Gaining a cognitive advantage for Special Forces: A conversation with Bill Wall, Accrete AI Government CEO & Co-founder
StoryAugust 10, 2026
There is a pressing need within not just the special operations community but also in all the U.S. military services and government to understand and identify the narrative messages U.S. adversaries are using against us, says Bill Wall, CEO and co-founder of Accrete AI Government, told me during a SOF Week Show Daily version of the McHale Report podcast. He and I also discussed ways the U.S. military can achieve information dominance, how artificial intelligence (AI) enables a cognitive advantage, and Accrete AI’s knowledge engine platform. We also talked about what had him most excited about SOF Week. Edited excerpts follow.
McHALE: Bill, can you provide a brief description of your role in Accrete AI Government, your experience in the defense industry, and the importance of special operations to what you and your company do?
WALL: I started Accrete AI Government with two other partners about seven years ago, as a subsidiary of a commercial artificial intelligence (AI) company that was starting in New York City in the financial markets. Our thesis was that the artificial intelligence and the emerging capabilities that Accrete was developing to help financial analysts – understand fast-moving markets, understand the search to find alpha in the markets, [gain] first mover advantage – could be ported over and brought to bear for the U.S. government and the U.S. military in particular.
To me personally, [special operations] is important because I served in the community and my friends and peers are still there. I understand how important the mission is, how important technology can be to increase mission success and survivability of the organizations out there, and to super enable the constant demanding requirements for special operations, which seems to only be growing day by day.
We have ongoing, vibrant conversations with the special operations community about artificial intelligence and the cutting-edge capabilities that our company can bring, and they’re invested in working with us hand in hand.
McHALE: How can the U.S. military dominate the information space with AI?
WALL: This is such a pertinent question. Three or four years ago, [the challenge was called] cognitive advantage, information operation, psychological warfare, whatever term you want to use of the day; I prefer cognitive advantage.
[This] was kind of a back-burner topic for the special operations community. It’s always been a direct-action counterterrorism advisory type of command, things that fly, things that shoot, things that blow up. But in the past four or five years we’re seeing growing requirements and the impact of information operations and cognitive advantage.
You don’t have to look any further than the recent conflict with Iran to see how nations who cannot match us peer to peer in a tactical battle are turning to the asymmetric use of information operations to attack the U.S. strategically.
I don’t know, John, if you’ve seen any of the Lego videos that have come out that have been sponsored by the Iranian government or people affiliated with Iran, and they are directly attacking the U.S. government’s legitimacy for the war with Iran, the U.S. [administration’s] competence. They are bringing up all sorts of issues to try to take away public and governmental support for the war against Iran, or the conflict against Iran. They know that they can’t fight us tactically, militarily, so they are trying to fight us strategically through social media and messaging to diminish the support that we get both internally and externally from our allies and our populace and internal to our own government. It’s a strategic level of war that’s only going to continue to be used more and more.
In some respects it’s working, and mostly it’s working because it’s a battlefield that the U.S. has not generally engaged in robustly, and I think we’re beginning – across the whole of government – to realize this is something we have to pay attention to.
In the past year, the National Security Council added a position [member], who owns the portfolio for cognitive advantage. That’s never happened before. That was not something that the National Security Council thought about much when I was in [the service] 20 years ago. Information operations was kind of an add-on. We’d plan a tactical operation, and then we’d say to the IO guys: “Hey, what can you do? You know, print up some leaflets.”
Here’s what’s interesting to me about the whole idea of cognitive advantage, In World War II, when we invaded Normandy, we had to have massive amounts of people coming across the shore. We had paratroopers from the 82nd and the 101st dropping in behind enemy lines, because that’s what we had to do to get there physically. Now our adversaries can parachute ideas metaphysically into our populace without any [risk] of being interdicted. [They do this through] social media, which a majority of Americans get their news and form their opinions from.
So, there’s a really pressing need within not just the special operations community, but across all the military services and the whole of government to understand and identify what are the narrative messages that our adversaries are putting out there against us. First, let’s identify them. If we can identify the narratives that our enemies are putting out there against us, then can we identify counter-narratives right in that information advantage world to fight that narrative, and then we can measure the effectiveness of those narratives. Are we changing minds? Are we changing how people act?
At a certain level of war the only important terrain is the six inches of gray matter between the ears of the leadership and the populace of your adversary, and so we have to be able to affect that. There’s lots of ways to affect it. You can affect it with action, you can affect it through tactical military operations, but we can’t ignore the cognitive realm of social media and information operations. It is of growing importance every day, and with the speed of information that goes around the world today with artificial intelligence, our adversaries are weaponizing it easily, and if we don’t do something to identify, counter, and measure, then we’re not in that fight at all, and we’re going to lose strategic battles. We’ll win tactical battles all day long, but we’ll lose strategic ones because we’ll never win popular support and we’ll never get into making the will of our enemies,
McHALE: Can you talk about your knowledge engine platform, Bill? How does that enable the information dominance you just talked about?
WALL: It’s such a great question for people to think about today. In the past couple years, we’ve seen social adoption and use of large language models [LLMs]. People are very familiar with them. People use them as personal assistants, and there’s a challenge with large language models. They’re only as good as the information that they have access to. And it’s been proven that if they don’t have the right answer, they’ll tell you anything, and people call that hallucination. So, a large language model by itself isn’t sufficient for our customers. What they really need to do is pull all of their different data out of all of its different silos into one large area and de-silo that data, extract entities out of it, extract relationships, and then put context around that. Then a large language model can be used to query that data. Large language models are built to do search. They’re not built to do analysis, so you need a knowledge engine to enable that analysis.
[For example] maybe a large language model could find out that you and I are neighbors, and it would connect us, and there’s a connection [that] we live next to each other, but that doesn’t contextualize our relationship. Maybe you hate me because my dogs are always in your yard and my kids jump over the fence or in your pool, or maybe we’re really good friends and we barbecue together on the weekend. Context is what matters here, not just the connection. To get value out of the information it has to be brought together, contextualized, and then made available, so that a large language model can then search against it and bring to an analyst the answers that they want. More data points don’t help an analyst. Data points that are contextualized help an analyst.
McHALE: Regarding large language models, you’re talking like Chat GPT, Claude, which are popular among consumers. My colleagues in sales swear by LLMs because they help them do data analysis or as a search tool. But as you say it’s only reliable as the information it can access and that can be verified. The other day in a meeting, colleagues were discussing how two companies are related. A search says they’re competitors, and on some products they might be, but they actually might be partners on some systems. It [calls for] further follow-up on what the LLM provided.
WALL: And with even deeper business relationships, what are their financial positions? Who’s invested in them? Is there a common investor? Are there board members in any way related? Are there board members from the two separate companies on the board of a third company? What is the nature of that relationship? A large language model by itself probably won’t find that because that’s a deep level of data and understanding that you’ve got to put together.
Pulling the context out of data is the true value of AI and what helps analysts and we want to make sure that people don’t think we’re trying to replace humans, we’re trying to super empower humans by helping them with this problem that everyone has of being overwhelmed by too much data that they can’t make sense of.
McHALE: How does the knowledge engine provide that context? Can it also verify whether what you’re getting from the LLM is accurate? I keep thinking back to the Reagan years, “trust but verify.”
WALL: I think that’s very important. A knowledge graph, if it’s providing the right context, tells you the source provenance of your information, what time that observation was made, the confidence level of the accuracy of the data, understanding the relationships in the data, the supporting evidence that’s there. And then a human feedback loop that can say, this isn’t what we think it is, there’s a mistake here. That might change over time. Data isn’t fixed in time. Our understanding of what something means might change over time, so putting that context into the data is really important. It takes a bit of subject-matter expertise, and a human somehow involved in a loop of building the right model and verifying it.
In some ways, it’s kind of like a thesis, you have your footnotes to show where you got your data from, and any product, any AI solution that says to a customer, “Hey, look, this is a black box, we can’t explain how we got to the answer” – that’s a bad solution. Humans have to be able to justify it, whether it’s a federal law-enforcement agent who [goes] in front of a judge and shows cause and evidence, or whether it’s an analyst in the military who has to justify to the boss why they’ve come up with a certain judgment.
I think it’s very important that we be able to provide the provenance and the context and the time and confidence level to all of that information. That’s a challenge with things that are on just the internet. Who knows what’s on the internet? Like, there’s all sorts of bad information out there. Anybody can put anything. Just being on the internet doesn’t mean it’s true. Having some sense of your data and making sure that your data has gone through a process to really look at it – and you know, confirm it to some degree – is important.
McHALE: So, along those lines, I could claim on LinkedIn that I’m an expert in just about anything like woodworking, and a search might pull that up, but it’d be very inaccurate.
WALL: Yeah, exactly.
McHALE: Context is very critical for special operations, because it could be the difference between life and death on a mission. Speaking of special operations in the knowledge engine, are there other solutions that you showcased and talked to operators and other partners about at SOF Week?
WALL: We’ve got two big things that we are looking at right now in SOF solutions. There’s a lot of words people use lately in the profession: psychological operations, information operations, cognitive advantage, cognitive warfare. I tend to veer towards cognitive advantage.
[Regarding cognitive advantage] the idea is, what are the trending narratives within a specific populace about a specific topic, and do we understand what the populace’s opinion of something is? Once we have that understanding, then do we want to change that opinion? Do we want to combat that opinion? Do we want to support that opinion? And if we do, how would we do that? What are the messages that maybe would resound in our populace? And then how could we measure whether or not we’re affecting those messages?
[A good example is what I mentioned earlier about] Iran and the videos they are posting that are Lego short movies. They’re Lego characters, and there’s a soundtrack to it, and it’s a very high level of production. What they’re attacking is the legitimacy of the U.S.’s purpose and reason for having the conflict with Iran. What they’re trying to do is diminish both internal U.S. and international support because they can’t fight us militarily, so they’re going to fight us strategically by trying to degrade the support, and they’re doing it in an interesting way – through social media with the videos.
The videos are very pointed. They bring up all sorts of relevant social commentary that you and I might hear on [X], or we might hear our neighbors talk about, and they’re bringing it to the whole world to look at.
How can we measure if this is effective? Is this degrading the support for the U.S. operations from our allies or internally? If we wanted to counter any of their messages, how would we do that? This is something that’s really important to me, having spent a career in special operations, I saw that one of the things that we did not do as well as we could have was fighting this strategic battle of narrative. We did not convince a significant amount of people in Afghanistan not to support the Taliban. We did not convince a significant number of people across the Middle East not to support Salafis jihadists. We fought the Taliban, we fought Salafis jihadists. We didn’t lose a battle against them, but we ended up not necessarily winning the wars the way we wanted to, because we lost the battle of public opinion – what I like to call that key terrain, which is the six inches of gray matter in between the populace and our enemy’s ears.
That’s a big thing for us right now. It’s a growing field. The information age that we’re in, where people get their news from social media, people express their opinions on social media, you can get a good feel for what a particular audience thinks about things on social media, and we think that’s a field that the special operations community could play in better.
McHALE: Regarding what they are doing with social media… we live in a free society, so everybody can get in. Then you look at Iran and their Lego videos. How do you combat that if they control – if they’re shutting down – the internet?
WALL: Part of it is, who’s the audience? The audience for the Lego videos from Iran is not the Iranian audience, it’s the international audience, and it’s the U.S. national audience. Are our allies beginning to say to us through diplomatic channels the same themes and messages that the Iranians are pushing through these videos? That’s what we have to look at – who’s the intended audience, and is it having an effect on that audience, and the different ways that we might be able to change that effect.
McHALE: Let’s look forward a few years. Predict the future. What do you think will be a disruptive technology or innovation or game-changer in AI for military applications, especially special operations?
WALL: The thing that is beginning to happen, and the thing that we’re going to have to wrestle with in the near future, is autonomous decision-making. At what point are we comfortable with machines making decisions for us. To some degree that exists in the U.S. military right now. There are a couple of weapons systems: the Patriot missile system, the AEGIS [weapon] system, which is on Navy ships, which can operate autonomously, because it’s a speed of action. If there’s a missile coming in at a ship, you maybe allow the gun to shoot it when it detects it, as opposed to waiting for a human, because it might take too long. So, within the decision-making processes that we’re beginning to bring AI into, we very much want a human on the loop or in the loop. At some point the speed of the information and the speed that the decisions will require is going to require that we wrestle with the idea of what are we comfortable with in the idea of machines making decisions and directing action for us, and the more that tends towards the lethal space, the more important that decision is.
I think that’s something that I don’t have an answer for, but I think it’s something we’re going to have to wrestle with more and more as artificial intelligence becomes more integrated into our formations, as it becomes more capable, and as we see more value in it. Somewhere along the line we will have to wrestle with autonomous decision-making.
McHALE: AI and autonomous decision-making capabilities are evolving quickly.How do we keep up with that speed of innovation with a slow acquisition system?
WALL: That’s always a challenge, but I think that in the past couple of years I’ve seen a tremendous amount of growth within government customers and the community of both understanding and wanting to understand innovation. There’s always going to be the people who don’t want innovation: Right, we’ve always done it this way. It’s been good enough for me for the past 15 years. Why change the system now that I’ve worked my way up to the top of the system? Don’t change it on me. Particularly with the special operations community, there’s always been a bias towards innovation because it drives speed and efficiency. Those are kind of the hallmarks of the community. Innovation is going to continue to drive that, and the special operations community has always been the cutting-edge adopters of innovative technology.
McHALE: It appears this current administration is pushing for ways to get things done more quickly for autonomous systems and AI – and of course, special forces – being right at the front of that.
WALL: For years the industry has complained about how the acquisition community adopts innovative technology. We’ve always said it’s too slow, and it’s too bureaucratic. But there have been, in the past couple years, in the past year in particular, some great changes. You’re seeing memorandums coming out of the Department of War that are pushing organizations to adopt innovative technology and work with smaller companies and find more flexible contracting vehicles and fund growth and innovation. I think it’s moving in a good direction.
McHALE: What had you most excited about SOF Week this year?
WALL: SOF Week is just a special week. A couple things really excite me about it. One is on a personal level, it’s just always a great opportunity to see old friends and old colleagues, some who are still serving, some who are out like me, in some cases meeting the children of people I served with, who are now serving. It’s just amazing personal connections; that’s one of the things that’s special about the special operations community is the deep personal connections. Humans are more important than hardware.
I love the opportunity to get to talk to customers and potential customers, and really get a feel for what their challenges are. We always think we’ve got the right answer until you talk to the customer, like, “ah, I missed that.” I’m always excited to talk to customers and potential customers and see if there’s a way that we can empower and enable them to be better at their jobs.
The SOF community continues to grow and support each other. The SOF Week experience is unparalleled.
