WSJ What’s News - Booz Allen CEO on Silicon Valley’s Turn to Defense Tech: ‘We Need Everybody.’

Episode Date: June 29, 2025

This week we’re bringing you an episode of our podcast Bold Names, where hosts Tim Higgins and Christopher Mims interview leaders of the bold-named companies featured in the pages of The Wall Street Journal. In this episode, Horacio Rozanski says he is obsessed with speed. As the CEO of Booz Allen Hamilton, a company that helps government agencies leverage the latest advances in technology used by the private sector, he has insight into the global race to develop artificial intelligence–especially in the realm of warfare. How does Rozanski see the relationship between the U.S. government and Silicon Valley evolving?  Learn more about your ad choices. Visit megaphone.fm/adchoices

At Desjardins Insurance, we put the care in taking care of business. Your business to be exact. Our agents take the time to understand your company so you get the right coverage at the right price. Whether you rent out your building, represent a condo corporation, or own a cleaning company, we make insurance easy to understand so you can focus on the big stuff, like your small business. Get insurance that's really big on care.

Find an agent today at Desjardins.com slash business coverage. Hey, What's News listeners. It's Sunday, June 29th. I'm Alex Osala for the Wall Street Journal. This is What's News Sunday. On today's show, we're bringing you an episode of Bold Names with hosts Tim Higgins and Christopher Mims talking with Horacio Rosansky, CEO of Booz Allen Hamilton, a company that helps government agencies leverage the latest advances in technology used by the private sector. They discussed big questions, like how the US can compete with China on AI, and how Silicon

Valley is getting into national defense. It's an interesting indication of where technology is headed in areas of government, national security, and more. Hey, Mibs, before we talk to today's guest, did you ever think about how you would unclog a toilet in outer space? Honestly, that is not a question

that has kept me up at night, but of course, now I have to know. Well, that is not a question that has kept me up at night. But of course, now I have to know. Well me too. But in today's episode, we tackled this difficult challenge and more with the CEO of Booz Allen and also how the firm is handling cuts with Doge and the Trump administration and competing with AI in China and why it's partnering with big names like Meta and Palantir. There are a lot of surprises in this conversation. You're not going to want to miss it.

That's next. A quick note. Before we get into the conversation, we recorded this interview prior to Israel and the US strikes on Iranian nuclear facilities. Now onto the show. The arrival of the Trump administration and emphasis on finding government efficiencies has affected a lot of people's lives.

But some have really felt it, like Booz Allen's CEO, Horacio Rosansky. 98% of the company's revenue comes from the US government. And much of the work it does is so secret that Rosansky can't even talk about it or disclose what those giant contracts are for. The work, as the kids say, is mysterious and important. And Rosansky tells us he generally agrees with efforts to make the government more efficient. There's so much duplication in the way we are regulated

and the way the government does things that a lot of money could be saved. Rozanski calls Booz Allen a tech company that helps government agencies leverage the latest advances in technology used by the private sector. And his role gives him unique insight into the global race to develop AI,

especially against China. If we all believe, as I do, that AI is the dominant technology of the next 20, 25 years, we need U.S. leadership to be in front there. And to me, it goes back to speed. We have the goods. We just need to deliver them faster. From the Wall Street Journal, I'm Tim Higgins. And I'm Christopher Mims.

This is Bold Names, where you'll hear from the leaders of the bold named companies featured in the pages of the Wall Street Journal. Today we ask, how is government consultant Booz Allen navigating the Trump administration's cuts and looking for opportunities in Silicon Valley's turn to defense tech. Welcome Horacio. Looking forward to talking about how you see Booz Allen as a tech company and the threat of China's AI development to the US. But first, let's discuss the elephant in the room when it comes to your company and the current situation in Washington DC. When 98% of the company's revenue comes from U.S. government-related work,

and the Trump administration's efficiency efforts have really targeted the likes of your company and others. What's the situation when it comes to government efficiency? I think for a lot of people, the image of Elon Musk waving around a chainsaw comes to mind. Now he's pulled back from Doge. Is everything better for you now? Yeah, did you get doged and then he exited?

Let me start by saying, I think we all should be in favor of efficiency. Every large institution always has to look for ways to become more efficient. And if you're the federal government, there'd be one way to think about it is your board of directors has 535 members, 435 in the House and 100 in the Senate. It is actually quite difficult to be as efficient as possible. And so we've been on the side of change and the side of efficiency from frankly long time before. I think we can all argue about is this particular way of doing it is that particular way of doing it? What's better? And that's that's I think fair game, but I don't think we should anybody should oppose Government that is large at a time where we have significant deficits looking for efficiencies, right?

Absolutely, and we're a little bit of tongue-in-cheek. They're talking about Elon waving around a chainsaw, but this is a real issue for you all having to deal with this and you have proposed a lot of cuts. The General Service Administration, which helps oversee the procurement for federal government has been pushing the likes of you all to propose cuts and whatnot. What's your total run for how much you've proposed cutting? What we propose is different ways of doing our work and the federal government contracting with us to make us more efficient and to be able to pass the savings on to the federal government.

There's a significant portion, if you're in an industry that is highly regulated, there's a significant portion of our cost and the cost of everybody in this industry that comes from this regulation. You know, there's so much duplication in the way we are regulated and the way the government does things that a lot of money could be saved. So you're working with GSA, but are you working with DOGE? Are those kind of one and the same now? You know, we're working and Doge is in the agencies working through these issues agency by agency. And we interact with with people at all levels in the agencies, including the members of Doge, you know, a lot of our work and our technology has been reviewed. I'm happy to say a lot of it has passed muster

with flying colors. And so we you know, I have a personal philosophy, which is I will talk to anybody who will talk to me. And that's with Allen's ethos. You know, we, we want to engage and we want to be part of the solution. I mean, I think it's interesting kind of situation here. A lot of the stuff that we heard out of Elon Musk and Doge, I think you probably could buy into the idea of looking for ways to do things better and more efficient using technology as a way to improve processes. I mean, listening to Elon Musk talk about some of the horror stories of the things he's found, he's always pointed this old mine in Pennsylvania where government workers are processing retirement

paperwork by hand and pencil and whatnot and saying, this needs to be more efficient. This needs to be modernized. Yeah, and Tim, I will say this. I think one of the big value added of all of these processes and what Doge has brought is we are now having this conversation. A lot of the things that people are finding out and getting interested in, we've known for years. all of these processes and what Doge has brought is we are now having this conversation. A lot of the things that people are finding out and getting interested in, we've known for years, we've argued for change and modernization of these things for years. I didn't get invited in podcasts until now to talk about it. And I think this is a good process for the American

people to know how our government operates. Well, let's lean into that because, you know, I think a lot of people think of you as a consulting firm, but you talk about how the company is really a technology company. And I heard you the other day talk about how AI has increasingly become foundational to how the government operates. And I'm curious to hear more about that. how your company is developing AI and how you're seeing the government actually deploying that technology. So let me give you quickly the sort of the arc of Boozal in the 21st century.

This is a 100-year-old company. It is a 100-year-old company. And I won't take you all the way back to the beginning. But we are a pioneer in management consulting. I joined the company, by the way, as a summer intern in 1991 in our private sector business as a consultant. And I learned a lot doing that work.

And I think I added a lot of value. Wow. So straight out of university. Straight out of, I was in grad school. Yeah. And I guess I must be not so great at writing resumes or interviewing because I stayed and frankly have loved my time at Booth Allen. But around, you know, post 9-11, we became much more central to some of these national security missions. Counterterrorism, defeating IEDs in Iraq and Afghanistan, a lot of work around cyber security.

And the closer we got to those missions, the more the equities and the needs of the government and of the national security apparatus tend to conflict with potentially some of the work you're doing in the private sector around the world. Case in point, I've been sanctioned personally by Iran and by Russia, and there's many places in the world I cannot travel to. A lot of our colleagues had the same dynamic.

So in 2008, we actually separated the two businesses, the commercial consulting business that most people knew, the business that we started with in the 1910s, from the business that was doing increasingly more and more mission technology work in the federal government. If you look at our business now, about 75% of what we do is national security oriented. Almost everything that we do is about taking technology, whether it's technology we develop or one of our commercial partners develops and making sure that that technology gets into these missions. Tim, to your point, on that journey, we've been pioneering new

technologies really as they've come up. I began our investment on AI in 2014. 2014. 2014. I was already just like looking at these things. I have a part of my degree in statistics and I had looked at neural networks and I saw some of that stuff. I was deeply interested. If I'd been born 20 years later- So you saw the transition to deep learning and its broad applications, especially in

national security. Absolutely right. If I'd been born 20 years later, I would have been a data scientist. We didn't have cool names like that when I got through school. Right. When you got through school, it was just like actuary. Yeah, I didn't want to be that. So I got a degree in business administration instead,

but I took all of the coursework that could have made me an actuary. And, you know, as I saw this journey, we were really interested in that. One of our scientists tells a story that she spoke at a conference and talked about CHAT-GBT-1 back in, I think it was maybe 2018 or something, and people were looking at her like she had two heads. And then, you know, when LLMs finally broke into the public consciousness, we were ready. And if you look at us now, we're working on agentic, we've been doing that for a while, we're working on physical AI. That's everything from autonomy to humanoid robots, to digital twins. And

then we're also working on adversarial AI. This is the stuff out of science fiction, right? We think of science fiction, whether it's robot cars or drones flying through the air having combat or humanoid robots doing human-like things. I mean, this is the cutting edge you're investing in there. We are. And it's going to be really important for these models to understand the physical world in the way we understand it, to be able to interact with it,

to be able to become really useful. If you think about a world where we want to reinvent manufacturing, if you think about a world where we want to keep our troops safe in the field, and we want to have drones firms as opposed to individual drones and so forth. We're going to need physical AI capabilities. There's a ton of investment going into that. We're certainly not the only ones.

But in the national security space, we really understand it. We just heard how Rosansky sees Booz Allen as a tech company. Next, why people in the tech space are becoming interested in national defense, an area that until recently was definitely not very cool in Silicon Valley. The fact that a lot of these companies and a lot of these brilliant minds have taken an interest in our national security, we welcome them and we want to work with them because we ultimately need to be faster as a nation, we need to be more successful as a nation

against these new technologies, and we need everybody. Stay with us. There are a lot of other folks investing here, and you're partnering with some of them, right? So Palantir, Shield, Meta for AI, and Space. What role do you play in those partnerships? So on physical, we're partnering with Shield, we're partnering with Scout AI, which is on the ground base. And then we're also partnering with NVIDIA as we bring their capabilities to bear here.

In all of these things, we play multiple roles. One of those roles is we actually build our own technology on top of their stacks as opposed to starting from scratch. I'm sure we're going to talk about this when we talk more about AI, but I'm obsessed with speed and reinventing the wheel is not what we're about. So we work with these companies to take what they do and help transform it so that it is more valuable to these missions. You know, Shield is unique because they have a national security first footprint, but a

lot of companies are dual use, right? And I think a lot of the investment that the government needs to take advantage of is this dual use investment where something is built for a commercial application, but it can actually be used as a national security application. But help me understand then what's your, I mean, Shield, right, they started making drones, they're very national security focused, but Palantir, they're no strangers to national security, Meta obviously is for now. So what's your role? Are you an intermediary? Are you a translator? Are you an integrator? It depends, but by and large we're a co-developer. A lot of these

technologies were not architected to serve in the battlefield as it exists today. So a lot of these amazing technologies and some of the great companies that you talked about, their technology was built thinking of the cloud as it exists in a commercial environment, where you're always going to have connectivity, where the thing doesn't have to get wet or dirty, where the thing doesn't have to operate from space. So we close those gaps by building technology that actually makes that happen. So I'll give you an example. A lot of Edge applications, even today for commercial use, assume constant connectivity.

And just to translate for our listeners, edge just means there's a device out there doing the computing, as opposed to just like a dumb terminal that's talking to the cloud. Yes, right. And so you have this device and it wants to connect with the cloud. And when it connects with the cloud, it's beautiful. If it loses connectivity, it can take 45 minutes to resync. Which is no good if you're a cruise missile or something. Or a soldier. Soldier in the field, or if you're a commander out there

trying to make decisions and trying to make sure that your information is up to speed to what the headquarters has, they can't wait 45 minutes. 45 seconds can get them killed. And so we have built technology that essentially addresses that specific gap. In a recent exercise, one of these great, there are two instances of the technology running in a military exercise. Think of a wargame. One was built on our stack and the other one was

built on the traditional stack. Well, guess what? When the power went out, when the bandwidth went down, when they simulated an electromagnetic attack, our technology kept working. The other same, same underlying system stopped working because it was not it was never been prepared to operate that way. Think about space, we just put LAMA up in the International Space Station working with META. It's very cool. You know, after we... So is that so the astronauts don't get lonely? They have somebody to chat with?

So this is on the research side, but what this thing does is it ingested all of the maintenance manuals of everything. And so now you can query them all and say, hey, how do I unclog a toilet in the space station, as opposed to having to go read a big manual? It's the same thing I do at my house, but it's, you know, getting that up in space and making it you're on clogging the toilet with AI at your house. I will ask you at GPT, how do I fix this? How do I do that? I know how to unclog a toilet all by myself. Thank you very much. But you know, these are the kinds of questions that so this is really just what we

wanted to do is make sure we understood on the compute environment that you're going to have in space, which is by definition always going to be old, right? Because you launch something, and chips change on Earth every six months, that's going to be up there for 20 years. How do you make the new technology operate in those environments? That's what we bring to the table because, you know, the guys at Meta are incredible and what they built with Lama is extraordinary. But they never thought about, okay, how can I make it operate on a 20-year-old computer? Because nobody around here has a 20-year-old computer. So that's the gap that we closed to make these things mission ready

in a national security problem. So just to put a fine point on it, like you got Lama to run on the equivalent of like a Commodore Amiga or something? It's not quite as bad. A 486? But we're working on all of these things so that they can really operate across satellites that have been launched at different points in time. And therefore you're gonna have completely different computer infrastructures. And at the same time, you need to make sure

that it works, that you can patch, that you can update, that you can protect these systems across what is a very, very, you know, again, if I look at Booz Allen as a company and every company is the same, right? I mean, you have some old servers and some new servers, but that stuff could be three, four, five years old out of sync. When it's 20 years old, 20 years out of sync, you almost have to rebuild this technology to operate differently.

And that's, if you think about what do we do, we make sure that these things can run They almost have to rebuild this technology to operate differently. If you think about what do we do, we make sure that these things can run in the environments that our war fighters are going to need so that this is successful in the field. It's interesting. You have a venture arm, so you're out there scouting startups to invest in. I'm just curious what the mood is like among these young tech companies when it comes to defense and defense contracts. Because I think a few years ago, that was going to be a tough sell.

I mean, you listen to Palmer Luckey, who's the founder of Androl, really one of the hot tech defense startups out there. He talks about it like being out in the wilderness and not being loved and kind of being shunned. But now he's getting his picture taken with Mark Zuckerberg and they're partnering up as well. I mean, this seems like it's a hot area right now. What are you finding?

It is a hot area and I'm so glad it is. Back when I was getting sanctioned by Iran and unfortunately my kids had to go to school with security detail. All of these tech companies were saying, well, you know, not for us, not interested. You know, we're happy developing games and all of that. I think the fact- At tech or something. Yeah. The fact that a lot of these companies and a lot of these brilliant minds have taken an interest in our national security, we welcome them and we want to work with them because we ultimately need

to be faster as a nation. We need to be more successful as a nation against these new technologies and we need everybody. But what's changed? I mean, where is this kind of, I don't know if you'd call it new patriotism or kind of new sense of kind of America versus others. What's changed in the minds of these folks, do you think? I think it's come from both sides. I think from the government and from the Department of Defense, you know, when they established the Defense Innovation Unit, they began to demonstrate that they really were interested in working with these companies and that there was an opportunity for them to serve the nation. And then on the other side, you have a lot of veterans now starting these companies and coming at it with that mindset.

They're also looking at what's happening with the large companies. The hyperscalers are also much more interested in national security than they were five, eight, 10 years ago and seeing real opportunity. I think that the combination of the opportunities there, the mission is so important and some of the people coming in, come in with this mindset of service. All of that has segregated into this huge bow wave of tech coming into government that, again, is both excellent and much needed. So as somebody who is clearly a nerd for edge computing.

Are you talking about me or you? Yeah. Always talking about Jim. We've definitely brought the minds together here. And you know, tech for warfighters and everything, you must have found the recent attack deep inside Russia, including an attack in Siberia orchestrated by Ukraine, publicized really by Ukraine, including the details. You must have found that electrifying because I mean, it's everything you just talked about.

I mean, it's these trucks that opened up and it's drones where it's all edge AI that have been trained to, you know, find and destroy airplanes. A lot of people like me are saying this feels like a new era in warfare because we've never seen off-the-shelf AI and computer vision used like this before. Do you agree? Did you find it surprising? Is this a new era? Ukraine has demonstrated that necessity, once again, is the mother of invention.

And they have brought things together that were meant to be together as commercial technologies to fight a war that is not of their choosing, but it's existential. And so what they've done, I think, is with drones, with autonomy, with coordinated fires and a number of things, using all commercial technology, because that's what they have available to them, I think has been impressive. And as a result, you know, we now are talking about things like a treatable mask, meaning you can put a lot of drones up there, and you're gonna lose the majority of them, and that's okay. As a new way of thinking about defense.

And I think it's going to be very powerful. And the whole world is watching. When I traveled last time I was in Taiwan and in Japan, they're watching so closely from the very beginning about what's happening in Ukraine. They're watching very closely about what's happening in Israel and really trying to understand the lessons of that

and how those may apply to them. And we certainly are, as a nation, need to watch, we need to learn, and we need to lead. In particular, Taiwan is interested because concerns about China, right? And what issues are going on there? What is the next kind of threat possibly there, I would imagine? Exactly. Exactly. And so, you know, they are threatened by potential invasion and they're trying to figure out how they would defend themselves in much smaller country right next to a very large and powerful superpower. And so that's, you know, they've been thinking a lot about that. And then this is again, I mean, going back to the national security point that I was making before, there's a lot of companies that because they have large businesses in China, will not serve the national security needs of, of Taiwan. And as we

configured our company, we wanted to make sure that we would never have that hesitation of doing something with an ally of the United States because of a different business interest. We reached out to the ministries of foreign affairs for both Japan and Taiwan. They did not respond. Now that more tech companies are interested in national defense, Rosansky sees a need for them to think about how the government will use their products to help speed up deployment, something he suggests is currently more of an afterthought. Instead, the government comes in late once the technology has been proven. And you know, again, that minimizes risk, but it also slows us down. That's next.

You've raised the alarm in the past about China's progress on AI. It really feels like the gap between the US and China on that is closing as it has closed on many other technologies like EVs, like energy production. What are you concerned about there and why? Let me pick it up in a couple of different directions, but on AI specifically, if you look at what's been happening is again, once again, in the necessity being mother of invention, China is specializing in models that are potentially less powerful, less exquisite, but operate on cheaper, simpler hardware. DeepSeek, I think, brought this thing to the fore.

But if you look at what Alibaba is doing with Quinn, if you look at what Baidu is doing with Ernie, if you look at what Tencent is doing, ultimately, they're building models that can operate more cheaply. Why is that important? Because if you look at the rest of the world that at some point is going to want to use this technology, is going to need these technologies, they sometimes cannot afford

the power requirements of a big data center. They don't have reliable electricity, reliable communications and so forth. And so cheaper footprint models, even if they're not as good, are going to be necessary. I don't want to live in a world where Chinese AI models of suspect provenance that are always calling home, we've seen that in their APIs, that have code that may make them do things are going to potentially shape public opinion, ingest massive amounts of data from around the world and so forth. I think we all take for granted the fact that the internet is such a vibrant and open system, but that's because the US created it in its image. I don't want to live in a world where we all operate in the Chinese internet, where the

government decides what we get to see, what we don't get to see, and when we get to see it. And if we all believe, as I do, that AI is the dominant technology of the next 20, 25 years, we need US leadership to be in front there. And to me, it goes back to speed. We have the goods. We just need to deliver them faster. Let's talk about speed for a second, a real quick second, if you will. You wrote in a piece for Fox News back in February, and I'm going to quote it here, accelerating the military's

responsible use of AI is not a threat. it's an imperative. Speed is our most important weapon in a rapidly evolving geopolitical and technological environment." This was really kind of in response to the idea that maybe people should slow down on AI, I think. How does the US military speed ahead with AI while still evaluating the risks, avoiding catastrophe, just making sure it's safe, right? I mean, I think we all agree we want it to be safe. 100%. And I think it's okay if we slow down right before delivery to the warfighter.

Right before delivery to the warfighter, we need to be absolutely sure that the technology works exactly as intended. That to me is a risk we should never want to take. But upstream of that, you know, can we try multiple things? Can we recognize that some will fail? Can we in fact look for those failures and learn from them as opposed to avoid them like the plague? That's what's going to accelerate us as a nation.

This idea that more companies are going to compete for these things. I've floated the idea that if you look at the US government as an early adopter, as opposed to coming in late as a scale buyer, that they could help accelerate not just the defense requirements, but the broader set of requirements. It'll both accelerate the markets and the market instincts to go faster, but also it'll insert some of these unique requirements of national security earlier in the conversation. Right? I mean, if you knew that the government was going to be one of your largest buyers of this and is getting in with you early, you would create the architecture so that you don't have the problem of disconnection,

so that you don't have some of the issues around cybersecurity that some of the cloud providers have. You would solve those early because that would be a requirement for doing business with the government. Instead, the government comes in late once the technology has been proven. And again, that minimizes risk, but it also slows us down. Well, it's interesting. I've heard you talk about your company doesn't do business with China because you have a huge part of your business is with national security oriented stuff, really classified stuff that you're not even allowed to talk about, which I think sometimes is a criticism

against the company. It's kind of a black box, if you will. Do you think some of these USAI companies should be thinking about where they're taking resources, whether it's investment or they're setting up shop, being worried about the national security risk that they have? Absolutely. I think every company needs to be thinking about that. I think, you know, especially young technology companies, if you think about what they are, you have a bunch of

brilliant people, a great idea, and some intellectual property. And then if you look at the way that they run their cybersecurity, for example, it's in many cases somewhat primitive. And that exposes their intellectual property, their biggest asset to espionage, industrial espionage from nation states as an example. And again, I mean, there's a whole topic of other serial AI, for example. This is something we've invested in and I believe we're leading on, which is how do you make sure that your models have not been corrupted by an adversary? And if you have access to the model weights or the training data or both, you can do a lot of things to a model that nobody will ever know until it's

too late. For example, that if you affect the training data for them all, you can teach them all to not recognize, say, a tank. You can teach them all to think that something that is a weapon is not a weapon. The drone won't see it as a tank or something like that. Right. Exactly right. Or the camera that is providing you security won't see that somebody...

We've actually shown this in our lab. You have this camera that's supposed to detect if somebody has a weapon. And then you put almost zebra markings on that weapon. And all of a sudden that weapon becomes invisible to not to the camera, but to the algorithm. So if a person is looking at it, the person would know for sure that this is just a painted gun, but the algorithm no longer recognizes

this as such. And if you have 6,000 cameras all going at the same time, you're not gonna be able to look at them all at once. And so these are sort of the simple examples. Now, think about if our algorithms all of a sudden become untrustworthy because we're not sure if they've been affected by an adversary, then a commander in the field has to think twice about whether the information that he or she is getting is valuable, is helpful. Somebody giving advice on a course of action to the president has to say, how much do I trust the information that's coming to me if some

of these algorithms have traveled through it? And that hesitation costs time. And time is lives in national security settings. And so these are the kinds of things that we are thinking about, looking at, and trying to make sure that very early in the development of these technologies, we start to embed the protections that are needed. You know, everybody who's ever built a house knows that it's a lot easier to wire it for an alarm before you put up the drywall. for an alarm before you put up the drywall. This is a similar concept. I mean, can we embed the security at the source in the way that things are written, the way that things are managed? I think that's a good place to say thank you for your thoughts and thank you for coming on, Horacio. We appreciate the time. Thank you guys. It's great talking to you.

A spokesperson for the Chinese embassy in Washington DC said its government has never and will never require companies to collect or provide data located in foreign countries that violates local laws. He also said the Chinese government manages the internet according to local laws and regulations and the Chinese netizens enjoy full freedom on the internet. And that's bold names for this week. Our producer is Danny Lewis. Michael Laval and Jessica Fenton are our sound designers. Jessica also wrote our theme music. Our supervising producer is Catherine Millsop. Our development producer is Aisha Al-Muslim. Scott Salloway and Chris Inslee are the deputy editors.

And Philana Patterson is the Wall Street Journal's head of news audio. For even more, check out our columns on wsj.com. We link them in the show notes. I'm Tim Higgins. And I'm Christopher Mims. Thanks for listening.