2024 Air, Space & Cyber: The Digital Enterprise

The “Digital Enterprise” panel at AFA’s 2024 Air, Space & Cyber Conference featured Peter Kunz, vice president and chief engineer for Boeing Defense, Space & Security’s Air Dominance division; Mark Andress, global vice president of defense and intelligence at Oracle; and Sean Moriarty, CEO of Primer AI. The panel, held on September 16, was moderated by retired Maj. Gen. Kim Crider, founding partner of Elara Nova. Watch the video below:

Panel Moderator: Maj. Gen. Kim Crider, USAF (Ret.) Founding Partners, Elara Nova:

Please join me in welcoming to the panel. Mark Andress, the Global Vice President for Defense and Intelligence at Oracle. Peter Kunz, the Vice President and division chief engineer for air dominance at Boeing Defense Space and Security. And Sean Moriarty, the CEO at Primer AI. Gentlemen, thank you so much for joining us here today.

All right, so we’re just gonna jump right into it, but feel free to offer any additional remarks about what you do in your role as we get into this and how you see this whole idea of a digital enterprise emerging. The first question that I have is really for each of you. From each of your perspectives, what facet of a digital enterprise have the greatest impact in creating competitive advantage? Mark, we’ll go ahead and start with you, and we’ll just go down the line.

Mark Andress, Global Vice President, Defense and Intelligence, Oracle:

Yeah. When I thought about this question, I harken back to when I was in the Pentagon for almost a decade when we were going through the Budget Control Act sequestration, and it really made me think about how we approached everything back then. Real hard budget numbers we had to hit while at the same time I was on the Navy staff, we were trying to launch the Navy into unmanned. So, it was this crunch the budget but yet grow new capabilities. It’s probably the same throughout time, this challenge.

But what struck me was that every single aspect of enterprise optimizing was pursued, right? There was no stone that wasn’t worth turning over to see if you could try to get real cost optimization through some kind of enterprise or digital initiative or looking for some kind of competitive advantage through advanced networking, through space-based ISR, et cetera.

So, I think in these two areas, one on cost optimization, I was fortunate to sit in with our CEO, Safra Catz, when she was meeting with some senior government folks, and they were asking her, “How does Oracle approach this? How did you get your start?” And she goes, “Well, it was really simple. I was told I got 18 months to take a billion dollars of cost out of my company. Just go do that, right.” And she talks through how it seemed very simple, but she looked at a global company and said, “Wow, I cannot believe how many snowflakes we have.” This country claims they must do it this way for some kind of law on HR reasons, and this country has an exchange peculiarity. She just refused to accept them and kept rolling out enterprise capabilities. But then she thought about how to ensure adoption. And what she would do as she picked a couple of winners, the early adopters, to show that she could substitute your unique Snowflake capability for a larger enterprise capability and then would return part of that savings back for investment, either in tech, the tech or the sales force, or whatever that country needed to expand. She really tried to make it a win-win.

So that’s just one example of what she was able to do. But I do think it’s both. I think it’s cost optimization as well as mission value.

Excellent. Thank you very much, Peter. Your thoughts?

Peter Kunz, Vice President and Chief Engineer for Boeing Defense, Space & Security’s Air Dominance Division:

Uh, well, I’ll start with digital enterprise as a super broad topic, right? So, you can somewhat fit everything under the sun, but when we think about it as the enterprise in support of the US warfighter, it runs the gamut, but sums up around the idea of rapidly accessible, common, understood, trusted, defensible knowledge. It’s about speed to knowledge, that might be operational knowledge against a threat profile that we all see evolving really quickly. It might be that support of open system architectures as they emerge. Even though we may think of those as a mission systems element on a platform, in effect, those are an element of a digital enterprise if they’re to be effective.

And then as we work into the business, as we try to support change, can we do so quickly, efficiently, and accurately? If you all decide you want things a different shade of gray, what does that break? What does that fix? How long does it take? And what does it cost? All of those are falling in under our very broad focus, which is how to approach digital and what it can bring to the industry and to you as our customer, covering that gamut in very different ways across the spectrum.

Panel Moderator: Maj. Gen. Kim Crider, USAF (Ret.):

Thanks, Sean.

Sean Moriarty, CEO at Primer AI:

Yeah, I think, you know, I’ll echo some of what Peter said. I think the most acute problem we’re facing today is the lack of speed. It’s actually not a lack of capability. We have profoundly powerful technological solutions to many of the challenges we face. But the time that it takes to get those capabilities in the hands of people who need them is far too long.

*Sorry about that, there we go, that sounds a little better. I would have been happy to pick up the voice, my indoor voice at home although no one listens to that either. But back to speed. We’ve got the capabilities—profound capabilities—in this country being developed, both public sector and private sector, but the time it takes to get our capabilities in the hands of the people who need them is far too long. And we all know that the single greatest impact we can have today is to get these powerful tools in the hands of the user as quickly as possible so they can make and implement better decisions, right? Fundamental to that.

Now, underneath all of that, you need to make sure that you have the infrastructure, the policy, the access to data. One of the things that you hear bandied about quite a bit is within the DoD, estimates are, I don’t know, less than 10%, and perhaps less than 5%, of all potentially useful data is either very difficult or impossible to access. And if you think about what that does, standing in the way of decision advantage, it’s profound.

But I think the second piece of this is a cultural shift with respect to expectations. I think people are in violent agreement on the threats of the age and what must be done, but cutting through that and moving quickly is really tough. And, you know, part of this, again, it gets to expectation, which is how long should it take for us to deploy a state-of-the-art technology when we’re accustomed to things taking five to seven years, but profound technological changes happening every 90 to 180 days? We’ve got to reframe our expectations around what we must do to deliver capabilities within timeframes that matter.

Panel Moderator: Maj. Gen. Kim Crider, USAF (Ret.):

Yeah, thank you very much. So, what I’m gathering from you guys is this whole idea of competitive advantage really points us to a lot of things. If you want to get competitive advantage, if you really want to be able to be way out in front of your competition and create the kind of capabilities that are gonna allow you to continue to, you know, sustain that advantage and prevail in the case of the military, prevailing in a conflict, you’ve got to be very focused as an enterprise.

I’m thinking about a few key things. You talked about mission value. What’s the mission value of what you’re bringing to bear? And how do we continue to focus on that as the areas that we’re gonna drive our digital enterprise, cost optimization? Another key piece of this, you can’t have competitive advantage if you’re gonna run your costs into the ground. And then, of course, access to knowledge, right? We’ve got to have access to the information. And as Sean pointed out, some of that is still very, very difficult.

We’re gonna drill in a little bit more here on the data challenges in a second, but making sure that we’ve got rapid access to some of that knowledge and that we are positioning ourselves to act more rapidly, to be more agile, to take the culture and evolve ourselves into a more rapid, agile, responsive entity is clearly all-important facets of how do we drive competitive advantage in the organization.

From what you guys have seen, and I know you’re out there working with programs in the Air Force and in the Space Force, and Sean and I were talking earlier about work that they’ve been doing in the Intel community as well. Where do you see the biggest challenges or hurdles beyond what we’ve talked about here today, and potentially some of the biggest opportunities to get some of this flywheel going and to start moving in the right direction?

Mark Andress:

I think I’ll start with opportunities. My last job, I just retired a year ago, and I had the honor of being the CIO out at the National Geospatial Intelligence Agency, NGA—great mission, great people. And I think I used to think a lot about the opportunities that were happening, these exponential opportunities that were happening in that space, at the same time. And I really had not seen them converge in this manner. There were three big ones, right?

We talked about it a lot here, and that is the commercialization of space. That was exclusively government for so, so long. And now, we were dealing with commercial imagery all the time, normally one or two big ones. And then, as I was leaving, there were not only commercial satellite providers but also commercial services companies—companies that would use imagery to derive a service and just give that to the government, an intel agency. So, this proliferation of commercialized space.

The second piece of this was cloud. An Air Force program manager back in the day would have to go put a procurement contract, buy a piece, or get some time on an HPC farm somewhere, in order to do the types of things needed for training models, for example. Now, they can just simply dial up what they need, how they need it, from multiple cloud providers. Now, it’s truly a commodity that’s available.

So, I’ve got proliferated sensors and communications, I’ve got a commodity for any type of compute storage I need. And the last thing that’s so exciting is the commercialization of AI. The whole investment by the commercial industry on defense is different. The attitudes are different. Companies are not shying away from saying, “I’m going to be a part of a defense contract.” I’m going to allow my software to be used to protect my homeland, to protect the airmen.

You take those three things combined, and it really does unlock an explosion of data, the ability to process it in a rapid cycle of new and innovative algorithms that you can test. It works great; if not, throw it away, get the next one, and keep going. So, I think I’m most excited about the opportunities that are coming in the next few years from those three things.

Panel Moderator: Maj. Gen. Kim Crider, USAF (Ret.):

Awesome. So, as-a-service capabilities, you mentioned imagery, cloud, being able to access that information very rapidly, and then being able to integrate that with AI and get that learning process going—some real opportunities that we want to continue to pursue, right? And all commercial.

Mark Andress:

All commercial, yeah.

Panel Moderator: Maj. Gen. Kim Crider, USAF (Ret.):

Right.

Mark Andress:

It’s unreal.

Panel Moderator: Maj. Gen. Kim Crider, USAF (Ret.):

So Peter, how would, what would you add to that in terms of the real opportunities that you’ve seen happening in the Air Force and in the Space Force today, where we’re really starting to get the flywheel going, move faster, move more effectively towards a digital set of capabilities? And you can even speak to the challenges too, if you like.

Peter Kunz:

Yeah. To some extent, there’s sort of two categories we look at, and one are the green fields, the new starts, the emerging programs. In that case, we are really starting to see the potential become real, right? That ability to merge up what might be program management functions and finance functions into impacts, opportunities, and risks on the technical and on the design side—that’s starting to happen.

In those cases, you have the luxury of that green field. On the legacy programs—things that have evolved and continue to evolve—there’s different opportunity and different challenges, and in some cases, it’s the same thing. The opportunity and the challenge is the data, right? We have programs that have evolved over decades. We have decades of data. And to your comment on pace, the pace of some of these programs, the tack time is orders of magnitude slower than the supporting technology cycle tack times, right? When we started the F-15, we didn’t have AI, and in fact, we barely had computers. So that has evolved over time.

But within, there is such a rich history of experience that the opportunity is, as technologies such as AI—whether currently sort of narrow AI or reactive AI—come to bear, to be able to bridge that. Things like understanding what’s the best answer to resolve a defect or a question now is no longer a bunch of engineers trying to solve the problem for the nth time over 30 or 40 years but leveraging that historical experience to try to get to the answer faster and get to a better answer that we can trust.

Sean Moriarty:

You know, this is something I’m privileged to sit on, the Atlantic Commission’s present council on software-defined warfare, and you’ll have upwards of 60 profoundly capable people working on this problem set for the better part of a year. And it breaks down into probably a dozen constituent parts.

One of the areas, though, I think particularly when you’re an early-stage company like we are, bringing emerging technology to the environment, the practicality for us from an opportunity standpoint is that we’ve got to identify opportunities where there is clear urgent need. Our capabilities can solve for that need very well. You have to have a champion who is very comfortable with reputational risk. They also need to have the span of control and authority to make that decision. And you also need an environment to deploy into—for example, access to data, compute, and even the ability, even if the data is in useful form, to get access to various compartments against the kind of rules and requirements of the privileging of that data. Understanding those preconditions for success is really important.

But the other thing that is important is that total cost of ownership. AI is filled with the smoke and hype of any new technology where commercial companies are seeking to make a fortune by bringing it to market. But underneath it all, represents extraordinary potential, and we are very much at the dawn of a new age with respect to these capabilities.

And, I think we can’t be afraid to take the swing to maximize the opportunities that these capabilities present. But going back to that fundamental issue, the risk calculus of the adopter needs to change, and the organizations need to support that change in risk calculus because we learn by doing. These technologies will deliver for us to the extent that we actually bring them into environments and put them to work right away.

From a cost perspective, one of the things we focus on is how can we dramatically lower the cost of what you can get. For example, with a cloud-based LLM for the customer, not everything requires a large language model. You can do work with smaller models that are specific to the context and can give you far greater performance at a fraction of the cost.

Now, unless you are a willful experimenter with these capabilities, working with a trusted partner, you won’t be able to see those gains. Those opportunities will just be theoretical. So really pushing an early adopter mindset is a fundamental precursor to developing the pace we need.

Peter Kunz:

Yeah. If I could just add to that, it’s spot on. A lot of it is around intentionality and understanding of the tools. And at the end of the day, and I say this from a supportive vector is, these are tools, and there’s a spectrum of tools within that portfolio, and knowing which one to use when is critical to success, right? You can hit nails with screwdrivers all day long, and you’ll say, “What a terrible hammer.” Well, it is, it’s a screwdriver.

So, within the spectrum of tools, spectrum of AI, there are different tools for different problems. And working with folks who have that expertise to know which, where, and when is critical to success and critical to frankly not becoming disillusioned with some sort of high-level macro premise of what it should do for all of us.

Panel Moderator: Maj. Gen. Kim Crider, USAF (Ret.):

Mark, uh, let’s continue this theme a little bit on, you know, kind of talking about AI and how can we use AI or what are some of the challenges in the application of AI that you’ve seen in your government work?

Mark Andress:

Yeah, obviously the most exciting thing happening as I was departing was the work that was being done with Maven, in particular starting out on FMV, but it transitioned over to NGA as I was leaving. And, kind of echoing what you were saying earlier, I think the quickest advances we would have are when you try to tackle as discrete a problem as you can, set that box around it, make iterative progress against a fairly tightly defined box, celebrate your success, and then grow. Right? It’s this constant little iteration.

So, for example, some of the things that really impressed me about Maven were, number one, the data is the gold. Your data is the gold. So your imagery, your data tags—that is what you have to protect the most. They would set up these test areas where the tagged data would be there, and it was government-owned, government-operated, and a commercial company could bring in their algorithm, run, and then leave. But knowing that area—that is the unique value proposition that will give the government its advantage and allow you to move quickest.

I was just very impressed with how they were doing that.

Panel Moderator: Maj. Gen. Kim Crider, USAF (Ret.):

Yeah. But as we’ve mentioned, you hit it here again—the data can be one of the biggest challenges. And certainly, in your line of work, Sean, I’m sure that what you guys tackle with, in providing solutions for the government, you know, you really have to deal with the data challenge. What are you seeing in terms of the challenges that we have with data in the government, and maybe some examples of how working with the government—Primer AI in particular, but industry in general—has been able to help work through some of these data challenges?

Sean Moriarty:

Yeah, you know, I think most folks—certainly probably everybody in this room—is aware of the constraints and the limitations of the data that exist in the environments, and your limits in getting there. You know, where we are, if you move kind of beyond AI, the digital enterprise, we’re at the beginning of—well, I don’t know if we’re at the beginning, but we’re still very early in what ultimately will be a multi-decade IT transformation of the government enterprise.

And so, I think understanding that data needs to be accessible but also protected. And when I think about that, that’s really about access control. And whether it’s high side or low side, or which user group should have access to information, you certainly need those frameworks. But the data stores that we seek to access in some cases—we know they’re disparate, with very different technologies used to get them. Standardizing APIs, enterprise data models—those are all essential, and people are doing some of the work on them.

I think one of the advantages of today’s technology is the ability—provided you have an interface, it doesn’t even need to be a very good one—to ingest massive amounts of data, unstructured or structured, and make sense of it much more quickly. And so, I think opening up and exposing data, still providing the policy and the protection around it so it doesn’t end up in the wrong places, but that basic exposure can now unlock extraordinary capability.

I’ll give you an example from an open-source perspective. When you’re dealing with massive amounts of unstructured data of unknown quality, we have the ability to go through hundreds of thousands if not millions of accounts—whether they’re news sources, social media, dark web—put that in the hands of an operator or analyst in a real-time stream, and they can get real signal from it. So, a lot of the data is about policy and then basic exposure. I think sometimes people worry they’ve got to get it into the exact right single format to take advantage of these capabilities, and they don’t.

Panel Moderator: Maj. Gen. Kim Crider, USAF (Ret.):

Yep. You want to add anything to that, Pete?

Peter Kunz:

Yeah, just similar. So, as you alluded to, sort of security enclaves—we also sometimes struggle with creating the right partitions or enclaves where we want to have that common and effective working space between the customer and industry, right? So, these enclaves tend to come into industry and include and incorporate a program, but not necessarily the entire company because we have lots of programs with lots of customers.

So, making sure we create those pools of access, both on the customer side and on the industry side, so that we’re doing the proper due diligence to our variety of customers, shareholders, and supply base becomes a complexity and an ongoing cost of those programs.

And the cost element also factors in when ideally, you’d like this to go from ceiling to floor, right? That’s the utopian solution. At some point in our supply base, we hit Bob and Mary’s welding shop, right? Bob and Mary’s welding shop doesn’t necessarily have the return on investment of the business case to align to the digital enterprise. So also, being intentional about which point it becomes paper and sneaker net, driving that as low as we can, but understanding the cost to, at the end of the day, the nation, to support that penetration of the digital enterprise is something we have to be really careful about.

Mark Andress:

Yeah. Okay, I was just gonna add something on the data front. I think that’s an important topic, and it’s an observation since getting out of government and learning a little bit more on the industry side. So, you know, I’ve been a part of massive XML ontology drills, and you know spend—you’ve got an army of people that are figuring out the structure, and you’ve got a compliance thing around data, functional data managers, ETLs—it was a very manual process.

Even at NGA, I had analysts who wanted spatial reference data, graph data, vector data, relational data, and everyone had their own special thing, special backend data store. All that’s changing, and we talk about AI on data. Spend some time looking at some of these modern data platform companies. They’re not database companies anymore. They—like at Oracle, we have 23 AIs, and all the investment has been in the AI within the data platform so you can take any structure, any format, anything, and it is constantly learning about your data to the point you just ask it questions. You’re not writing complex queries.

So I think there is technology within data platforms that are out there, that are just as exciting as, you know, applied AI for computer vision or other areas that can really help.

Panel Moderator: Maj. Gen. Kim Crider, USAF (Ret.):

Okay, great. So, let’s drill a little bit more into the cloud piece of this. You know, clearly we need to be able to get information and insights to the edge, right? And we need to be able to leverage the cloud to do that, both with the data and with the AI. Can each of you speak a little bit about what you’ve seen in terms of the evolution of the application of commercial cloud and its ability to support, you know, really edge advantages for the warfighter?

Peter Kunz:

Well, you don’t want to start, I mean, I’ll talk all day about cloud.

No, so I mean, the promise and potential—and we’re starting to see it realized—is around disaggregated data and the robustness that a cloud architecture approach to insights can bring. There’s a flip side to that, which is also cyber threats, the potential increase in apertures, but those have to be managed in order to get that opportunity side.

So, when we look at what we believe to be the most likely or highest probability battle spaces and environments, we’re talking about the robustness of our networks, the robustness of our decision-making data, and how to get it wherever you need it, whenever you need it, will be challenged, right? And those cloud architectures create a mechanism to disaggregate that and provide an element of robustness or redundancy.

Sean Moriarty:

At the same time, it is important to recognize that we will often, in many cases, in the most urgent scenarios, be working in disconnected environments, right? And so one of our design principles is actually to work back from that end user at the tactical edge and start to think about, okay, in a world where this person, this operator, this analyst may only have a laptop or a handheld device, how can you give them most of the power and richness they could enjoy if they connected to the cloud in a world where they are not? And a lot of that comes down to making sure that you’re shipping deltas and information very quickly. If that network connection is dropped, you have the ability to operate, and you have access to these smaller models, which will give you much, in some cases, more power than a large language model.

I should point out that your large language models, in virtually all know at least all known cases, are cloud-based. So much of the promise of that power requires real-time connection. But there are tools and techniques that allow you to preserve much of that power in a disconnected environment, and that really needs to be brought to bear when you’re assessing the capabilities you want—which is, do you have the ability to operate in disconnected environments and retain much of that same power for decision-making?

Panel Moderator: Maj. Gen. Kim Crider, USAF (Ret.):

Yeah, absolutely.

Mark Andress:

Yeah, so on the cloud front, I think we’re going through kind of this generation two of cloud, right? Generation one was all about hyperscale consolidation. So, the value proposition is pulling it all back into these massive data centers. You expand your marketplace, increase security. Like I said, you could dial up a program manager and they could dial up whatever they need.

That was kind of gen one. Gen two is now on us. And this generation two of cloud is being designed, has been designed from the ground up with disconnected operations in mind. And it came a lot out of commercial. So, what you get now is a common set of infrastructure, a common control plane, and the ability—based on, interestingly enough, finance. So, when you go from a billable consumption model, you have to maintain some connectivity.

They’ve set up new finance models that just say, “Hey, look, that piece of cloud that you now have out forward on Oahu or Japan or wherever, it’s paid by the day. We don’t care about your consumption. Set up your control plane, do your big compute in the hyperscale, run it out forward if you need to, if you need to disconnect for a month or six months, no problem, go ahead. But that common control plane allows you to resync once the connectivity has been reestablished.

So, it’s really unlocking what the warfighter needs when we talk about DDill in ways that before, it was kind of hyperscale plus a large USB stick out forward. So that’s an area that you really need to explore.

Panel Moderator: Maj. Gen. Kim Crider, USAF (Ret.):

Yeah, absolutely. Good.

So, we’re coming up at the end of time here. This has been a really meaty discussion here. But so, in a word or two, I’m gonna let each of you, starting with Sean, sort of foot-stomp. What would you have this audience of government and industry folks take away from the conversation we’ve had here today? If they really wanna move to digital at some level—enterprise or more locally—what would you have them take away as something that they’ve really gotta bear in mind? And just give us one or two words on that.

Sean Moriarty:

Yeah, I would say overwhelmingly, I’m a broken record on the need for speed and willful experimentation. And I think the right starting point is to look at the most critical tasks that are going on in your organization and see how much of that can actually be done in software so that people can actually start their days where they actually used to end them, and they can leverage their experience, their knowledge, and their creativity to really solve hard problems and have the machine do much of the pre-work necessary to enable people to do that.

Also, find opportunities where you can gain meaningful advantage where the stakes are low, so to speak, from a risk perspective, particularly life, limb, and money. Organizations are full of slow, unwieldy, yet necessary processes that software can give you massive amounts of time back, and as you get an understanding of the capability, it becomes much easier to figure out how you take that into areas of greater need more quickly.

Panel Moderator: Maj. Gen. Kim Crider, USAF (Ret.):

Thanks. Pete?

Peter Kunz:

I guess I’d say the biggest lesson that we and many others have learned, and often relearned, is that as you’re architecting these systems, it can’t be just top-down. It can’t just be bottoms-up. It has to be both. Right? At the high level, it’s how to connect; at the bottom level, it’s how the work actually is getting done today and what the result has to support. And if you don’t approach it from both, you’re going to end up with a broken solution.

Panel Moderator: Maj. Gen. Kim Crider, USAF (Ret.):

Thanks. Mark, last word.

Mark Andress:

Yeah, I would say the best lesson I have from dealing with a lot of these enterprise digital enterprises is to spend some time deciding what is your enterprise. Literally, what is in your box? Because if you try to put too much in your box, you can get some cost optimization that we’ve talked about, but inevitably you’ll lose agility, you’ll start to slow down. And then if you have a tiny little tactical box, you’re really cool, you go super-fast, you’ve got some great demos, and then you have a hard time scaling.

So put some real thought into what it is within your “enterprise box” that you really want to value, what you want to get out of it. Set that boundary, go as fast as you can inside of it. You’ve got the freedom to move fast, and then focus on the relationship between what’s outside your box—APIs, engineering standards, or whatever—so that you don’t become just a closed-loop cycle there.

---

• This transcript was auto-generated, and may not be 100 percent accurate. The source audio and video can be accessed above.