Quad Charts be Damned: Data Meets the Mission

For years now, the question for federal agencies has been the same: is your data ready for AI? In Episode 9 of The GIST of Govt IT, Brian and Sean sit down with Andrew Churchill, who leads Qlik's Public Sector Business across the US and Canada, to dig into whether the answer is finally shifting from "not yet" to "we're getting there" — and what's actually driving it. Andrew shares how policy moves are pushing the idea that data belongs to the mission, not the system owner; how AI is automating the mundane data prep work; and why the trust score on the data behind an AI recommendation is becoming the single most important factor for senior leaders making decisions. Brian, Sean, and Andrew unpack the agentic identity challenge nobody's talking about, the IBM Think team that burned 20% of its annual Mythos token budget in a single weekend, why a federal employee who knows the mission plus AI beats a forward-deployed engineer every time. Why federal data progress is a war of inches and why leaders need to turn the people delivering small wins into heroes — quad charts be damned. Plus, live music recommendations for the DC region.

----------

RESOURCES MENTIONED IN THIS EPISODE

Featured Guest
- Andrew Churchill, VP Public Sector, Qlik
- Qlik Public Sector
- Qlik Data Literacy Program (free resources)
- Qlik FedRAMP authorization status

Hegseth's Advana Memorandum
- Advana restructuring memo and program overhaul (Jan 2026)
- Hegseth's "Transforming the Warfighting Acquisition System" memo (Nov 7, 2025)

DoD AI/Data Programs Referenced
- Advana (DoD enterprise data and analytics platform)
- Project Maven (DoD computer vision/AI)
- Project Bravo (Air Force innovation initiative)

Federal Data Policy
- DoD Data Strategy
- Federal Data Strategy (strategy.data.gov)
- Data.gov

Live Music Featured
- Tedeschi Trucks Band - Bound for Glory 
- Tedeschi Trucks Band Live from Red Rocks - Full Concert
- Point Break Music Festival
- Warped Tour DC

References & Concepts
- Gartner BI/Analytics Magic Quadrant
- Stuart Wagner (Navy, former Air Force CTO)
- Lori Mangold (Booz Allen, former Army Chief G3/5/7)

Previous Events
- Transforming Data Into a Strategic National Asset w/ NASA CDO
- Qlik Public Sector Summit

Life's a Game of Inches

- "Al Pacino" Epic Monologue from Any Given Sunday 

The Hosts & Show
- Swish 
- GIST 360

Come find Brian, Andrew, and Sean at a show this summer. Quad charts not required!

CONNECT WITH US

Got an idea for a future episode? Want to be a guest? Let us know.

Brian Lake - blake@swishdata.com

Sean Applegate - sapplegate@swishdata.com

Subscribe wherever you get your podcasts: Apple Podcasts, Spotify, or gist360.com.

Transcript

Cold Open The Federal Data Dam

Brian Lake 0:00

Ever since AI exploded onto the scene, every federal agency has been asking the same question. Is my data that is needed to power this transformation truly ready for this moment? And the answer for most has been well not quite. But something is finally shifting. New policy is breaking down data and information silos that have stood for decades. AI is doing the mundane prep work that used to take armies of people. And federal employees who know their mission's cold, paired with the right tools and data, can run circles around forward-deployed engineers and platforms that have been sold as panaceas. So to find out if the dam is finally breaking and discuss the war of inches that's reshaping federal data, you know what we have to do.

SPEAKER_02 0:45

Let's get down to the gist of it.

Meet Andrew Churchill From Qlik

Brian Lake 1:15

Hey Sean, how are you doing today, buddy? Good morning, Brian. Well, Sean, today I'm really excited to be here in Washington, D.C., back in the studio. We've got a great guest with us today, Mr. Andrew Churchill from Clik. Uh Andrew, why don't you tell everybody who you are, what you do, why you're here, and uh what we're gonna be talking about today. Yeah.

SPEAKER_02 1:32

So uh I'm a I'm a Fed IT lifer, uh only job I've ever had. I've been at Click about 11 years and spent my entire career in the data space. I lead our public sector business across U.S. and Canada, and in that capacity, I have the opportunity to walk work with 100 plus customers and some really interesting data analytics and AI missions. It's uh uh it's uh something that really helps me get out of bed every day, enjoy what we do. Uh very exciting.

Brian Lake 1:59

Aaron Powell So why don't you tell the folks before we dig into the show a little bit about what what Qlik does and what's your primary mission and what you're doing for federal agencies.

SPEAKER_02 2:07

Aaron Powell Yeah. So I mean, Click's been around 35 years. I mean, we're I think it was our 16th year in the upper right quadrant of uh you know the Gartner uh BI. Trevor Burrus, Jr. It's a good place to be. Yeah. Any place else is the wrong place to be. But you know, we compete against big guys like Salesforce Tableau and Microsoft. And you know, we kind of separate ourselves as the built-for-the-enterprise scale uh type of solution, and we've earned that spot uh in in the federal government. You know, we came out of the gates really as one of the BI 2.0 vendors. How do I turn business people, the analysts, into the developers of those information products? And through acquisition, we've been owned by Toma Bravo for you know since uh 2018. Um, we've just you know grown through acquisition, expanding our portfolio to really get to the data acquisition side. So how do I move data to where it needs to be, to Databricks, to Snowflake, to AWS, to Azure? Uh and then how do I improve data? Uh you know, recently, obviously, like everybody else, our big focus is on how do I turn analytics capabilities into AI capabilities in this agentic moment. And you know, that's sort of where we are. We're uh aisle four on our way to aisle five. We're FedRAMP moderate, and our customers are adopting that and beginning to make that journey.

Brian Lake 3:24

Aaron Powell

Summer Concert Season!

Brian Lake 3:25

Excellent. Well, I know that uh we seem to have two types of guests on this show: either folks that have uh hacker names or live music lovers. And so I know that you and I were at the Fed 100 a couple weeks back, and we kind of discovered that if we had the Venn diagram, there's a lot of in the middle kind of similarities between live music. So it's summertime. We're obviously sitting here, we've all just recovered from sweating, trying to get to the studio. So what do you got on your uh summer music schedule?

SPEAKER_02 3:49

So I think the one I'm most excited about isn't necessarily my choice, it's my daughter's. So my daughter's 21. You know, she's kind of come around to listening to a lot of the music that uh that I like, but she identified this point break festival in Virginia Beach at the end of the month, which is big reggae festival with a whole bunch of different bands. I'm just really excited to get down there, enjoy some live music with her. Uh beyond that, Tadeshi Truck Band uh at uh at Wolf Trap at the end of August. Love that big band experience. I think they just tear it up.

Brian Lake 4:22

That's such a great band because when uh Derek Trucks and Susan Zedesky got married, I don't know if people know this, they both had separate bands, right? They got married, and instead of saying, Well, we're gonna take my guitarist and your bass player, they just took both bands and put them together. So there's like 13, 14 people on stage, right?

Sean Applegate 4:42

So Hey, speaking of big concerts, warp tours this weekend in D.C.

Brian Lake 4:46

Oh, now we're calling we're going way back here. Okay.

Sean Applegate 4:48

My daughter's gonna be at that one, Andrew. So uh I won't be there.

Brian Lake 4:51

I'm gonna be able to do that. So this is one of those like whatever's old is new again kind of thing, right? So excellent. Well, listen, I know, and we'll we'll make sure, folks, we put some really choice uh tracks in the show notes for you guys to check out. So

Policy And AI Speed Data Prep

Brian Lake 5:03

uh let's dig into what we want to talk about here, which is really about data, right? Um we've been asking the same question for about three years, four years now. Is the data, is your data, your agency's data, ready for AI? Um every round table, every webinar, it seems to be so that this is the core topic we get back to when we're talking about agentic AI and AI as a whole. So let's really start there. Uh Andrew, are we any closer to an answer? Is the question itself broken?

SPEAKER_02 5:36

Uh so we are absolutely getting closer to a state of readiness. Uh and you know, to put some perspective on it, I mean, we've been getting data ready for something for the better part of the last 50 years. Probably, you know, I've been selling software to get data ready for analytics and reporting and data warehousing for 25 years, and the ball wasn't moving very much. So here we are, we arrive at this AI moment, and you know, the I think the thing that's shifting is one, policy. You know, so I think we you know, you take Hegseth's memorandum, hey, the data belongs to the mission, and and uh Kat Hicks had something similar a few years back. But AI is coming to the rescue of AI. The fact is that the hardest work that was preventing data from being made ready uh was because it was a very laborious, mundane type of task. And AI is taking on a lot of that. So we're automating vendors like Click and many of our peers that are operating in these niche spaces are automating so much of that hard work that we're going to, you know, between that policy and between the people being able to provide 10 minutes of time to be able to accommodate the need of that of that uh task, I think we're we're moving the ball.

Sean Applegate 6:53

Yeah, I think one of the things we've definitely seen, for example, is the ability for AI to do simple things like data labeling, where you'd have to throw hundreds of people at tens of thousands of images for computer vision or looking at you know, taking data in and try to normalize it, wrangle it, figure it out, clean it up. The AI is very powerful, saving a significant amount of time there when you think of real people time and resources from a cost perspective. Aaron Powell Yeah.

SPEAKER_02 7:16

And I may a lot of people on my team thought I was the biggest nerd in the world when we introduced this metadata tagging feature where you could go in and look at a piece of data in the glossary and say, you know, tell me what this is. And it went out and it you know through AI, brought back the answer, recommended the solution to you. And you're like, I'm like, that's game-changing. They're all that's boring. I'm like, you have no idea. You just wait and see.

Sean Applegate 7:39

This is going to move the needle. Well, the other part too, I think that's important is the ability to trust the data. We may get into that further later, but having a trust score in your data catalog or in your system. So not only do you say the data is clean, but you know, what's the lineage? Do I trust it? What's the quality of the data? Really important when you get into the legal side in the government of you know, is it you know what's the source of the data? Is it able to drive a good decision? Because if if you get AI, you need responsible and ethical AI. And that really does start with having great data that not just the mission can can trust, but people like the lawyers or the uh the OIG can really support and back up and get behind. And a lot of the CTOs I've spoken with say they spend as much time with lawyers in some cases as they do with engineers writing code. Yeah, for sure.

SPEAKER_02 8:23

And I mean, like if you look at the way that senior leaders make decisions right now, they have a they have a group that they trust, advisors that are recommending what decision they should make. And you know, data is more and more becoming one of those advisors. So how do I, you know, how do I develop a trusted relationship with that advisor? Well, I need to be able to understand what's coming behind it. And I think this is one of the biggest obstacles to scaling AI is going to be do I, as a decision maker, believe that this recommendation coming from the system is something I should trust? You know, and that begins with the data that is driving that recommendation.

Brian Lake 9:02

Yeah, I mean, I think the question that I that I also think through a lot, we heard this on a roundtable we did with your team, is just there's so many different data sources. I mean, if you look at Department of Energy and these labs, I mean, going, they're going back decades of hundreds, if not thousands, of different types of data that you need to then try to aggregate and bring together. How do you think about for federal agencies? I mean, is is AI the great equalizer to finally extract that information, that value to be that trusted source across all of these different types of data? Um is that where they're going to be able to drive that ability to actually finally trust your own data? Yeah.

SPEAKER_02 9:44

Well, I mean, so again, it begins with people and policy. You know, the biggest obstacle in the government is people, policy, and politics. Uh as we see those things move away because there's more directives that come down that say this data belongs to the mission and not the system owner, it peels away that barrier. And that's where the opportunity, you know, with you know when you look at Databricks and Snowflake and you know, everything native in AWS, the potential there for whether it's physical aggregation or zero copy kind of representation of that data, it it just melts it away. And at that point, you then are dealing with, you know, so now you've taken care of the technical access issues. Now you're dealing with the, do I trust this, do I, do I understand the relationships between this data to be able to represent that in an AI decision, an analytics dashboard, whatever it might

Trust Scores Lineage And Zero Trust

SPEAKER_02 10:36

be.

Brian Lake 10:36

Aaron Powell Sean, are you seeing uh conflict between owners of the data from either the C-suite to the mission owners? I mean, we used-we used to we talked about this years ago, how chief data officers had no budget authority, even sometimes respect in agencies, and now these folks have the some of the largest budgets now in federal agencies.

Sean Applegate 10:54

I think we've made a lot of progress. It's not perfect, but I think most CAIOs or CDAOs have the secretary-level coverage and now executive orders that simply say, hey, data is important, it's the government's data, let's all share it. I would say, you know, three to five years ago that wasn't necessarily the case, and somebody would have to walk around with a letter they'd keep in their coat pocket. So when somebody said it's my data and you can't have it, they'd whip it out and say, well, the secretary of this department says it's my data and not yours. You know, would you like to play ball with me, or do I need to go to the top? And um, yeah, normally you show up with a velvet hammer and hopefully you don't have to whip it out, right? People want to work with you instead of maybe push back. But uh again, I and we've done some surveys across the different verticals, right, Brian, in the past. I mean, we see statistical differences in places that are maybe a little more generative and collaborative, where that's maybe a little more open and easy, and maybe not even as sensitive. And then you have places that are maybe a little more bureaucratic or risk-adverse cultures, and that becomes a little trickier there. Although I think, again, I think we've made a ton of progress. The the bigger challenges I think we face are around funding and building momentum or or you know, upskilling employees so they can kind of row together in a collaborative way that's uh collectively coordinated versus maybe very bespoke and scattered, meaning you gotta herd the cats to get things built where there's a lot of departmental level AI use cases being done, many cases hundreds of these in agencies, but not every agency has a mature platform and strategy that's government furnished. They might have a whole lot of, I'd say, FSI thought-of and driven use cases that aren't coordinated at the enterprise level. So you'll see that varies widely at agency levels. And you need to kind of meet them where they're at and figure that out and then work on it together.

SPEAKER_02 12:35

Yeah. And I you know I think the there's a there's a gap that I don't think we've fully discovered yet in that. So let's let's say, for example, there is a memo that says, hey, the data belongs to the mission. Uh well, what does that really mean? Uh I I might be a system owner that says, hey, every 30 days, I'm gonna give you a, I'm gonna FTP you a full copy of my massive database that is going to be so hard for you to swallow and ingest in in a way that it supports the actual mission outcome that you're trying to achieve, versus I'm gonna open my system to whatever means of modern integration uh might be best, the best fit for that capability that is is needed to support that mission area. Um that that is a uh a real challenge that you know is going to start to pop up more and more. And you know, we'll we'll tackle it as you get to it. I think the other part you and I have spoken about it in the past, Sean, is the zero trust piece, because um you know Agentic is going to have uh systems like Click's Analytics platform with users knocking on the door saying, Hey, um I'm I'm an agent representing Tim. Uh well, Tim's not in my system. Like this is going to be one of those challenges. How do I provision access to this port unknown portfolio of systems that these engitic uh sort of workflows may be going around knocking on doors for?

Sean Applegate 14:01

Yeah, absolutely. So I think I think what we are seeing is maturity in the data pillar from I'd say lower-level data areas up from our RBAC or attribute-based controls as well. And those take time to build and maintain. It's a lot of back to governance and having a team that actually maintains that with ways. And again, AI can help do that when it can re read files, understand it, and apply intelligent policies. So we're seeing some of that come up. And then on the agentic identity side, I think we're seeing good progress in the space, but it is a very early days. So doing true agentic identity base for like non-person entities, NPEs, is making progress, but it's not fully figured out. So there's I think we're seeing a lot of early ways to approach that, but um, not all of those are maybe enterprise scalable or or enterprise adoptable today for different customers. So it is tricky.

Brian Lake 14:48

Uh real quick, I want to go back to something you said about the need for upskilling. And I kind of look at there's some some interesting counterpoints that it seems like for upskilling you need data, and then to learn how to leverage all this data, you need upskilling. So which which needs to come first here?

SPEAKER_02 15:04

I mean, I I don't think we we've been talking data literacy for about as long as I've been at click. And I think you know it it that upskilling is essential right now to so we used to say, you know, data literacy that we talked about five, ten years ago was how do I argue with and consider the data being presented to me? How do I argue against it? How do I contemplate what it whether I trust it, that type of thing. So now it's less so about how do I interpret this complex scatter plot. Man, what does this mean for my mission in my personnel readiness, for example, to now the system saying, you know, your personnel readiness status is XYZ, and you now need to be able to interpret the validity of this recommendation. Like how do I interrogate with that agent? You know, tell me more about how the how we arrived at this. Uh and again, that gets back to this whole trust thing. So the upskilling there is one, how do I uh understand what this means? How do I interpret something that's already been interpreted for me? Uh and I don't I don't, you know, when you say which comes first, I don't think you can delay on either.

Sean Applegate 16:16

I think they're parallel tracks. Yeah, I think the important thing too is when you think of uh an agent as a as a virtual assistant, you want them to go to the right resources to get the answers. And this is really about designing your your agentic platform, and I'll call it your advanced analytics and data framework to be thoughtfully designed so that agent's not going to some petabyte of PDFs to make some random decision based on the context window and rag that you happen to pull randomly. If your agents are able to interact with structured data, go to the right data source, the right answers that are authoritative and has the skills and the right integration to do that, that's really powerful. And it's really about getting the at most accurate answers possible. And often that doesn't come from unstructured files, it comes from structured data sets that are again a perfect a well-thoughtful, governed data product that's authoritative. And in many cases, for that agent, they might need to go to a couple different data products to do analysis. When you think of a knowledge graph that a human would think about, how do I ask the next level question, connect the dots, do the thoughtful analysis, bring that back and provide an answer. And if they need to go to some rag unstructured stuff, that's fine too. But in many cases, the the authoritative data we run our missions out of, we think of readiness for Department of War or defending a border come out of pretty, pretty well-designed, huge government investments that we want to be able to take advantage of. And um it's not not always apparent when we're dabbling with use cases versus building enterprise systems that we're making very strategic decisions from, if you will.

Modernization Under Budget And Legacy

Brian Lake 17:49

Andrew, I've heard you mention this previously, and I and I think Sean just kind of started to lay the groundwork for my next question for you. Is you described modernization as kind of a rock and a hard place. Sustainment costs eat the budget. New technology is expensive, there's no Pan C or Magic Wand. When you're trying to build a modern data platform, what's realistically the right moves for CIOs who have to do deliver not just this fiscal year, but in the next fiscal year, not necessarily in five years. What do they really need to prioritize and think about to build this modern data platform?

SPEAKER_02 18:25

I you know, part of that rock and a hard place conversation is how do I uh you know, well, should I, and if I should, how do I sustain legacy systems? Um just two days ago, I was down the uh claw and ChatGPT rabbit hole of how much mainframe is still running across uh you know the public sector in the United States because we were deciding what we were going to present at an upcoming AWS DC summit. We we're finding ourselves with an increase suddenly in mainframe data integration. I've got DB2 on ZOS. I don't think it's going away anytime soon, but I need to get that into Postgres in my modern cloud stack to be able to support this outcome. And you know, that's just a realization that the cost of removing that mainframe capability, you know, uh the the original cloud, right? Uh the mainframe, uh, you know, if we can sustain those things and move them forward is going to be a better outcome than I'm going to build from scratch and and build all new. Uh I'm so happy that I'm in the cheap seats, you know, sort of playing sort of a supporting role because uh to be a portfolio, you know, especially the IT portfolio manager, uh look managing the investments across one of any of the large cabinet level agencies or DOD services organizations, it's gotta be the hardest thing in the world because you're the the demand for mission supporting capabilities is just the pace is 1020x what it was. And yet the sustainment cost of everything we've got and can't and having a hard time retiring is there. So I wish I hadn't, you know, how do we how do we thoughtfully approach it? Um, you know, find technologies that help you facilitate the usefulness of those legacy systems that are still serving their purpose in the modern modern AI agentic era. I think we're doing a great job uh with that, whether that's data movement or being able to, again, provide a you know a data AI capability that facilitates the getting of information from those systems. Um it's a balance.

Sean Applegate 20:34

Yeah, I love I love if if you're in a big enterprise role, having the modularity and the right tool sets to fill in some gaps. One of the things we often look for is switch when we we create partnerships with folks like Click, are is that solution flexible and adaptive enough for both our large civilian agencies that often might be more cloud focused, and in this case FedRAP, high and moderate matter a lot. And will I able to am I able to work with that partner also in Department of War or intelligence community customers? The federal market's complex and they all have unique needs. The ability to take what you've designed in the cloud potentially, or maybe what you built on-prem and then move to the cloud later and operate it there, but be able to still deploy it air gap downrange or in a more sensitive environment is really powerful for somebody that might be a CDAO or CAIO that has maybe a general need in unclass environments, but then very specialized needs and more sensitive parts of the environment where they have a trusted partner they can make work everywhere, and they don't have to, I'd say, downgrade their capabilities when they go to that air-gapped environment. And that's not simple. And that's typically where folks that have been around a long time, they're mature, they have a broad portfolio, provide the most value to a CD AI, CDAO or C AIO versus a uh maybe very cloud tethered new modern solution that's hot, but maybe not as robust and not as proven at the enterprise scale, and often not Fed ramped either.

Modern Data Platform Beyond Dashboards

Brian Lake 21:56

But what can you guys can either of you or both of you kind of just Describe to me what does that modern data platform look like from a data set deep in the recesses of an agency environment to a dashboard sitting in front of a mission user trying to make make sense of what this means for trying to execute their missions? Like what are the different pieces of this puzzle?

SPEAKER_02 22:20

Aaron Powell So I mean, you know, again, if you if you talk about a mosaic of data systems, and you know, we're all supporting a lot of that type of thing. It's one, the number one thing is availability. Can I can I make that data available at a low level of latency that supports the what's happening now view that I think AI is going to set an expectation for? How do I get that data so that it connects to the rest of the data sources? I mean, the fact is, you know, we're we're dealing with systems that, again, either are very old and and you know, have some bring a different uh level of context uh than the than the more modern ones. Um how do I bring those together? How do I create some trust on it and then how do I facilitate it into a system that is consumable in many ways? Because that that's the modern data system right now is not about dashboards. You know, when you when people think about click, oh, click helps me build a dashboard. Dashboards are we've we've I think the the industry is tapped out pretty much on the audience that we're going to serve. You know, mu uh you know, when you look at a maturity model, it's like 20% of an audience of a total organization is probably high maturity uh in terms of adoption. But the fact is that 95% of the people that work there make decisions of some variety. So this agentic moment is going to be how do I take data and put it into that decision workflow, whether that's an Appian or a ServiceNow, whether that's a CRM system like a Salesforce or a custom application? Uh how do I make that easy to be consumed by everybody? And that to me is the that modern data system is going to be able to handle the the context for that user and provide the the trust and and and uh governance to make sure that I can deliver that at scale without worrying that what uh Andrew shouldn't see and Sean is capable of seeing is not going to suddenly be available to Andrew.

Sean Applegate 24:28

Yeah, certainly. The one thing that we often find that is overlooked in the data space is that infrastructure to move a lot of stuff around. And I mean that at an enterprise end-end system architecture perspective. So we do a lot of work in the OT space for both connecting networks and building edge environments, like in the organic industry-based Department of War, uh, the wide area network kind of monitoring and optimization connectivity space. And then obviously in the data space as well. And when you think of that at a CIO enterprise architecture level, some of the biggest challenges, especially in parts of the Department of War, are things like D deal environments or modernizing base infrastructure. And unfortunately, you have to have some highways built, you know, network paths to be able to get stuff out of a uh an IoT sensor at the edge or a camera at a manufacturing facility on a artillery manufacturing arsenal in the Army, for example, where you want to be able to do security at scale at the at a global level, not just locally in some air gap network. And so I think building the right infrastructure is really important. This is one of the things I think Mayor Talk identified early in in fiscal year 26 was executive leaders had AI as like a top priority, super important, and data comes along with that. What was on the lowest priority for agency leaders, meaning the secretaries, for example, was like network infrastructure, infrastructure spin. But if you looked at CIOs, the CIOs were like, hey, we're coming a couple years back from COVID. I have a lot of people returning to the office, have a lot of demand on my networks, I have a lot more data being moved around. And all of a sudden you're like, ooh, my highways, my network highways are clogged. That's not sustainable long term. And this was an issue in has been an issue in Department of War as a Marine back in the day, sucking data down a little SATCOM straw that was like 256 kilobits per second. And you know, an imagery interpreter might download a nice image of a target, and that might take 45 minutes to get across, right? We don't have 45 minutes to wait to drive AI decisions now. So being able to do those type of things when you think of just making sure I have good infrastructure as a CIO on a network team is important. What's using it? How do I do quality of service is important? And then once we get all the data back to where it needs to be, whether it's local, regional, or the global headquarters for processing, which is probably all three, uh you you then got to worry about, hey, do I have the servers to compute, the right performance requirements? So the AI agent platform making big investments in can make a decision in seconds, not hours, for example. Or it can't make a decision because the data never got there. And those things happen too uh in in real-world networks and real world systems.

SPEAKER_02 26:54

Yeah. And I by the way, I mean, I think this is going to be a really interesting, you know, sort of learning era, you know, as we try to so now I expect this agent to be able to bring me what I'm looking for in a reasonable amount of time. Well, the dependencies could be network, uh, you know, could be a poorly designed, you know, database, uh, could be, you know, the query system. You know, click you know magically. You know, I I I think we really just lucked out. Our engine that we was developed 35 years ago, it was almost like it was purpose-built for this agentic moment. Like our our performance versus a SQL-based tool like a tableau or a Power BI is going to be 7 to 10x in responding to that query. And it's going to be 7 to 10x less expensive from a token burn perspective. These are all these things that were, you know, you read these stories about this token burn problem across large organizations. It's only, you know, we're we're at the you know the foot of that uh learning moment. Uh and again, you know, there's so many dependencies that we don't even understand yet. You know, we were three years ago, we were sort of hyper-focused on ingress-egress costs as we move data between these things. And that's gonna be that that's gonna seem like a really easy problem as we reach you know these next wave of challenges.

Brian Lake 28:14

I mean, we just talked about this yesterday, Sean, about IBM burning 20 percent of their token budget for the year in one weekend doing a single agent running on mythos. I think I think that's what the story related to me.

Sean Applegate 28:27

Yeah, they it was at IBM Think. So it was there Tuesday, their CT federal CTO was telling a story, but he basically said, hey, we have uh like five people and about five million bucks in tokens budget with mythos kind of set aside. We we set this agent loose uh for the weekend on finding vulnerabilities and patching them on one software applications repo. So this is you know, it could be a really big repo. I don't know what it was, but they came in on Monday and and looked at the bill and it burned a million dollars and basically Saturday and Sunday. Yeah. And they go, wow, that was 20% of the whole budget for this project, like for all of what we were going to do. It's nuts.

Brian Lake 29:03

They were like, yeah, we still have 363 more days of the year to figure out how to spend the rest of these tokens, right?

Sean Applegate 29:09

Yeah. So I mean it look, the bottom line is you I'm a big fan of continuous improvement and building efficient systems, but it also means you need to balance that with wise budget spend. And when we talk, Andrew, when you mention things like, hey, we can do very targeted decisions and get answers back, it's 700%, more efficient than some other, well, I'd say more legacy tool sets. That's really important. When we talk to a lot of CIOs and other folks that are building systems, they have limited budgets. Those budgets are not unlimited. We represent the taxpayer's investment in in the US government through tax dollars. We need to spend as efficiently as possible. So building systems that are are performant, well governed, trusted, and don't cost a bazillion dollars is is important. And we have to approach that wisely. So looking at token consumption, along with is it a a good answer, right, from a LLM as a judge, or evaluating the output is an ethical answer, and is it based on trusted data, are all really important. I think that's the the challenge for a lot of C D AOs today or C AIOs is balancing competing priorities and having a set of uh uh experts that can help guide them down that path because it's not it's not simple is the answer.

Mission Experts Beat Platforms Alone

Sean Applegate 30:22

Yep.

Brian Lake 30:22

Andrew, I've had the pleasure of being able to see you speak at a bunch of different places across the industry. And I think I've heard you say many times that don't underestimate a federal employee who understands the mission with the right data in front of them. A lot of the conversations we've been having is that AI is gonna solve all this. And now just the past few minutes we've been talking about the AI is giving the answer, right? So it how do you just kind of talk about the friction between people who just blindly believe that the platform or the AI solutions are gonna replace the mission-focused, mission-driven people that have been doing this their whole career? Yeah.

SPEAKER_02 30:59

I mean, so when you think about um you know AI and in particular generative AI, it it really these are these are knowledge systems, and they're building knowledge off of something that's been documented, something that someone knew. Uh the fact is that at the moment in our government, the the the knowledge still exists in those subject matter experts. You know, that you're net like you take a problem like logistics, you take a problem like recruiting, you take a problem like the DoD audit, you take a problem like drug regulatory approval. There is a person uh in that organization that knows more than any system will in my in the time that I'm operating in my career. And you know, I think the misstep here is that we just say, oh, just forward-deploy engineers, they sit in a room, couple of interviews, uh, you know, MVP evolving into final capability. No, that you're never going to do that. You can do you can get that absolutely critical 20% off the top type of solution that way. But the bulk of what these what government agencies do is so dependent on those expertise, those experts being able to apply what they know to the system that is made available to them. I'll take any any day of the week. I know some people in government that are, you know, there's there's a guy in logistics, he's also an AI group, he's become an AI guru, self-taught, but I would have taken him and put him with a legacy set of tools and bet on him every day of the week over uh a forward-deployed engineer with uh the most productive AI just because the guy knew what happened in that organization. Now, him plus AI, it that's that's gold. Things are the thing, the the hard work that he used to have to spend three days doing now take him three hours, three minutes, whatever it is, he's just more effective. But we we must really double down on empowering those subject matter experts to do more of what they've already been doing.

Brian Lake 32:58

Context is critical in this state, right? Absolutely.

Sean Applegate 33:02

Yeah. So yeah, maybe our quick story. We have a we have a customer, we've deployed Qlik at the enterprise level, and so they've made a choice to do that. Um they had one of their lines of business that had some, I'll say data scientists for deployed engineers, if you will, that were from a different uh integrator. They didn't understand how to use the enterprise platform. They had chosen, because they had been educated in college on these things, to use open source data science libraries, which are really cool. They can do neat stuff, they don't cost a lot of money. But if you're struggling to deploy a product after weeks of work and get it to deliver on the mission, you probably haven't hit the target efficiently. We brought in one of our expert SMEs, who happened to have a master's in data science, who was well educated like them, but he knew how to use the click platform to hit the target. And he also had background in ship manufacturing as well, luckily, uh with the Navy. So he understood the Coast Guard mission. And in about three quarters of a day, he solved the problem for that line of business that they wanted to solve with knowing where to find the data, where it was trusted, knowing how to build the dashboards and reports and the dis make the decisions they needed. So what they found was wow, hey, Swish, you guys were able to come in, solve a problem in three-quarters of a day or a day maybe. And we were struggling with this multiperson team using, I'd say, outdated tools that maybe weren't enterprise ready, and they had been struggling for weeks. They were super happy. But I again having the right that not only the right employees, but having those employees skilled in the platforms that you've made big investments in to drive outsized returns at a very rapid pace are important for people to consider. That's why often we get engaged, we'll bring in a small center of excellence, right? A three to 10-person team. And one of the most important things we try to do when we're hiring is bring in mission knowledge with the engineers along with to technical skills. Those two things together are extremely important. Um, as an old marine, for example, I understand the Intel community, and a number of the folks we hire are veterans, love hiring veterans, so they bring a lot of mission knowledge because they live the mission. They're also passionate. So those guys will go take the hill with the warfighter next to them and do whatever it takes to accomplish the mission, right? They're not checking out at 4 p.m. and worrying about what time of day it is either.

Brian Lake 35:14

I think that's also a lesson for those folks that have recently left government over the past year or so. There's really great uh intellectual talent and property out there, and organizations would be foolish not to try to get those folks into their ecosystems as quickly as possible.

Sean Applegate 35:31

Trevor Burrus, Jr. Yeah. And they aren't just people either young or old. I think it's people that are curious and want to continuously learn, and they want to try to solve a problem. Give them a real problem, they're passionate about, they'll apply the skills, learn it, and they'll be able to re-regurg that later. You know, one of the other other customers we have is a Department of War customer, and you're pretty familiar with them, Andrew. Um, but they sent all of their executive leaders to a three-day AI course at a, I'll say, top-tier university, and they came back and applied those things immediately in the click platform. I met with uh two of the two-star generals that I think went through the training a couple months later, and we were talking about kind of solving problems. And these two individuals are like, oh wow, yeah, I've already applied those things in our dashboards. I built my own dashboards as a two-star general. Had that been 30 years ago, you would have never heard a two-star general say, hey, I'm building my own dashboards, doing agentic things, and I built this little assistant and I can ask you questions, and I get questions back immediately. Like that is super powerful, but it really speaks to the ability of the not just the analyst to get the answers quickly, but but a person running all of operations for a big agency to go, yeah, like I'm gonna go build some cool stuff that makes my job easier. And they're really happy with that. And they get personal satisfaction out of it, that they can still sit down and kind of work the mission side by side with the people they're they're going to fight the battle with, whether it's a colonel or it's a PFC around the corner that they're representing, right? They are a doer, not just a leader or a figurehead.

Brian Lake 36:52

Yeah.

Sean Applegate 36:53

Quad charts be damned.

Brian Lake 36:55

Yeah, I'd love to get rid of quad charts. That would be nice. Our

Breaking Silos Through Leadership And Law

Brian Lake 36:59

whole careers, we've been talking about breaking down the data silos. Like it's been every couple months, you see the new headline, we're gonna tear down the side, we're gonna tear down the walls, tear down the silos, and it never happened. It it feels actually like it's happening. What's driving that? And what are you seeing from your from your perspectives that is one occurring, and what are the drivers behind it that is achieving this dream that we've been talking about?

SPEAKER_02 37:29

Yep. Well, I mean, so one, you know, you got data silos and then you've got mission silos. And I think you know, AI is breaking, you know, the the the expectations that people have, that senior leaders have of AI are that I'm going to be able to understand my entire area of responsibility, not, you know, I'm going to get a snapshot of seven different pillars that I then try to understand what the influences are. So it's necessity. I mean, you know, there's this moment that we're in where I understand that data is absolutely critical fuel for these, you know, these insights, these uh, you know, for decision advantage. Uh you know, they're they're trying to get to the most real-time comprehensive view that's there. And so one, it's the people, policy, and politics are are breaking down, and that is that is it. You've got whether it's uh a memorandum, a change in the way that funding is allocated, or modernization. In some cases, it really is some of these systems folding together. Uh you know, take you know, Army EBSC, you know, we'll see how that emerges on the other end. But you know, you had systems owners that said mine, and for may per perhaps good reasons, concerns about performance, concerns about changes to the contract that wasn't in scope, uh, you know, change about governance. I'm I have a responsibility to protect this data. And when I, you know, when I release it out into the wild, how am I certain how it's going to be controlled? So I mean, these are the things that are evolving, and you know, technology is playing a role, but ultimately it is it's leadership. You know, the when when leaders understand you know why that's important and choose to take action despite all of the pushback, we're going to see results.

Sean Applegate 39:22

Yeah, it would double down with if the leaders are wanting to build data products and they put, I'll call it, you know, product owners in place of that, where they think more like a traditional tech company or tech culture that leads the build great data products or great applications would be another way to think about that. We see a lot of momentum around focused execution because they think more about how do we make decisions, how to re-engineer the business process, how to use the data across the organization better instead of thinking about the individual team silos of people and teams in a more isolated sense. And so when you think of kind of driving the handoffs across the value chain, for example, they get a lot of value from that. I will say from a policy perspective, the thing that I think often does slow us down are things like how do I navigate some of the different um legal limitations of how we who can access data and where. And that's sometimes at a cross-agency level. When you think of somebody like CBP working with FDA, like what are they allowed to do? What types of data do they have? Where does it fall inside of legal limits of the law? So, like Title X, for example, versus other areas, can I blend that data, or can certain people have access to it or not, depending on their jobs in the government that laws allow them to have access or not in some cases? So when you look at things like data sets at CGIS or other parts of the government, you know, certain people can have that, but you can't necessarily let everybody have it. And so you have to follow those rules. And that's where things like, again, being able to configure those things in a DOP policy, a data catalog, have the metadata associated with it makes that really easy for others to plug in and operate within the boundaries of the system, but for a governance team to understand that they've given the appropriate people access and authority to use the data, but they've also said, hey, I've got controls in place that limit the uh use of the data so I can go look, it's designed responsibly, ethically, and we can stand behind that as a U.S. government agency that we can tell the U.S. citizens we're not doing things we shouldn't be doing based on the laws that have been passed by Congress. And those are things that often we do have to make decisions about that are not always the most intuitive decisions, they're not always the most efficient for the mission, even, but they are the right decisions based on the laws that are on our book. And those things need to be respected and accounted for. But it's not easy, right? That's a lot of why we have lawyers that get involved in designing applications or how we use the data, not just the engineers.

SPEAKER_02 41:45

Yeah. Yeah. I I mean it's particularly when you have person data in um and particularly citizen data. And it's one thing. Take the DOD the data about you is essentially their data. So that you know you they employ you, they pay you. You talk about citizen data, you know, completely different challenge. I mean, DHS and privacy impact assessments. I mean, that that occupies a lot of the systems development lifecycle is you know, can we bring these data sources together for this use case that will be served to these audiences to make these decisions? I mean, that's a that's a very, you know, interesting challenge that they're up against. Um but again, you just you take uh so Advana and Maven, two two two different systems, two different purposes. Advana was literally, you know, mostly the boardroom side. Yeah. The business of DOD. Maven was the battlefield side. Uh battlefield side moved much more quickly. Yeah. There were a lot less mine, mine, mine, because you know, the you know, it would the it was natural because the outcome systems side, uh that get equated to money and power and who, you know, who who's you know, rice bowl was going to continue to be bigger. Uh that you would look at that as a case study, it's a I think it's a magnificent, you know, sort of and I've we've been there since 2016, since inception, to be able to watch that has been just remarkable. And it and I'll just go back to my statement, leaders make the difference. Uh the reason in in there it was not a linear shift. You know, we went plateaus and growth in terms of getting more data sources and and more openness to the you know the use of that data. And it was because of certain people that arrived and backed up, you know, the you know, the the path towards the ultimate goal. Um and you know, that I think that's where we're going to be. You're gonna see organizations that have cautious leaders and they may be right to be cautious, and you're gonna have people that say, this is so critically important, you know, we are going to move forward absent some concrete evidence for not doing so.

Brian Lake 43:56

I mean, it sounds like then when I think about the people process and policies, it's the policies that is the hardest thing to overcome here for changing to drive this stuff forward because it sounds like the people have the drive, the processes can be changed in real time. It's the policies that sometimes get in the way here.

Sean Applegate 44:15

In a lot a lot of cases, the the policies are not in their direct control. Right. It might be ran by Congress or some other organization where they have to it and it purposely is designed to move slow in some cases, adapt slowly, be have checks and balances. So you have to operate within the boundaries of of what you're given in many cases.

The War Of Inches

Brian Lake 44:33

Right. So let me let me ask you this, Andrew, I know we have a we have a a breakfast briefing coming up in the fall, which uh we'll make sure we let folks know about in the show notes and we'll get you back on before that time frame. But you just had your public sector summit. You had both customers and non-customers alike speaking at your summit. What was some of the one of the biggest takeaways that you that you really drew from from folks that were sharing at your at your public sector summit?

SPEAKER_02 44:58

So if you looked across the content, you know, absent, you know, if you take out Click obviously talked about Click's products in our roadmap. I mean that that that that that happens at the event. But you know we had a panel on data literacy. We had uh Adja Kakara who's the the the CDO for uh Deloitte uh global public uh government and public sector um we had you know multiple other presenters that if I had to distill down what they you know said was the people are going to matter the most at this moment. Like they we we've always every year that we've done that event, we've always said our main purpose is the community that we're trying to build, getting them together, share lessons learned, you know, good and bad. And it was the always the reason that we that we ran this event uh and you know they were the fuel for what was happening in our data movement. I th I really think that it it is two to three X more important right now that the people are going to be the ones who either put in place these little building blocks, and they are going to be little building blocks of this agentic future, you know, people that are again curating data, breaking down data silos, building it data products that are representative and trusted, developing integrations between the LLMs and other AI capabilities and systems like Click. These are going to be so many small things, but the people are going to make that difference. And it is you know it's like it's the war of inches. I get in by you know we do the Robert De Niro any given Sunday locker room speech. I really think that that's you know where we are because there's the the obstacles that we will overcome are going to present themselves every single day. And the people with the fortitude and the drive and the passion that to to progress through are just going to keep on making progress. And we're see like I I love working with those folks. I love hearing from them about the the outcomes that they're achieving the the surprise moments when senior leaders go, you built this yourself in the last 36 hours? Yes I did, sir. Those those are you know golden

Monday Moves MVPs Learning And Wins

SPEAKER_02 47:20

to me.

Brian Lake 47:20

Yeah that that to me is what leads me to my final question for you both is so if you're a CIO, C D AO, C AIO, you know if you're if you're thinking about those small building blocks and you come in on Monday, what are what are some of those small building blocks you should be tackling or some of those priorities you should be thinking about Monday morning that you could ideally now accomplish in a week if not less? What are you thinking about from that perspective?

Sean Applegate 47:48

Yeah for for my guys it's normally you know what are the the biggest mission priorities and how do we build something small that provides value? So think of MVPs and empowering your workforce to experiment, fail fast, have a safe environment to experiment in. So you have to give them air coverage a leader, I'd say the second thing is think of how you build a culture of learning and so you let make sure they make some time to learn and apply their learning. And so there's some great resources to do that. If you're in the Department of War most folks have access to O'Reilly.com or if you're a veteran you can get to that normally as well. Most civilian agencies give access to that they have a great data literacy learning track to start with but a lot of really deep resources to get hands on if you really want to go deep down the rabbit hole. I'm personally a big fan of Click's data literacy program, which there's a lot of free resources for on your website. And then there's some other maybe paid training paths that are more professional, enterprise scale, come in, teach the broader teams how to do it, apply it. I think doing those things collectively as a team where your leader can grab the torch and say let's show up for a lunch and learn or hey set aside two hours a week to go learn something new or you know come show what you built, right? Do a little uh little little kind of a you know hack hackathon internally or some of the great events. Stuart Wagner, for example, uh he's over the Navy now but he was the I think CTO of the uh Air Force at the time did did Project Bravo and they did amazing stuff having airmen come together with an idea um support them with some blended teams that you brought people in that were engineers and applied and built things in real time like in 48 hours of of time. Those are really fun projects for a CTO. I'd love to see more of that across the government.

SPEAKER_02 49:23

Yeah um you know for me uh it's you know celebrate the wins uh I mean the fact is is that like I said there are people making progress all the time. Um there's a Lori Mongold been a long time you know she was Army G 357 now at Booz Allen and just also just a you know big data warrior uh always been a big proponent of data. She often gave a speech about first fan, the the you know one person dancing at the music festival that all of a sudden erupts and and grows virally and it's I think that's the the you know the moment that we're in we've got to shine a spotlight on the places where we're having success because again there's a lot of folks that probably feel like they're kind of boxed in from doing the things that you know that they're hearing they should be doing how do you highlight the folks that are having success doesn't matter how small uh how do how do leaders really make empower those folks, find those lighthouse things that inspire that that next moment you know obviously as a backdrop to that, you've got to be taking feedback, understanding what the obstacles are and determining within the confines of the policy, you know, we we operate, we're a highly regulated industry, probably the most highly regulated there is within those, you know, how can I move some of those barriers safely without introducing risk that outsizes the the the the the benefit of the outcome but again I'll I'll say if you would come in Monday morning, it's turn turn those people that are delivering those wins into heroes in your organization.

Wrap Up Resources And Listener Prompts

Brian Lake 51:02

I love that and uh obviously I love the uh the viral dancing analogy Andrew this has been great conversation. We got to have you back in the fall where we'll get to hopefully we'll preview our breakfast briefing and be able to talk a little bit about what we've seen in the few months between this conversation our next conversation. So thank you Andrew for joining us today. For those of you listening obviously everything's in the show notes. Let us know what we missed. Let us know what you want to talk about next we'll have uh some links to both some of the presentations from the Click Summit, from our webinar with Kevin Murphy, the CDO of NASA we'll make sure we have some really great music clips for you to check out but come find Andrew and myself at a show or even we're gonna get Sean out to a show this summer too. So come find us out there on the dance floor go to gist360.com for anything else that you might need. Like follow and subscribe wherever you get your podcast. Andrew Sean thanks guys great conversation today as always pleasure to be here have a good one