Terry Gerton The Aerospace Industries Association has a new white paper out talking about agentic AI, that it is this solution that is going to speed up decisions across the defense industrial base. When you think about all the issues that DoD has in terms of procurement and industrial base, understanding supply chains, how confident are you that AI is going to solve the root problem — and that it’s not something else like organization or culture or just legacy processes?
Tim White Well, I think you’re highlighting there one of the most important characteristics of what we’re really working on, and that is that we have a complex system that we’re all working in. And what I believe, and what I think we’re seeing now, is the fact that AI is an enabler. It is an assistant. But it’s not going to solve everything that is already out there. We do need to work on culture and we need to work on process and we need to work on workforce. All of those things are with us right now. But I can foresee that AI is really going to make a significant difference in the way that each one of those problems is being addressed. So not the end solution, not the only solution, but certainly a powerful enabler.
]]>
Terry Gerton Let me lay in a real-world test that’s playing out in public, kind of as we speak, and that’s the Pentagon openly clashing with Anthropic over how far military use of its AI model should go. DoD is pushing for access to Claude for all lawful purposes. Anthropic is drawing a hard line against fully autonomous weapons and mass surveillance uses. I don’t want to ask you to comment on that disagreement specifically, but given the fact that this is playing out in public, how do you think it shapes the path for agentic AI in defense?
Tim White I do think that, in general, every company that operates in the public space and most certainly with the federal government and the Pentagon needs to be aware of what those uses might be, and we need to be explicit about that. I think many of the people who are in this space right now are very comfortable with using it in the right ways, and it really just hammers home the importance of governance in AI. And really that governance is on both sides of the procurement organization as well as the providing organization. It says, are we all on the same page? And that is one of the areas that we really hit hard in our white paper, is to make sure — are we on the right page? Are we on this same page and do we agree on how this is to be used and what are the guardrails?
Terry Gerton Governance is a word that’s easy to say and sometimes hard to do. What would an appropriate governance structure look like here?
Tim White We actually about a year and a half ago released a paper on governance, AI governance in aerospace. And it walks through all of those different elements that you would need to have that would be effective governance. Because, really keep in mind, what you’re trying to do is to make sure that there is appropriate and safe usage of whatever tool it is. So you would think about including things like use case libraries, policy. You would include the ability to revise documents over time and revise expectations as new capabilities and new challenges arise. All of that really plays into, how do you govern this thing? And keep in mind, governance is very much a living process and a living capability. You know, agents are new, I would argue. It has changed the way that people are looking at governance. Used to just be that a year ago, you’d be looking at large language models. And now you’ve got agents that are actually taking action on these outputs from large language models. Well, that changes the equation and therefore it changes the governance.
Terry Gerton Agentic AI, as you just described, is a system that actually takes action, often without human prompting. What safeguards would you want to make sure that a governance structure puts in place so that the AI doesn’t kind of run off the rails?
Tim White First of all, you would want to make sure that you have the right process, but more than anything, I would think about policy. People, as they’re looking to adopt any AI, any capability that is new or emergent, they want to know what the guardrails are. And it’s incumbent upon leadership and experts to come together and say, we understand what the risks are, we understand what the possibilities are here, and let’s provide our community with the answers around what is appropriate usage — with the understanding that you’re going to have to have monitoring processes in place, you need to have training, workforce is a huge part of any AI roll out and most certainly agentic AI roll out. You want to make sure that all of these things that you think about as you get new tools are there in place as you’re rolling out. Not necessarily kind of emerging as you’re finding things out. It needs to be intentional. And I would hammer home the expectation of leadership engagement. This is the heart of what leaders do, is to bring their people together, enable them, provide them with new tools, but also provide them with guidelines and boundaries.
Terry Gerton I’m speaking with Tim White. He’s the vice president of engineering and technology at the Aerospace Industries Association. Tim, let’s pull that thread just a little bit further because we talk about governance. We talk about guardrails. But if agentic AI actually does get embedded in decision-making around supply chain or contracting or production decisions, where is the accountability if the AI makes a bad decision?
]]>
Tim White My sense is that accountability currently and in the future resides with leadership, because leaders are really the ones who are procuring the models. They’re providing the models, they own the policy, and then there is the expectation that the workforce as well is using these things appropriately, they’re not going outside the guardrails, they’re elevating potential concerns, right? There’s definitely a partnership there. But keep in mind that as we talk about agentic AI, there are different ways that it can operate. One way is that you’ve got a human in the loop. So the agentic AI will come in, it will take your prompt, it will provide a proposed action, and it asks you, “do you want to go and do this?” And so in that way, it’s a very clear responsibility that the user has to say, “yeah, actually, I’m putting my signature, my imprimatur, on this action,” right? The other way that it can operate is that you can tell it just go. There’s risk there. And with good governance, you would look at those risks and say, do we really want to do this? And I would argue that in some use cases, in nonconsequential nonassurance processes, maybe that’s okay. But in many cases — and I would be very adamant about this, and speaking about the aviation industry, the aerospace and defense industry, we deliver assurance, we deliver surety — and so we almost always are going to want to have a human in the loop and have that accountability that you asked for.
Terry Gerton Well, let’s talk about those humans. You’ve mentioned the workforce and the role that they play in this process. But DoD and its contractors consistently struggle with training. DoD systems themselves have trouble with data hygiene, uneven digital infrastructure. What is the practical path toward deploying something this advanced inside organizations that might not be ready for it?
Tim White I would answer that in two ways. And I think that you’re right, that there are always ongoing challenges with training. But then I would flip it to the other side and say one of the great advantages of AI is that it’s a pattern recognition capability. And what that leads to is the opportunity to use it as a tool to help people when they need help. So you might think about someone who comes in and they’re working through a new process that maybe they got trained on three months ago and now lo and behold, they need to go and execute it and they don’t remember everything. AI would be able to recognize, this is in effect a new user, let me jump into the process and provide some feedback, provide some immediate, relevant help. I think absolutely that as you think about the traditional agentic mindset of, let’s just go have the computer do things. There’s risk there, but I would argue that these challenges that we have, there’s actually tremendous opportunity within the AI to help us meet those challenges and address the workforce training issues, the workforce capability issues, to be able to accelerate and deliver much higher quality solutions with the workforce that we’ve got.
Terry Gerton So you’re making a strong, positive case for the roll out of agentic AI, and it may be something that we can’t even control, it’s going to be here whether we necessarily want it or not. But what does a responsible rollout strategy look like over the next three to five years? How do you see agentic AI actually being incorporated into, say, DoD’s workflows, and how will we see it play out?
Tim White I would argue that agentic AI is a tool in the tool set. And what that means is that you must begin with the end in mind. What is the end result that we’re seeking here? Is it faster procurement? Is it greater quality? All of those things can be addressed through many different methods, right? Six Sigma and technology and agentic AI, those are all options. So I would never suggest that agentic AI is the beginning of an improvement process. Rather, we would look at, what are the challenges that we have? What are the tools available that can address those challenges? Agentic AI is probably one of those that would enable us to, as an example, go faster, be more accurate, deliver higher quality. And if appropriate, then you move into the next-stage design and you would say, what is our policy that we need to implement? What is appropriate use of the agentic AI to address this problem that we’ve got? And then you move into training and eventually you’re in steady state and you’re confirming, you’re monitoring, you’re making sure that you’re delivering what you promise to deliver. And that’s really where it plays in the overall lifecycle of improvement.
Copyright
© 2026 Federal News Network. All rights reserved. This website is not intended for users located within the European Economic Area.