystem that you wanted to dive deep into. You wanted to be uh heavily invested in, you wanted to maximize value extraction from. And the reality is you were not likely to miss out on a
[10:17] tremendously transformative capability that was going to magically appear six months later and completely obviate the cho the choice you made of ecosystem to adopt. On the other side, you could make investments in building bespoke custom
[10:34] integration of these individual point solutions. You could make that investment with confidence that that individual point solution was still going to be relevant for two, three, five years. It may not be industryleading that whole time, but it was still going to be useful, relevant,
[10:49] and not completely obviated. So, you could amortize that investment out over a period of time, and you could get return on that investment. So you could you could land on either side and you could make it work. But the pace at which things are evolving now means that's just not true. And so we had to
[11:05] try to find a middle ground. We had to try to find a different approach both for ourselves and our customers that that allows us to be dynamic in the environment that we're all experiencing today. maximize our ability to pivot to be flexible and to deliver value at the
[11:23] um pace at which new capabilities are being delivered. And so we set out to think about well what are the core principles that would allow us to find that new approach. Um and I we have a bunch of them. I'm going to talk about three of them because I think they're
[11:38] the three most important. The first that I really want to highlight is this idea that we have to accept and embrace uh distributed data. We have to stop fighting against it and thinking that we can live in a world where we consolidate all of our data together and that that's how we're able to extract value and and
[11:56] analytics out of our data. We It's just a fallacy. It is not real at this point. the pace at which data is being produced and the places in which data is being produced continue to accelerate whether it's at the edge in the cloud on premise. It is not going to slow down
[12:11] and so we have to accept and embrace that fact. The second is that we want to focus on data products and not data platforms. So what do I mean by that? Well, what I mean is in a dynamic world where things are changing this fast, we
[12:26] want to focus our attention and energy on the things that have sustained value to the organization. That's the data. The data is the thing that has the intrinsic value to the organization. The platform is the tool to extract value from that data. And so we want to focus
[12:42] on data products, not on the data platforms. And then the third and it's a correlary to the second is we want to focus on portability of those data products. And the reason this is so important is because you can do all the right things focusing on data products
[12:59] and not the platform. But if you aren't thoughtful about using open standards and open frameworks to implement your data products with an eye towards portability, you will inevitably find yourself locked in to a platform. So you will have thought you were doing all the
[13:14] right things and still found yourself stuck in a world where you're not able to adapt uh in a way that you want to. And so uh distributed data data uh products not platforms and data product portability those are the core uh elements of our approach to this middle
[13:31] ground in this dynamic world. Um and so we set out to find partners that could help us implement that vision. Um and for for us at Lidos, nothing we do is small scale. Everything is enterprise
[13:47] scale or nation scale. And um nothing we do is green field. Every environment we are going to go into has pre-existing data, pre-existing uh systems that we have to adapt and uh absorb and otherwise transform. And so we needed a
[14:03] partner that understood how to operate in those environments. and we needed a partner that was aligned to our principles and understood and really embodied them. And that partner for us was data bicks. And so, um, I like to think of what we're doing with data
[14:19] bricks as a, um, open standards vertical integration, which may sound a little bit like an oxymoron, but what I mean by that is it's an approach for us to be able to get the value of that homogeneity system, that uh, vertically
[14:35] integrated system without the drawbacks of that system that enables us to have the uh, value creation of the heterogeneous system without the delays in the high cost of implementation of
[14:50] the bespoke integration that ultimately comes with that. And so that is what we've been able to achieve and ultimately been able to do with data bricks. Um and so that uh is not just a cultural philosophy. Really importantly, this is this is something that we are
[15:07] already delivering to customers and implementing internally, right? And what this allows us to achieve is something that has been really elusive. You know, Omar uh in the previous talk at one talked about this quite a bit about the idea of data silos. What this is
[15:22] allowing us to achieve is real true enterprise scale adoption. Um that is something that people talk about but has really really been hard to achieve when you think about what it really means to have enterprise scale adoption. That foundation of enterprise scale adoption
[15:38] allows us to accelerate the collection and production of intelligence and then ultimately tailor that intelligence specifically to an individual to an individual organization to an individual mission and thereby delivering outcomes at the pointy end of the spear where
[15:53] mission impact can be absolutely maximized. So uh we're really excited about what we've been able to achieve already. This is a new approach that we have have undertaken over the last 6 to 12 months that we think is purpose-built for the dynamic uh elements of today's
[16:09] uh environment. Um really really excited to talk to you all after the session about your thoughts and your experience in this how you're adapting to the pace of change that's occurring within the uh data ecosystem and AI ecosystem that we're all trying to maximize uh value
[16:26] from. But with that I'm going to hand it back over to Molly. Let her bring Justin up on stage and I look forward to talking to you all after that. Thanks. Thank you so much. So that was great. Uh, thank you so much, Eric. Uh, so, um, one thing I just
[16:44] wanted before we we bring Justin up for the fireside chat, I thought I would just kind of talk at a at a high level again, um, on our most common mission outcomes. Um, the real world mission AI across our customer base. So, you know, I kind of bucketed this in three three ways. The first one that we see our
[17:00] agencies wanting is how they modernize the way they deliver public services. So if you think about it, you know, think about agentic workflows for benefit eligibility determination or AI AI enabled health insights. The second way is we see these public sector
[17:16] organizations wanting to modernize the way they internally do business. So that could be anything from back office workflows, AI enabled resource management, real-time fraud monitoring. And third is we see our public sector
[17:31] customers wanting to deliver security and build trust to their constituents. So these AI enabled use cases aren't really futuristic anymore. Data bicks is already enabling these outcomes across both private and public sector customers. And something that we don't
[17:47] talk about enough that I think we should talk about a little bit more is that data bicks is bringing the same data intelligence platform that powers some of the world's most advanced commercial companies to the public sector. And you heard Rory chat uh talk about that today a little bit. Um so if you think about financial stability, you have large
[18:03] corporations like a JP Morgan Chase, a Capital One that's using our platform to process pabytes of market data to manage risk and detect fraud in milliseconds. And that same technology that allows the commercial sector to protect your credit card swipe in real time also enables the
[18:20] IRS to process your tax returns very fast and prevent tax fraud while also not slowing down a tax refund for a working family. So if you think about global resilience, companies like Petro Bros and AON Insurance use data bricks to model everything from environmental
[18:36] risk factors to industrial safety, predicting disasters and accidents before they happen. Again, those same tools are used at public sector organizations like FEMA and the Defense Logistics Agency where they can predict supply chain breaks, you know, before a
[18:52] storm hits. Also at the World Bank to process large amounts of economic data that drive policy decisions. Uh within the transportation industry, companies like Rivian and Joby are ingesting streaming telemetry from electric vehicles and aircraft to ensure
[19:07] safety and to optimize performance in real time. These are the same tools supporting mission AI at the Department of Transportation and the Department of Defense where they need to manage fleet readiness, predictive maintenance, and airspace safety. You can think about upskilling the workforce. So commercial
[19:23] companies like UDMI and Corsera use our data intelligence platform to personalize learning for millions and we apply that same data intelligence engine to the best universities around the world to personalize learning journeys as well as identify at risk students early. And finally, thinking about
[19:40] industrials and and manufacturing. Um, companies like John Deere and Skyideo use data bricks to power autonomous machines that can ingest pabytes of sensor data to train computer vision models. They might allow a tractor to identify what is a crop versus a weed or
[19:57] a UAV to map a 3D structure in real time. Again, these same capabilities are the same ones deployed at the USPS and Department of Transportation to manage their their fleet. So across all of these areas, the pattern's the same. Trusted, scalable AI accelerates
[20:14] missions by giving those teams fast insights that they can act on immediately. So hopefully by now you're starting to get a sense for how seriously data bricks takes delivering mission value to its customers. And in a minute you're going to hear directly from one of our favorite customers on how this his
[20:30] organization is accelerating mission operations. But before we do that, there's one thing that you can help us do to deliver even more mission value to the customers. And that's by helping us understand what mission outcomes you all are most interested in over the next year. This will help ensure that we're
[20:46] investing our resources in areas that are most important to you. So, if you can all scan this QR code that's surrounded by pictures of my children and Santa to make sure everyone is paying attention, I would very much appreciate if you can fill out the survey. Um, seriously, it will be
[21:02] really, really helpful. The data will be great. So, I'll just give you a moment to do that. And while we're doing that, without further ado, I would like to invite one
[21:18] of my favorite technology leaders up to the stage so you can hear from him directly on how the Navy's accelerating mission operations with AI. So, it's a privilege to host him. Um, I also promised him very loud clapping. So, if you could clap extra loud for him, that was part of the deal. Uh, please join me
[21:34] in welcoming Justin Finelli, the Navy CTO. Put your girl trying to
[21:51] carrot magic in the air. This is like very appropriate walk up music for Justin. I was very pleased. We can leave that in the background and kind of do this rhythmically. This is a hype group. I I do like that. Um, so
[22:08] thank you for being here. I know that you're very busy, so we really appreciate it. I sprinted and this group is worth it. Um, he absolutely did really sprint, so I want everyone to That is 100% true. So last night when we were talking about this, he made me promise the loud clapping, which you guys absolutely delivered on. And the second was hard
[22:24] questions. So hopefully I can deliver on that. I may have been dozing during that. I uh We'll see. So you're allowed to pass on any question, by the way. Um, I wanted to keep this snappy and moving quickly. Uh, so let me start. Justin, you have
[22:39] famously said that innovation adoption is a contact sport. In the Pentagon, that usually means endless meetings as we both know, but real contact sports leave bruises. Who or what process are you currently tackling the hardest? Good question. So um an interesting
[22:56] thing that's happening right now is I think there is um less contact uh it is more like uh uh special ops than it is uh contracted uh engagement
[23:12] right now. So the um I I think the what we're trying to attack right now is um the value per dollar. Uh and so like who's on the receiving end of that? Uh people who like meetings. Uh and so I
[23:27] literally had someone today, we said, "Hey, uh I I think we understand uh put together a proposal." And they said, "It sounds like you'd like to learn more of this about this." And I said, "That sounds like four more meetings. How about you give us a one-page lean business case on what we can do. We'll
[23:44] circle, we'll redline, and then we'll sign or not. And so like the status quo is what is uh like under siege at this moment. And I think we recognize collectively both in the government and outside. Um this can be power laws
[24:00] better. And a lot of the things that were stopping us before are still there, but they're either muted or we can run faster than them. Uh, and so like let's do that. I can talk later about a couple places where we've 10xed what was
[24:16] happening before both in terms of speed and outcomes. And so like just we need to shoot the hole. Okay, perfect. Um, so we talk a lot about mission AI at data bricks using data actually to do something not just report on it. When you look at the fleet today, be honest. Are we actually using
[24:32] AI to change mission outcomes or are we just writing better performance reports? Um, I love whoever just laughed. You are my people. Is there more that we can do certainly afloat? Yes. Uh and so what that most
[24:48] recently has looked like was um we had some uh we tackled some big winds ashore and then we compared them what was with what was happening afloat and and they said um like there are places where our data performance and AI was better
[25:04] ashore than a float. Like as a technologist, this makes sense. As uh someone who is supporting war fighters and uh and uh in a uh department that really really wants capability edge, it doesn't make sense. And so what we did
[25:20] there was we brought a couple uh chief systems officers um from carriers and other uh platforms forward and said um do you like this? And they said, "Yes, I would die for that on my ship." And we're like, well, here's what we have.
[25:35] Two of them took it back to their ship. It was on there uh in three months. In one case, someone tried to march it back off uh and then uh there was some uh potentially contact sporting uh and then it's so it's on all of the carriers now.
[25:52] Uh this is a case where um anything in that particular uh readiness um zone uh of operations is performing significantly better. We now have a lot of cases where the bandwidth on ships and edge as a whole allow us to
[26:09] do things that we couldn't do three years ago. And so what do outcomes look like? It literally looks like faster changes to the system. So we uh I teach a course called uh softwaredefined warfare with uh with a common friend of ours. Uh and and the culmination of that class was
[26:27] what if we could do changing configurations on the edge, changing the way that we make decisions overnight. There are places where we're doing that and there are so many more places where we could. It's just a matter of that sort of product market fit. Here's one
[26:42] thing that's different. Um we have sailors and marines who have ideas. We also have uh vendors and partners who have ideas. We now recognize um the requirements officers and uh people in my shop. We do not have all of the
[26:58] ideas. So what we need to get more outcomes at this moment is for people to be Nat's ass level of specific on their ideas and then we can kind of make it malleable from there. That's that's terrific. That's like I always try to joke on you, but then you say things like that and it's just
[27:14] incredibly impressive. Um, okay. That that leads me to this next question. What is a Gen AI use case that really excites you right now? So, um, I I want to answer this two ways. Uh, one is, uh, like based on genai.mill
[27:31] uh, coming through. Um, I I think we're going to, and so this is kind of obvious, but uh I think we're going to have that in the hands of more people in a streamlined way. And so just at the absolute baseline of this, um, the last
[27:47] speaker spoke about ROI. I love the use case that gives sailor back the most possible hours. And so we will absolutely rank them based on impact. um whatever one does the most for them is our favorite in
[28:04] terms of like and so like literally outcomes is how we're prioritizing what is most interesting to me well um there are some like sticky problems uh from an optimization perspective that we couldn't get after uh I spend a little bit more time with money uh than uh than
[28:21] I'd like to uh and so if we could tell the here is our buying power story within acquisition as we move to portfolio IOS. Uh I I think that would be a really interesting JI case to say, hey, here are our 12,000 systems.
[28:36] Yeah. Where is our best case for um doing that hunting, doing that contact sporting, uh and then scaling what really really works. That makes sense. Um feel free to pass on this, but everybody loves talking
[28:52] about innovation and adding new shiny toys. Nobody likes talking about divestment with maybe the exception of you sometimes. What is your favorite zombie system that you're trying to kill right now? Um, so we have uh a lot of data environments.
[29:07] We have a lot of cloud instances. Um, at one point one of the hyperscalers uh told us they didn't know how many cloud instances we had on their cloud. That's how many. Um, and so this is suburban
[29:23] sprawl that we just want to get in front of. Um I I I don't think anyone designed uh the city that way. Um but reducing the number of data environments without reducing capability will improve value.
[29:38] Yeah. Um and then similarly cloud instances like that is the blueprint. We've already done some of it. Um but like it's uh it's the best kind of like moment for if you're hiding data at the
[29:55] nippernet or uncclass level and someone can say hey we can do this more efficiently. We just like that here's the same problem. Let's get tactical for a second. The same problem is always we have an overlap of data but there are 2% of the system is transactional and uh
[30:11] nothing else does the transactions and so we can't move them and that has left 10,000 systems in their place for the longest time. Uh where we can consolidate that into a platform where we can rank order it based on ease of migration and return on investment. We
[30:28] want to do that divest invest. And so what's different now is we're doing a little bit more judo where it's not this team of government people who are trying to rack and stack and we'll get back to you in a little while. If companies are coming to us to say, "Hey, this one Dart
[30:43] takes out these five data environments," we're we're saying go. Oh, wow. Terrific. All right. If we gave a data bricks notebook and full data access to a petty officer on a destroyer today, what do you think they would build that a
[30:58] contractor never would? Um the uh so we know this about cyber. Uh I I think we um we can sometimes forget it uh as it relates to kind of everybody else which is um the unsexy
[31:15] stuff is how we spend too much of our time. Um I have seen a lot of military folks who are um like literally uh chomping at the bit to serve and protect all of us and we can weigh them down with overhead. And so I I would say that
[31:33] like one of the first uh last time I was on a cruiser and then um like almost every time I'm on an amphib they're like can we not do this thing anymore? And so um that that's maintenance functions. Um and so we have probably a few like
[31:51] decent um maintenance functions that are that'll work disconnected. Uh but if we could have something there that uh allows us to do more predictive on the supply chain side uh I I think that goes pretty differently. Uh and then similarly um hey any task that these
[32:09] guys are doing on a regular basis show here is all of the ways to do it. Here are the highest impact ways to do it and then take it down. So I think they just want more time back so that they can serve better sleep a little bit more uh and then just overall have situational
[32:26] awareness that we have the data for right now. Um, we need closers to help on the back end for that to be uh like kind of the the norm as opposed to um the heroics. That makes sense. Uh, we often talk about leveling the playing field. I
[32:42] don't really want a level playing field. I would like an unfair advantage for the US Navy. Where does data and AI give us that unfair advantage right now? It's a room full of the lights are bright, so I can't totally tell how many nerds uh are here. Uh I I am one. Uh
[33:00] I saw thumb some thumbs up. I feel like um the this is the coolest thing in the world. If we build a huge lead and increase deterrence, all of us are exceptionally
[33:16] valuable to national and world security, right? And so like my money uh often goes to um hey uh here are ways that we can improve our budget in general. Um
[33:32] the like I I think of the the purse as like a pretty powerful mechanism but um our decision- making uh the way that we're breaking this up in general is enterprise. So uh the back-end activities uh usually thought of as like
[33:49] I mentioned not sexy actually do improve our readiness, improve our supply, improve how we do everything that feeds more than it ever did before the war fighting and the intelligence. And so as
[34:04] we clean up enterprise and use that to inform warf fighting and as intelligence informs war fighting and potentially as enterprise in informs war fighting. So that's my answer. We can do all of this in a more integrated fashion.
[34:19] That's a great answer. Uh all right. So I feel like people right now they're talking about kind of lengthy procurement cycles. Um, and what I think there's a large narrative out there that like we need two weak threat cycles. When you look at your road map, what is
[34:35] the one thing keeping you up at night because it's moving at the speed of molasses versus the speed of relevance?
[34:51] so within acquisition, uh, we we have a workforce who is learning to live in this new world. Uh and I think what we can collectively do about that if the feedback I often get is people were trained in the federal acquisition regulations and so some of us have uh
[35:09] other transactional agreements and that doesn't work the same way. So two answers. One is um what keeps me up is that uh contracts and performance contracts still take too long because they're just um there
[35:25] aren't enough people doing that as progressively as possible. And then I want to answer that. So like the solution that I've seen is um if folks from the private sector come with a more baked solution just like it's not fair,
[35:40] right? Like we want an unfair advantage. Well, it takes unfair action. And so, uh, like it it's just not enough to meet us 50/50 because there are so many people that I'm looking at who are like very very capable. And so, getting a deal to the 20 yard line still means
[35:57] five more meetings. If people can bring deals to the one yard line to say, "Hey, um, here is a way that we can make decisions faster and get rid of systems and all I have to do is have this data
[36:12] shared with me and you have to write the contract this way." We are so much more receptive to that than any time in my 27 years. And so what I'm saying is if there can be a silver platter activity
[36:28] um that that the private sector leads, we'll still like again we can we can redline it, we can edit it, but like I just think that you guys are more creative than us a lot of times and and so that's what keeps me up at night is people are giving us something at like
[36:44] the seven yard line uh and I'm like ah this looks so good but there are 10 things at the seven yard line and it's 11:00 p.m. and I can do one more and then text you back or we could do three deals if they were at the one yard line.
[37:00] And so just being able to close that cost and value gap quicker. Um that's that's what I see as a as a bottleneck. Yeah. Um our chief AI scientist, Jonathan Frankle, asked me to ask you this question. What is your strategy
[37:16] when it comes to integrating AI in the Navy? as a whole. That's what he said. He would love to hear that. All right, in two minutes because I do have a fun like rapid iteration. Yeah, great. Uh so first one is um who thinks we're great at data sharing?
[37:35] All right. Um so uh a couple pieces to this strategy. Uh number one, name enterprise services. Uh so we have uh 50 uh business BMA business uh management area data environments. Uh if we have a
[37:51] uh overall uh here is the enterprise service that we're using. Here is the clock that transparently we are showing people either migrated or they have trouble migrating and then we're going to help them with carrots or sticks. That's number one enterprise services
[38:07] and transparency visibility. Number two is any policy impediments that we have data sharing or otherwise we want more light of day on that. And so we've drawn up some new policy changes. Um number
[38:22] three is we measure outcomes and that is less um like hey here's the system or here's the process. It's that forces us to fix the big problems and that's a little bit more um on the backs of a few
[38:40] people who get to escalate things really quickly. But we have shifted to talking about outcomes. we've kind of like prototyped. We're ready to scale that. And so I think that's the one, two, three on on how we do this differently. And then repeat.
[38:55] That's a good one. All right. The next few questions are kind of rapid iteration. Um tech debt or culture debt. Uh everyone's probably seen the innovation diffusion curve, right? Um we don't need a 100% of people uh to be uh
[39:12] like rowing harder than they've ever rode before. Uh we've had the 3% for a while. Uh I think we're in that 12 to 15% private and public uh who can bring big wins. If we scale that across the
[39:27] whole left side of the bell curve, that is enough to fix both of them. Answer. All right. You teach graduate students at Georgetown. What's one thing that a 22-year-old student understands about the future of tech that a senior government leader might miss?
[39:49] um how dynamic it is on like a weekly basis. So uh so from a uh like you have to pay attention. Uh I see um people who are learning new
[40:06] tricks faster than they were before. Uh, and so the way that I think about this right now is, uh, it's not about Molly or me, it's about me plus AI and what results are you getting from that. And
[40:22] so I think there's like an increased recognition um, in the young people. I need a team of three people. That's actually more effective than a team of 15 people. There are still folks in the government who are like, I think we need 30 more people. Brooks law is like 40 years old. more people um does not solve
[40:40] that many problems anymore. And so uh I think it's really well understood in industry right now the uh the revenue per employee. I think if we can get that down or if industry through wins can teach us to do the value per government
[40:57] employee or the value per dollar on a contract then we're we're absolutely winning. Okay, that leads into this question. Um, we're in a room full of industry. If Can you finish this sentence? If you want to work with the Navy on AI, don't bring us
[41:13] a slide deck. Bring us a lean business case that shows um either a divestment opportunity. This isn't fair because we understand our business. You might not. Um, please just help figure out our business. show us
[41:29] where we can do a divestment and then show or estimate the baseline and show how much better you are than the baseline. Um, and uh if there are wins you've already gotten like that, we're allowed to brag for you. Here's the
[41:46] deal. It's in everyone's best interest for this to be the golden era of public private with enough data to tell win stories. We know from relationships that you need to hear good news a lot more than you
[42:03] need to hear bad news because it sticks. Most historically most government acquisition or delivery stories are like lean towards bad. We need to blow it out of the water. We need quantitative success stories. So if you have those
[42:18] 2159179497 uh the the idea of tallying up the number of success stories above what we've done before, let's tell those three to 12 times more than any bad news stories. We can do this together right
[42:35] now and then we never have to go back to waterfall operations. I love that. All right, last question. If your job was a Netflix series, what would it be called and who would play you? That's uh that's going to require a little bit of thinking. That was a tough one. So, uh on my part,
[42:52] uh I see this picture of me without facial hair. Uh and uh and I I'm I'm channeling like I'd like to watch Steve Carell play me. Hopefully, he's not like too mean. Um uh but uh let's see.
[43:10] I'm going to run down the clock and like actually let this flow. um adapt and adopt. Uh and this is uh really the story of um letting great
[43:25] Americans at great companies help reinvent the way that the government operates as a whole. Terrific. Everyone, Justin Finelli, thank you.

Learn more about the Databricks Data and AI platform.

The information provided herein is for general informational purposes only and may not reflect the most current product capabilities or configurations.