Long-Range Kill Chains

Watch the Video

---

Read the Transcript

---

Col. Gary E. Roos:

All right. Well, good morning. I am Colonel Gary Roos, and I’m the Senior Materiel Leader of the Adaptivist Weapons Division in the Armament Directorate at Eglin Air Force Base. So obviously I’m over a portfolio of munitions. In my role, I represent the war fighters and the weapons systems that are ultimate consumers of the long-range kill chain, which is what we’re going to talk about today. We are on the receiving end of the systems and the data. Our focus is ensuring that when the weapon’s employed, it is guided by seamless, resilient, and a rapid kill chain. Our success in the battlefield is therefore fundamentally dependent on the innovations of our industry partners like the ones on the stage and the broader command and control and battle management enterprises to deliver that capability. With that context set, let’s turn it over to our distinguished panel. Over to you, Elaine.

Elaine Bitonti:

Good morning. Thank you for having me on the panel. My name is Elaine Bitonti. I lead our connected battlespace and emerging capabilities business at Collins Aerospace, which is a division of RTX. In my business, we are focused on developing the next generation of capabilities that will be needed for future warfare. So we spend a lot of time decomposing the operational problem set and looking at what technologies can solve those gaps.

Col. Scott “Fug” Gilloon, USAF (Ret.):

Good morning, everybody. I’m Scott Gilloon, Sector Vice-President for Air Force Capabilities at General Atomics.

Orlando “OJ” Sanchez Jr.:

Good morning, everybody. OJ Sanchez. I’m with Lockheed Martin, and I lead the Skunk Works.

Col. Gary E. Roos:

Great. Thank you. So we’re going to do a couple of individual questions to get it started. Now, Elaine, we spoke about this before the panel. I know you talked about the challenge of scaling from one kill chain to hundreds. So we’re going to be doing that in a magnitude of difference from what we have been. What is the primary bottleneck that you see, be it technological or procedural, to achieve this? And how can we incrementally demonstrate and help us to overcome this safely?

Elaine Bitonti:

I think there are both technological and procedural challenges. I think first from a technical perspective, there’s really two things that we think about and are focused on. One is when you’re going to have that type of, I’ll say, scale of event, how are we going to get all of the data from the different shooters, the different sensors, and how are we going to quickly process that into decision-making? We’re going to have to be doing that on a scale that’s never been done before, and so that’s really a difficult problem.

The next problem set is around the connectivity that you need to communicate those decisions, particularly when we’re talking about this level of engagement over hundreds of nautical miles. How do we have resilient connectivity that will support that? And so I think from a technical perspective, those are challenges that we’ve previously, I would say, not dealt with in our prior conflicts. We’ve had more single kill chains and single target engagements that we were looking to do, and now we’re going to be scaling that to the hundreds.

And so there’s been a lot of work done over the past few years, particularly in the Air Force PEO C3BM, with a lot of infrastructure and connectivity investments that I think we can leverage, but there’s still continued work to be done from a technological perspective there. And then I think just procedurally, how are we going to think about when we need to use new technologies such as AI and other things that will help us process that amount of data in the time it needs to be done, how are we going to do that with a human on the loop? There’s a lot of new TTPs and other things that will have to be looked at as we consider those changes.

Col. Gary E. Roos:

Great. Thanks, Elaine. Scott, so we talked about the importance of kill chain resiliency now. So for the long-range engagements, when the time to target is very significant, what are the primary threats to that resiliency, and what are the key strategies that you can ensure our kill change withstand the adversary’s interferences?

Col. Scott “Fug” Gilloon, USAF (Ret.):

So I think a couple of points. First, I would like to thank everybody for making the trek all the way down to Colorado B. I think it just shows the importance, just the number of people in the room, the diverse array of folks in the backgrounds. So when we look at long-range kill chains, I think we’ll start by just addressing the US and where we sit. So for the US, we’re not alone in our ability to press a button on one side of the planet and deliver effect on the other side of the globe. What does set us apart though is our ability to close the long-range kill chain. I think that’s important because in this case, it actually provides for potent deterrence, and it also provides for a significant operational advantage.

Now, if we took 10 people in the room and asked them to define long-range kill chain, I think we’d probably get 10 different answers. If we asked 10 weapons officers, there’d be 10 it depends, and then there’d be 10 answers. But within each of those answers, I think what we find are three foundational elements to each one of those kill chains, sensors, effectors, and command and control. And I think in a lot of ways, the panel you see here, and it probably touches every person in this room in some way, shape or form, is that array of sensors, effectors, and command and control.

So to ensure deterrence though and provide that operational advantage, those sensors, those effectors, that command and control apparatus, it has to be resilient, it has to be redundant, and I think probably above all else, it’s got to be free from single-point failures. A lot of times when we talk long-range kill chains, I think we hear about Indo-Pacific scenarios, we tend to focus there, or we tend to focus on space-dominant architectures. What I think of when looking at that is I think we need to take a little bit more complete perspective. We need to consider other applications. For example, in homeland defense like Golden Dome, we need to consider the Arctic approaches, because if we just limit our thinking, then we’re going to miss the synergies across the theaters. And I do believe that’s really important.

So if we take that different perspective, it helps us understand how diverse sensing layers, it helps us understand how more responsive effectors, it helps us understand how a more distributed command and control network are going to enable us to close those kill chains over vast distances and/or against sophisticated adversaries. And I think that’s probably why everybody’s here, but I think we’ll probably get into that in the next half hour or so.

Col. Gary E. Roos:

Thanks, Scott. We’ve come a long way. I remember a few years ago, long-range kill chain was just a lightning bolt on all of our posters. As magic happened, we were happy it happened, but we’ve come a long way since then. OJ, so we talked about prior to this discussion that the threat gets a vote, that our kill chain grows in complexity, and we are creating new vulnerabilities as it continues to grow. How must our kill chain design evolve to outpace adversarials’ way that they’re adapting to exploit?

Orlando “OJ” Sanchez Jr.:

And thanks, again, I’d echo. Thanks, Colonel Roos, all the prep and everyone for being here. It’s an honor to be here talking about this. Certainly I resonate with Elaine’s comment about data. I think first and foremost, the ability to realize that the data is going to matter is an important part of this. And then I also think there’s probably 10 weapons officers in here and 10 different opinions about how to solve the problem. But I think when we talk about the threat gets a vote, for those of you who have ever been to weapons school or think about weapons school, fundamentally, we always say maneuver in relation to the bandit. And we have to realize that while we’re having the conversations about how to leverage data, how to use the sensors, the effectors, and do it in a C2 environment that is resilient, the threat’s doing the same thing.

And so we have to be constantly thinking about how can we evolve. I agree with your overarching premise there. That means that the resiliency we’re looking for, it’s not a stand in or stand out. It’s not an air or space problem. These are and problems. And as the threat evolves, we need to build that resiliency into our thinking so that like weapons officers, we can say, “It depends, and we have a solution as they evolve towards that.” So I think to do that, largely part of the problem is more cultural than it is technological. There are some hard technological problems here, but some of the cultural problems, we have to go back to that basic premise, maneuver relation of the threat, understand what we’re trying to solve. It’s just a different problem.

So as opposed to thinking through siloed, platform centric or data centric, we need to think, what problem are we trying to solve? Pull cross-functional thinkers into that problem set and then design to that. And that’s a bit of a cultural issue. The Air Force is certainly taking it on through acquisition reform. Industry is leaning in. We need to lean in on how to move towards government reference architectures, open standards, designing with adaptability in mind. But at the fundamental element of it, because the threat’s going to change, we need to think at the outset, how do we get around the problem and then build teams who can adapt to that problem, regardless of the technology changes.

Col. Gary E. Roos:

Wow, that’s great. Thanks, OJ. So we wanted to talk a little bit about, we’re talking about long-range kill chain, the problem and how we’re fixing it. I think, Scott and Elaine, you guys both address different parts of the problem. So I wanted to pick at that a little bit. So Scott, you talked about the focus on ensuring the resiliency of the kill chain, which is making sure we have a exquisite solution for a problem to make sure that is resilient and redundant. And Elaine, you talked about scaling, from going to hundreds for concurrent engagements, which is a different problem set. However, as we’re solving these two problems together across the enterprise, do these two goals, do they create tension? How do we design an architecture that will achieve both the resiliency and the mass that are required for the future conflict?

Col. Scott “Fug” Gilloon, USAF (Ret.):

I’ll go first on that one. So I think when you have a resilient and a redundant kill chain, I think it scales its sensors, its effectors, its command and control apparatus. It scales those with and/or to deliver the effect, so it’s not something that uniquely needs to change if you’re doing it right. However, I do think that from an industry and a government perspective, there are some things that we have to be careful of. We have to be cognizant when we’re doing single-source manufacturing for anything that’s going to touch that long-range kill chain. We have to watch for fragile parts and our fragile elements, if you will, in our supply systems that are going to affect our ability to sense or to strike or really to manage those data flows because it affects software as well. We’re talking about a software-enabled transport layer.

But I think one of the things that is probably not as visible to the audience that we spend a lot of time on, on the industry side, is spending our money to find ways to achieve a more resilient kill chain. So that’s not something that’s talked about a lot is how does industry spend their money. I can tell you, we spend millions, I can speak for General Atomics and I know we spend tens of millions of dollars every year flying and evaluating these capabilities. From Lockheed, from Collins, from multiple other vendors, it’s hardware, that’s software, but we are also taking them, combining them with autonomy stacks and skills from all the vendors, at least most anyway that you’re going to find on the floor here. And we do that so that we understand where those single points of failure may be.

I think it’s important though, again, to stress that that’s industry working together. And I think it’s something to highlight here is where you don’t see, and I think Elaine would agree and probably OJ as well, that it’s just not something you see on the outside as much or not something that’s visible, but it’s happening, and we’re spending a lot of money to try and find those ways so that we can present them to the government as, “Hey, here’s an option,” or “Here’s a solution,” or “We’ve looked at where those single point of failure are, and here are some paths around it or to mitigate.”

Elaine Bitonti:

I think the other thing, and we, at least the engineers that support me talk to me about this a lot, as we naturally scale up with what we have to do, from a communications perspective and from just a processing perspective, you do ultimately meet challenges of physics. You can only have so much of that. And so there are real, I’ll say, technical constraints. I do think there’s a lot to be learned from the commercial sector on types of scaling that they have done for their communications networks across the world that we can leverage from a long-range kill chain perspective. Additionally, I think there’s been a lot of good work and technology that the Air Force and Joint Force are using today when we think about things like Starlink that came from more of a commercial background. They can help us overcome some of the previous limits we had, and they can help us have multiple pathways when we think about that resilience.

I also think OJ’s point on government reference architectures, there’s been a lot of work done there. I spent my entire career on the defense side, and from when I started to where we are now, the fact that we even have government reference architectures that everyone knows the standards to, we can all develop to, we can have some of that interchangeability that Scott talked about in autonomy stacks, in comm stacks, in processing stacks. We can have all of that and support the scaling if we have a common standard to do it to. And then, I would say from an industry perspective, we can focus our investment on those most pressing joint force problems. So I think there is natural tension in the system, but I think there’s also things we can do from an acquisition perspective, from a technology and architectural perspective that can help us overcome those things and make sure that we have the long-range kill chains we need to support what the Air Force needs to do in future conflicts.

Col. Gary E. Roos:

Great. Thanks, Elaine.

Orlando “OJ” Sanchez Jr.:

And maybe I’d add on that. I agree. We try to be practical, like how can we move forward? Because the reality, just like with physical systems, this long-range kill chain is going to be a mix of things we already have and new things in the future, all built on different approaches, depending on when they were embedded. So I think part of what we need to do is we also need to recognize that doing something now is better than doing nothing, waiting for a perfect solution. So working backwards to try to solve problems that are going to exist for a while is an important part. That takes collaboration because it usually means it’s not a program of record. There’s some kind of industry investment coupled with some kind of willing program office to go burn down what I would say, the minor miracles that are going to occur on an existing system or bring new tech.

So there’s a certain amount of collaboration now. An example I would use, last year we worked with the F-35 program and the Netherlands, the Royal Air Force out of the Netherlands, taking highly classified F-35 data off a foreign system, moving it through a multi-level security, taking it up through Starlink Starshield and then into a foreign C2 system. Is that long-range kill chain practice? It absolutely is. And we burned off that miracle, if you will, so that a future program doesn’t have to do that. So I think some of what we need to do is accept that we’re not going to create a perfect, resilient, adaptive system in a modern way all at once. So we got to try now to start burning these things down. And it is a collaboration game. This one’s a team sport. There’s not going to be the perfect RFI written to do this. Industry won’t spend their money perfectly. So everywhere where we can practice and collaborate to try to do those, we’re going to be better off to solve these really hard problems.

Col. Gary E. Roos:

Great. Thanks, OJ. So I have an analogy, so just follow with me here. When you’re traveling across the country and you got to stop at a gas station, you show up, and you know a couple of things are going to happen. One, the architecture of the fuel pumps are going to work for your car. If it’s unleaded and you know the field’s going to work for the combustion degree car as well. That architecture has already been done years ago where you can drive across country with full knowledge and confidence that you’ll be able to gas up.

Long-range kill chain could be something similar. I know right now modular open system architecture is a really big deal in the government and specifically with Arm Directorate, WOSA, Weapon Open System Architecture, is a really big deal for us. As far as the open system architecture is going, is there anything you think we should be doing differently, or is there anything you think the government should be doing differently to help out with that architecture, where if my munitions or my aircraft flying into an area, I end up with confidence I’ll be able to get my gas up, I’ll be able to get the kill chain required for me to be successful?

Orlando “OJ” Sanchez Jr.:

I’m glad to start. So completely agree. So everything about the future needs to be open. There needs to be standards, and we all need to be able to understand and plug in. So what the government can do, I think, is continue to convene and pull in the players and industry who are willing to do that. And I think it should be a red flag if you’re working with industry, whoever that is, and they’re not willing to sit at that table and make compromises because reference architectures means somebody has to compromise. We’re all going to be better if we compromise. We’ve seen that working with the Air Force Rapid Capabilities Office for decades working through CMCC and others, and it’s great to see that that’s proliferating.

So I think the government creating those forums, making a big tent, making not a sequential, but a parallel effort, we’re all working on it, would be a big help. And then I think everything going forward, there’s so many efforts that get started. We can’t let anything in the future slip through without that as a design criteria. So those are things that the government can’t help with. There’s just so much activity across the DIB. Convene the boards, keep everyone involved, and then make sure that everything new that’s going forward has that hooked into it so that we don’t design to a particular niche that a particular program or problem set might require.

Col. Gary E. Roos:

Thanks, OJ.

Elaine Bitonti:

The other thing I would add to that, I think there has been a lot of progress on open standards and government reference architecture. The one area where there has been progress and I think there’s still more to be done that the government can help with is particularly from industry, we work in several of our areas. We work across all the services. And so we do still see many times that different services are trying to solve the same or very similar problem in different ways. And so I think continuing to have more cross-service collaboration when you’re looking at the government reference architectures, particularly for weapons, for comms, for things where C2, there’s a lot of commonality that helps also industry understand that when you have to invest three times for three different services, that’s a different capability than if there was more commonality, an industry could be investing once to serve two or three of the services.

So I think there’s been progress there, but there’s still more that could be done. To the points about large scale exercises, we find all the time when we do them in REFORPAC, particularly on the networking side, every time you go to set up a bespoke network, there is what I’ll call NRE. And I’m not talking about from a cost perspective, I’m talking about from a technical perspective of how you actually facilitate that network, communicating everything it needs to communicate to who it needs to communicate. And so I think there’s been really good focus in certain areas around processing autonomy where we have really strong GRA. I also think there’s maybe at the different networking layers, that’s maybe an underserved area that the government could also start to look at that would be really beneficial for long-range kill chains and for interaction with our allies and partners.

Col. Scott “Fug” Gilloon, USAF (Ret.):

So I think I’ll take a little bit of a different aspect on this, and it’s something that it affects the kill chains now. As Elaine mentioned, when we start talking about, I’ll call it intelligence software, every aspect of software is going to impact the kill chain. Whether you like it or think it does or not, it just exists and it’s there. And so what we don’t often talk about is how the GRA, if you will, would extend into what I’ll call data governance. So when you think about a long-range kill chain, think about all the data that goes into that. So to affect all the ones and zeros that are moving through that network, I’ll just call it from a global perspective, from the United States at least anyway, we don’t have a really unified data governance model.

And I think that’s going to be a challenge. When you talk about threats and challenges, I think that’s going to be a significant challenge. We see it in autonomy. We have an autonomy government reference architecture and government reference implementation, and I give the agile development office high marks for that. For those who aren’t involved, it’s a pretty significant effort, but it’s delivering results and it’s allowed us to move forward in a unified way. But in the broader sense, those autonomous things, everything, every piece, every sensor, every widget that’s out there is collecting data, and how you manage that data matters. For those who have Teslas, if you’ve ever checked your wifi, watch when you come home and see how much data that vehicle is downloading. It’s a fun exercise, but I will tell you that there’s a governance model for it. Now think about that.

Every Tesla that drives everywhere has the same sensor package, so there’s your GRA. Same sensors, same hardware, different shell, but it’s collecting data, and it’s sending data to someone who knows how to use it. And I think that’s where we, industry anyway, we’ll turn to the audience and I’ll say, “How are you going to use the data?” When we look at how we spend or how we make different, I’ll call them hardware or software solutions, I think the question and the feedback we need from the community, the war fighters, all of you folks out there in uniform, is we need to know how do we manage that data? How do we govern the data? What’s going to give you the results at the end of the day to affect the kill chain? But that’s something that’s important. I think certainly it’s, I’ll call it maybe an unrecognized threat in some circles, but it’s very much real, and it’s something that we’re going to have to deal with.

Col. Gary E. Roos:

Great. Thanks, Scott. All right. I have a question about the commercial sector now. So for each of you, do you have a concrete example of a commercial innovation that you could think it could be a game changer for something that would make our kill chain faster, more lethal, more resilient? And we’ll start with you, Elaine.

Elaine Bitonti:

I think the one I would talk about, I’m very excited about it. It’s still nascent, but when we think about the application of AI, particularly to resilient connectivity, one of the biggest challenges with a lot of the communication systems today is you go, you deploy the communication system, and once you deploy it a number of times, it can become susceptible to things that the enemy would like to do to it. And think about in the future, if we had, I’ll say, AI-enabled resilient connectivity where you could actually be generating new waveforms based on what was happening to you, so that you could be doing that on the fly. You don’t have to go back and have, I’ll say, engineers do that in the lab and take time. It would be actively responding to the environment around you.

So I think that, as we look and go forward and how do we take a lot of the things that have been done in the AI space for commercial use and apply them to our problem here for long-range kill chain, I think that’s one of the most exciting examples that could very much be a game changer to how we think about connectivity today.

Col. Gary E. Roos:

Thanks. So Scott.

Col. Scott “Fug” Gilloon, USAF (Ret.):

Example, so I think there’s probably a bunch of people in the audience who would think large language models. Software, and we’d say, “Give me ChatGPT or whatever for my work function.” And certainly there are going to be some applications there. And so I’ll take maybe a more boring approach, but something different. And one thing that I think is underlying that framework, if you will, that allows or enables folks to scour the internet or search across the globe, no matter where you are, are things like laser comms. So when you look at, we talk about secure communications, we talk about a way to operate in contested or degraded environments, think about laser communications. There are commercial providers who are leveraging that type of capability today. How many of your platforms have it? It’s really difficult for someone to actually intercept laser communications. The commercial world is experimenting with even geekier way of doing it with free space optical comms.

So I guess in your free time, maybe you can Google that one or ask ChatGPT, but there’s significant capabilities, and they’re going to come from the commercial sector. That’s where they’re going to innovate and they’re going to find ways. And so I’d turn again to the war fighters and say, “Where does that make the most bang for the buck in the kill chain?” It’s a hardware, it’s a widget, it’s a thing. From an industry perspective, when you come to us and you say, “Hey, what can I get? ” we can integrate most anything, I think. And with open standards, it’s allowed that to become much faster and lower cost. But I’d turn back again to you and say, “Hey, with these really cool commercial applications, where is it really going to make the difference?”

And that’s the thing that from an industry side, we don’t see this kill chain maybe the the exact same way as you do, because you’re out there doing it every day. And so it’s that feedback loop, I think, that’s going to help us. There’s hundreds of billions of different widgets and things, but that’s, to me at least, the really important part.

Col. Gary E. Roos:

OJ?

Orlando “OJ” Sanchez Jr.:

It’s a good question. Maybe I’ll take a slightly different tact. There are a lot of commercial applications, but one thing commercial does well, because they’re in a competitive market, multiple, all the dynamics of their user base always changing, they focus on changing the how, not just the what. Breakthroughs happen, we love using them, but behind the scenes, the way they did it changed in order to give them a competitive advantage in the marketplace. So there’s a few things that we can lift from that methodology, and I’ll just use a couple of examples you all did. So in the AI space, for example, we’re all working with large language models, but in the commercial sector, they’ve now moved from working on CPUs to super massive approaches on GPUs. You can go from looking at large data problems that took hours in a super massive approach that would have taken us years.

So we’re now applying that, and we need to do more of that in the defense side. So we’re working on that, for instance, with Vista, with the X-62 VISTA, we’re doing GPU-based AI. That’s a how, not a what, and that will enable us to do it differently. And then also the focus, we all have to acknowledge even if the budget were to go to 1.5 trillion, which is amazing, that’s an incredible amount of money, I don’t think I could process that, there are always top-end budget challenges. So we have to think about how commercial does affordability different. A lot of that has to do in producibility. So at the beginning of a problem set, designing for producibility, whatever that is, whether it’s a software-based stack or whether it’s a hardware-based stack, there are some real practices in the commercial space that we can lift the how and experiment with those so that everything we go forward has an affordability aspect to it that we didn’t have before.

So those are the things that I look for. And in addition to novel approaches to technology, in Skunk Works, I look for how people are doing things differently and then how can we apply it to defense problems. And I think that’s an important attribute here for long-range kill chains.

Col. Gary E. Roos:

That’s a great distinction to make. We just talked a lot about machine language learning and artificial intelligence. So I wanted to pick on that a little bit on the human machine balance between the two. So as the machine learning, we’re going to be able to adopt faster. We’re starting to get them onto networks where can do other things where we haven’t been able to do in the past. But we still want a human in the loop. That seems to be a desire that we still want to have. So where do you strike the right balance between the machine speed and the human oversight to be able to do things that long-range kill chain are going to be demanding in the future? We’ll start with Scott this time.

Col. Scott “Fug” Gilloon, USAF (Ret.):

All right. So we experiment a lot with human machine teaming, both in terms of, I’ll call it crewed-uncrewed teaming, and we take uncrewed-uncrewed teaming. And so what we see from just a, I’ll call it, I’ll start at the airframe level, it’s the open architectures that Elaine referenced. And then just overcoming the ability to put two widgets together and/or generate data from those and the usability of that data. So that’s the starting point. We spend a lot of our time integrating those because as it turns out, integration can be hard. Making things work together in spite of all the open architectures takes a little bit of skill. And so the challenge I think is one, when you look at that, maybe take a step back, the challenge is one of, do we have the right level of patience with what we’re asking?

And so we’re asking to create, pick your framework or pick your function between, and again, at our level, let’s say crewed-uncrewed teaming. Do we have the right frame of reference, or do we have the right level of patience as we begin to integrate and iterate down this path? And so in a very real sense, it’s what you’re seeing with CCA. I don’t know how close many of you are to that particular program, but just the other day was the first time we were able to take a CCA and integrate with an F-22. That’s a pretty big deal. I think it’s high marks to the Air Force for enabling that. We’re just a couple of years into a program, and we’re already connecting those. That wasn’t by accident. That’s a lot of smart people figuring out how to leverage existing things that OJ talked about to make that teaming work.

So from a challenge, it took a lot of patience. It took a lot of smart engineers to go back and figure out how to move data across networks that were never designed to move the types of data that we’re moving. But I will point back though, but it’s the ability to have that patient thought and go, “Hey, we’re going to proceed down this path.” And to take that risk, as OJ said, it’s a calculated risk, but a measured risk in order to deliver the effect back to the war fighter.

Col. Gary E. Roos:

Thanks, Scott. OJ, on the human machine balance.

Orlando “OJ” Sanchez Jr.:

And thanks for reminding me of that. That was great. Another example of open teaming, Lockheed working with GA, connecting two products, that’s the kind of collaboration we need in experimentation. I take it, and again, I think I’m adding the human element to this one too. So if I think about, and I’ll use the Raptor as an example, when Lockheed early designed the Raptor with the F-22 program office, we thought we needed to go high and fast. We had all this OA that said, “You’re going to use supercruise. We’re going to fly at 55,000 feet at 1.6.” You know how many times we fly at 1.6 at supercruise on the Raptor? Are there any Raptor guys or ladies in here? Not often because once you gave it to the war fighters, they figured out that there was actually a better way to apply the technology that was relevant.

In this space with human machine teaming, I think we can’t overthink how we’re going to get there. So ultimately the war fighters are actually going to adapt, and they’re going to use these technologies in a different way than we’re thinking up on this stage or than the folks that are here at AFA, and getting it to them sooner is probably the better approach. What we can do as an industry side, I do know what’s going to happen is there’s a huge element of trust involved here. Just like in any other team, we use the word team, but we’re not talking about how do you build trust between these teammates. A human who has to trust a machine in order to do critical tasks in a dynamic environment, it doesn’t matter if we can drive the technology. It could take us decades if we don’t start working on the trust problem.

So however we get after the trust practice, I think there’s going to be a balance of human on the loop while we build trust, and then the war fighters are going to figure out, “Hey, I trust this teammate, this machine teammate. I’ll offload more capability based on the problem I’m facing.” So right now, our approach, I think we have to build in, don’t shoot for the moon on the first approach. If we build a system with you all that completely keeps the human off the loop because we can, it’s not going to get used. Not just in the Air Force, in other systems, you’re going to have ROE problems, et cetera.

So we have to build in that reality of build the ability for the human to stay on the loop and then let the war fighters figure out how much they’re going to come on or off the loop over time. I think that’s going to be the most successful approach to the application of this, whether it’s in the air vehicle piece that we’re working on in CCA or in the command and control piece of how much do I offload to an AI agent or otherwise. We got to design it now with human on the loop with the ability to be able to dial it and then let the war fighters figure that out. So I think that’s going to be the most effective approach.

Elaine Bitonti:

I think I’d add two things to that. One, I think a lot of what we look at with where we’re spending our AI investment is there is so much, I’m going to use this word, toil. There’s so much toil in even operating a long-range kill chain, doing the things you have to do, moving things in spreadsheets. How do we use AI in these things to reduce the toil so that we’re presenting much better quality of data for humans to make decisions on the thing that matters? We don’t need humans engaged in all of the toil that currently has to happen today for us to do these things. And so I think that’s one part of the balance we can look at is let’s eliminate all of that and have much better data presented so humans can make decisions on what matter.

And then the second thing picks up a little bit off what OJ was saying around trust. It’s also a new and different way of working. We’ve been doing a lot with AI-assisted coding for our software engineers, and what we found is it can be very, very beneficial. We can increase productivity significantly, but it’s a brand new way of working. The way that you wrote code when you didn’t have these AI-assisted tools is not the same way that you’re going to do it now. And so in addition to building trust, you also have to build new ways of working and new ways of thinking about things when you’re introducing these new items. And I think that’s been one of our lessons learned is we didn’t spend enough time contemplating what that means and the change to the way people work to fully take advantage of the opportunity that the technology offers.

Col. Gary E. Roos:

Great. Thank you, Elaine. So finally, considering all the points that we discussed here today on the kill chain on scaling, autonomy, resilience, and the adaptive threats we’re going to be seeing, to close, what do you think would be the single biggest cultural organizational barrier within the industry in the Air Force that we might overcome to realize the kill chain of the future? We’ll start with OJ this time.

Orlando “OJ” Sanchez Jr.:

Sure. I’m glad to start. So what I would say as we think about these barriers, this is a team sport. And in many ways, all of us have contributed to narrowing our teams over time. So as we’ve evolved in our war fighting construct, we have a pretty program centric or siloed approach to developing capability. Everyone in this room, you hear it in every conference, is trying to break that down, but if we’re pragmatic about it, we have a long way to go. And so if we want to actually solve this long-range kill chain problem, we’re building the weapons. We’re building the systems that can persist or sit outside. We’re building the C2 architectures, but we are going to have to work on breaking down these silos, pull people into a problem-first approach. So use OA in an integrated way, look at that problem and then pull cross-functional, cross-program, cross-company teams together to look at that problem.

And I think if we work on that more, we’ll bring to life a bit more of the acquisition reform and a bit more of the desire to get the best out of industry faster to this specific problem, which is pretty hard. Closing this long-range kill train with resiliency at volume, at place and choosing, that’s a hard problem. So that’s what I would say is probably the biggest practical thing we can do.

Col. Gary E. Roos:

Awesome. Thanks.

Elaine Bitonti:

I would say I think really two things. One is today, as OJ noted, we primarily develop, require, and acquire systems from a service level, but when we will fight a war, it will be at a joint level. And so I think from the government perspective, that’s something that we need to continue with the cross-functional teams and the joint view on things as we continue to develop things in the service because the way that the money is allocated, and that’s where the priority goes. And so I think we need to make sure that we’re also allocating money on what are the joint problems that need to be solved and how do those actually get effectuated.

The second thing I would say is there’s been a lot of talk and a lot of really good progress around how do we get MVP capability out to the war fighter, get it in their hands. It doesn’t need to be perfect, let them use it and evolve. And I’d say that’s gotten a lot of traction, and our contracting system, our requirement system and our test system is still not really caught up to that way of working. And so I think we are making progress there. Acquisition reform has started to touch on that, but there’s still a long way to go between being actually able to get that MBP fielded in the time needed and not have a test program that takes two to five years and not have contractual requirements that aren’t really moving the needle for the war fighter, but industry has to satisfy them. And so I think there’s continued work to be done there to get after the speed element and getting things into the hands of the war fighters as fast as possible because they’re the best judge of what’s going to be useful.

Col. Scott “Fug” Gilloon, USAF (Ret.):

All right. So I’ll try and wrap up my best in 60 seconds or less, I suppose. Circling back to the whole question at hand, long-range kill chains. In my opinion, they’re about building dilemmas for your adversary, and that’s right through a mix of diverse sensing layers. So that’s not putting all your eggs in one basket. That’s the single point of failure that we talk about. So when we look at things like global communications, they’re great, but we put a lot in the low earth orbit. We have dominance there today, but the adversary probably knows it. And so we’re not building a specific dilemma the more that we pile into those types of layers. So it’s diversifying both your sensing layer and making what I’ll say is responsive effectors. So responsive effectors, meaning you’re listening to the combatant commanders and what they say they need for a weapon or other effect that needs to be generated.

And then the last part of that, from a command and control apparatus, we’ve got to have distributed command and control. The things we’ve talked about up here, we talked about a software-enabled transport layer, all of that’s great, but it has to be distributed, has to be resilient. That’s the only way I think we’re going to pose dilemmas is by having resiliency and redundancy in our kill chain. And if you’re not seeing that and you’re finding the single point of failures, be honest. I think that’s something from the industry perspective we’d ask is be honest with us. Tell us what the problem is. Give us a chance to go use our money to go solve it and we’ll do that. That’s why we exist. So I’d put that pitch to you guys.

Col. Gary E. Roos:

Fantastic. Thank you. So it is clear to me we’re going to need relentless teaming to be able to get this done together from Air Force industry, within industry and within the joint services. So we are definitely looking forward for future partnerships and continue to get this serious challenge done. So thank you to the AFA and thank you for everybody in attendance today. We appreciate you coming today.