Loading summary
Transcript60 lines
- [00:01]
Brian Kenny
This episode is brought to you by the center for Creative Leadership. Doing the next right thing has never been more complex. CCL's research based solutions help global organizations and leaders navigate uncertainty and exceed limitations. Learn more@ccl.org possible welcome to HBR on strategy, case studies and conversations with the world's top business and management experts. Hand selected to help you unlock new ways of doing business. Getting a big bureaucratic organization to innovate or adopt new technologies is hard. That's why Harvard Business School professor Maria Roche wrote a case study about U.S. air Force Major Victor Salsa Lopez. He helped launch a program that uncovers ways to use AI to strengthen US Defense efforts. Professor Roesch and Major Lopez talked about the challenges of fostering innovation within a large bureaucracy in a conversation with host Brian Kenny on cold call in 2023.
- [01:11]
Maria Roesch
Maria, I'm going to start with you. Can you just tell us what the central issue is in the case and what your cold call is when you start the discussion in class?
- [01:18]
Victor Salsa Lopez
Of course, the real central issue in the case is thinking about the perils of digital transformation in a large bureaucratic organization. As you already mentioned, this also falls on the broader theme of inn. Often we think it's only small organizations that can innovate startups, but a lot of innovation actually comes out of these big organizations. And so thinking about the trade offs, also of how you organize it and set it up, because in a large organization you may have to do it in a bit of a different way than in a startup. The cold call to really get started in the case is thinking about is it the right choice to set up the AI accelerator as an innovation unit outside of the organization? And then after that we really want to dig in because it's not clear that this is really the best way to do it.
- [02:01]
Maria Roesch
Yeah. How did you hear about this and why did you decide to write the case?
- [02:04]
Victor Salsa Lopez
So I decided to write the case because one of my MBA students who I taught in the required curriculum in strategy came to me with this idea because he was a fellow in the AI accelerator and had all this great experience. I was like, of course. This is so nice. This is the best experience you can have actually writing with an MBA student a case about this. So we dug right into it. And it's also really close to my own research because I'm a strategy scholar and I'm an innovation scholar at heart. So for me, really thinking, thinking about these questions of how can you incentivize innovation? How can you actually adopt these new technologies is core to My research. So it was like the perfect match.
- [02:43]
Maria Roesch
I would venture to guess that most people don't think about innovation and DoD in the same sentence, right?
- [02:49]
Victor Salsa Lopez
No.
- [02:49]
Maria Roesch
So I'm wondering. I think people might be surprised to know that the DoD does invest heavily in this. Can you talk a little bit about their commitment to finding ways to be innovative?
- [02:58]
Victor Salsa Lopez
This also goes back to one of my research projects that is looking at the RAD lab, which was stood up in World War II. It was actually one of the first times you had big science happening in the United States and it was powered by the military and it happened right here. The legacy still lives on today in the Lincoln Lab and also part of AFWorks and the AI accelerator. They touched upon these things that were happening in the 1940s and that's really when things started with military support for research. But there's many, many more examples. I'm happy to tell you more later on.
- [03:31]
Maria Roesch
Can you list us a couple?
- [03:32]
Victor Salsa Lopez
Of course. There are so many. So I cherry picked a coup because it was like, which ones are the best ones? So one of them is the field of material science that actually came out of government funded research thinking about how to create this new field. Another are the first weather satellites. The world's first large scale project on personal computing in the 60s called Project Mac. Also government. You had the first computer mouse. No, it was not Apple. It was government funded research. You had the first mobile robot that walked around using AI actually a long time ago called Shakey Arpanet. Without that we wouldn't have the Internet, the personal assistant that learns pal. Without that we wouldn't have Siri GPS Sematech, which was fundamental in creating the semiconductor industry the way we have it now. I can go on.
- [04:23]
Maria Roesch
I mean, that's a huge sort of legacy of innovation within a place that's not really given credit for that, I guess in popular context. So, Sasha, let me turn to you for a second. We'll switch over to AI about how the DoD has approached AI to this point.
- [04:36]
Victor Salsa Lopez
The AXL was born out of the American AI initiative and it was the nation's strategy in artificial intelligence. That was back in 2019, February. At the time Secretary Wilson at the Air Force answered that call and said, we want to. The Air Force needs to be able to answer this, this executive order on how we're going to move forward. And we've really tried to work on AI ethics and AI safety as well as AI development concurrently. The thought was we need to do this in a way that is open to the world and there's risk inherent in that. We are at the heart a military organization focusing on our nation's defense. And that means at times keeping secrets from those who wish to do us harm. However, this is so important that doing AI in a place that was in the public eye, that we could be questioned in a good way, that we could have peer reviewed research and that we could work in an academic environment at the place where AI is really at the forefront of thought was really important to the Department of Defense. And in doing so, the Department of the Air Force decided that MIT was going to be this home of this AI accelerator. So Lincoln Laboratory, the Department of the Air Force and the Massachusetts Institute of Technology all together working on these problems completely in the open, completely unclassified.
- [05:41]
Maria Roesch
Yeah.
- [05:41]
Victor Salsa Lopez
Publishing academic papers with peer reviews that you can go look online so that we can get that feedback about what is the best way to use this technology, what are the wrong ways to use this technologies, what pitfalls are we going to have? And how do we advance not only the state of the art, but also educate and train our own folks into how to use this and put it into practice in their everyday environment?
- [05:59]
Maria Roesch
Yeah, there's been a lot of concern raised about AI and the potential for it to do harm to. And some of that's come from within the industry itself. I'm wondering for our listeners, can you maybe talk a little bit about some of the types of AI that the military is interested in?
- [06:13]
Victor Salsa Lopez
Absolutely. It's very boring. I caution you. I think the real promise of artificial intelligence in the day to day, right in the small innovations and the same things small businesses and large businesses care about is the ability to make things a little bit faster, gain insights a little bit better and make better business decisions. At the end of the day, how do we spend the money better? So at the AI accelerator, if you go on the website, you'll find a whole bunch of projects. We started with 10 initially. I think they're up to 13 now. I am out of the organization now. I had to move the military moves every few years. But for them, one of our top things was, for example, pilot training schedules. If you ever go into a squadron in 2023, there is one to three people who are in charge of having a massive whiteboard with pucks. And on those pucks are names. And on the lines on the board there are airplanes. And it is some poor soul's job to take each one of those pucks and put them in the right place so that you have a pilot and a co Pilot and an engineer and all of the people that need to fly that airplane ready to go. Until someone of course gets sick, their child gets sick, they have a problem. And then this beautiful schedule that you had float out for the last week completely goes up in the air because everything is broken. Surprise, surprise. We're also seeing this in industry. There may have been a few meltdowns of our own aviation industry recently about pilots getting. This is a really hard, massive problem and the DoD is not immune to it. How do we better optimize for a pilot training schedule? And how can we use artificial intelligence machine learning technology to better do that? That's a very mundane but very important task.
- [07:39]
Maria Roesch
It also sounds relatively harmless.
- [07:41]
Victor Salsa Lopez
So I think. Yeah, absolutely right. And again, something we can put in the public eye and something that we can get. In addition to that, we also have a navigation problem there that's really interesting. Gps, as we're seeing in some of the conflicts around the world, is becoming a more and more contested domain. And we might not have gps. For the everyday person walking around the United States, your GPS probably isn't going to go out unless there's some strange exercise happening. But for the military, we are worried about our potentially losing access to GPS in this signal. So how do we navigate? Using the Earth's magnetic field, Much like a compass, but a little bit more refined using machine learning technologies. We published a data set at the accelerator. That data set with that published got picked up by multiple people around the world. Some very large companies picked it up. We talk about this in the case there was a small business that actually ended up spinning out using that data set. And now that small business is on contract actually through Affworks, moving that technology to navigate very precisely within about 100 meters using the Earth's magnetic field and nothing else. No gps. Right. But this is using machine learning technologies to clean the signal of the Earth's magnetic field while we're flying and all of the other intrusions that happen on the plane with every other electronic piece of equipment.
- [08:47]
Maria Roesch
So there's no super mega laser that's being developed?
- [08:49]
Victor Salsa Lopez
There is no super mega laser. And I think we're. While I think, I think this is something unique to Western cultures too, I think our sci fi really puts a bad taste in our mouth for artificial intelligence. But as Maria mentioned, we had this quadruped robot that the DoD built back in the 60s using AI. AI is not new. We've been talking about this since the 50s or 60s. There's some really funny videos out of MIT actually of professors talking about the dangers of AI, even back then, and we just haven't seen them come to fruition. And I think part of it is when it really comes down to the nuts and bolts about how do you make this technology the best for my organization? It's about the same stuff we all care about. DOD or company, how do I save money, how do I make better decisions, and how do I optimize?
- [09:30]
Maria Roesch
Maria, the case does talk about the difference between legacy companies and digital native companies, and that's important in this context. We'll find out why a little bit later in the conversation. But can you just lay out what the difference is between those two things?
- [09:42]
Victor Salsa Lopez
Yes, of course. So digital first or digital native companies are organizations that were born to be digital. The way they were set up, all of their systems, their business models, how they interact with their customers, are already digital. Whereas you have more established organizations, older organizations. I mean, it's not really their fault. They were born before this was available. They have to kind of catch up, and they keep changing their systems. It's more like a patchwork, almost Frankenstein, like a thing that starts to happen because things keep getting better updated, and then what do you do? You try and fix here and there, but then in the end, you're, like, left with this weird structure, IT structure where everything's kind of all over the place. And so there are these big differences in terms of those kind of companies you can think of, like Spotify, Facebook, Google, even. These companies were born around being digital.
- [10:33]
Maria Roesch
I think many of our listeners can relate to what you just said, though, and I would say Harvard Business School probably falls into that category of a legacy firm, because we have many different systems and we try to make them work together, but it's hard. I can only imagine what it's like at a place like the Air Force or the dod.
- [10:47]
Victor Salsa Lopez
Absolutely. We run into these problems every day. In fact, one of the new efforts at AppWorks, Autonomy prime, is looking at specifically how do we get after connecting all of these systems together. Many of our aircraft and many of our systems have different languages that they speak. So I need a translation layer between them. There's always a fight about standards, but the fight about standards for communication is we see all the standards on the table. We realize that we need a standard to unify them all, and now we have N1 standards to unify them all, and the problem continues. Right? And so this is such a hard problem within the Department of Defense, but having the education, I think, to understand where the technology is good and where it is bad and where we need to put our resources, just like any business is really at the core of the education part of the mission for the AI accelerator and now AfWorks as well as we start to look to try and integrate some of this technology with the warfighter.
- [11:39]
Maria Roesch
Yeah. Can you tell us a little bit about your background?
- [11:41]
Victor Salsa Lopez
I am from South Texas, grew up in Houston, born in Corpus. I studied astronautical engineering at the Air Force Academy. Started my career a little bit before that as an intern at NASA working on mostly earth science. As I Transitioned into the DoD space. Really loved spaceflight. I'm still a nerd. I did not do well. I think I had like a B minus as my gpa barely. But I love it. Oh my goodness. I'll talk your ear off about it forever. And then moved into Georgia Tech systems engineering master's program before actually coming to the AI accelerator at MIT and then working a few research initiatives there on artificial intelligence. Mostly focused on human machine teaming and also multi agent reinforcement learning for kind of the nitty gritty.
- [12:22]
Maria Roesch
Yeah, that background certainly makes sense to have you involved with the Innovation Accelerator. Can you talk a little bit about what the goal, what the purpose was for the Innovation Accelerator?
- [12:30]
Victor Salsa Lopez
Really to conduct fundamental research, to move forward the state of the art for really the good of all mankind. Again, very similar to the RAD lab, very similar to the 1940s. And as we continued that mission early in the early days, we realized one of the big things that we needed to do was start educating our people. It was really hard at the core. For the airmen who were on the team, it was the first time we embedded active duty members into a university. Not to go to school, not to get a degree, but to just be there and liaise and work with the projects. And in a project. We haven't done that since World War II. From the air Force perspective, at least as we got there, we realized that this education piece was really central to the mission. This is a prototyping organization. So there's 10 airmen, there's 400,000 or so people in the Air Force. There's no way that we can do it for everybody. So we said, we're going to prototype. How do we do this? What is the best way? And there's actually, believe it or not, an education AI research project. There you go. Apps like Duolingo and whatnot have really refined. How do you reach out and give people the right notifications at the right time to get them the right education? That same research is being conducted through the AI Accelerator with mit, because we need to know how to do that with our members, not just for things like language, but also for digital transformation.
- [13:35]
Maria Roesch
What were some of the projects that you pursued and how did you arrive at the ones that you thought were worth pursuing?
- [13:40]
Victor Salsa Lopez
Yeah, so this is the fastest I've ever seen the government move. So I mentioned in February, this executive order came out. I think in March, they decided that MIT was going to be the home. In April, they yanked 10 airmen. I was one of them, and we were off to the races. So what we did is we put out a call out to the professors on campus and said, what do you want to work on? And that was kind of the first thing. We are great airmen. I had an engineering background. I have not spent the last decade or four decades working artificial intelligence. So we let the experts be the experts. That was key. They come to us. We had over 200 proposals that were given, and then the airmen had to decide what we thought was the best. So we were reading through them. Having that technical background obviously helps. And then we went down to the force and said, war fighters, which one do you care about? How do we match you with this? And then we went to. This is the weird part. For the Air Force, at least the program office. These are the people that hold our requirements documents that actually decide what needs to get funded, of the one to end things that are needed. Right.
- [14:33]
Maria Roesch
So as we talk about hierarchy and bureaucracy, you're entering into that, Correct?
- [14:37]
Victor Salsa Lopez
Right. So there's the tactical level of people doing the work, and then there's people that really need to pay for the things to do, to work. This is important in any organization, and their mission is very important in the Air Force. But they have a list of things to get done. So we said of these projects, this is what your tactical users want. These are the things we'd like to research. Is this something that you would enjoy as well? Would this benefit your mission? And it was really important to have that bureaucratic at the top support, plus the tactical level support, because if we're missing any one of those, we likely will not get things from academia over into the institution. And so having a deep understanding of where that moves and having the authority to speak to them was really crucial to the success of transition.
- [15:18]
Maria Roesch
You mentioned earlier, Maria, that innovation can happen within a large setting like this. But I would imagine it's not easy, particularly if you're one of 10 airmen and you've been assigned to this new unit and you're trying to strike up relations in an environment that you're not used to being in and then sell that back into the organization. Can you talk a little bit about some of the challenges that units like.
- [15:36]
Victor Salsa Lopez
This would typically encounter thinking about the larger organization? That there is this concept called inertia. When you have things in place, changing anything is going to be incredibly difficult. So you can think about it as cultural inertia. That is a lot about the tacit knowledge, how do processes even work? So even thinking about how do you change any of that, then you also have the architectural or administrative inertia. This is much more about the formal process. So as Salsa was just mentioning, you have this hierarchy. Everyone's been doing this for years, for decades. How would you even change that? And then thinking about technical inertia or technical debt, as I was mentioning before, if this is the way the structure is been built out, how do you think about changing it? Do you shut down the organization for a week or two, which is going to be incredibly difficult if we're talking about the Air Force. Yeah. So how do you then actually do that? And so there are different ways of how you can organize for this kind of innovation. The AI accelerator and AFWorks are these kinds of approaches where you actually take a little group and take them out of the organization. Now you don't have to worry about the formal hierarchy. You have a small group of people, you can be a lot more informal. You don't have to use the normal official way of talking to people. Things are faster. Salsa also mentioned earlier the openness. If you are in the Air Force, think about all the secrecy and like the data that you're not even allowed to talk about. But in these smaller organizations, it may be a lot easier to actually have these conversations. And when we're thinking about how technology or innovation is produced, it's fast, has a lot of feedback, often involves face to face interaction. Being close to universities, which are very open in that sense. Right. We disclose our work publicly.
- [17:19]
Maria Roesch
Yeah, it's like the complete antithesis of.
- [17:21]
Victor Salsa Lopez
Exactly. So it's like polar opposites. And so how do you do that? And then maybe the best way to do it is actually go out of the organization.
- [17:28]
Maria Roesch
Yeah, yeah. Salsa, how much autonomy did you have to make decisions to move things forward?
- [17:33]
Victor Salsa Lopez
Quite a bit. So normally, as any good large organization has almost every role in the DoD has what's called an Air Force instruction attached to it. And we have inspections that go through of are you doing everything that your job requires. So the scheduling people we talked about with their big old whiteboard will Have a checklist of things that they are required to do. And they will have an inspection to ensure that all of those things are being done. And you can get in trouble if you don't do them. And you'll need to show how you're improving. There was none of that for this job. There was no Air Force instruction. We were the first cadre. Our mission was to go accelerate AI, which is so broad and massive and so sourcing the right talent, because it takes a special kind of person to go into that environment with no instructions, no real formal goal. And so there was a lot of how do we make the best value in a short amount of time? And balancing that with how do we start building these long term, ideally, relationships that are going to continue providing dividends back to the organization well after we move on. And so lots of autonomy to keep it short and lots of risk here.
- [18:39]
Maria Roesch
Yeah. Maria, the case talks about rewiring the workplace. This is another concept that I think broadly applies beyond just the DoD example. Can you talk a little bit about why that matters?
- [18:47]
Victor Salsa Lopez
This was really important to us to think about it because as I was just mentioning, when you have to change things, you're doing something different from the source organization. You also have different goals, you have different mission, you have a different vision and mind. And so the activities you do are going to be different from what the source organization does. So you have to think about a different way of getting these activities together, choosing the activities you do, but also how they align with each other. And a way that you can think about it is really thinking about four components, people, process, product and place. And kind of thinking about these different parts of how you want to structure them and how you want to change that from how it used to be. So thinking about the people, Salsa already mentioned this, who do you hire? Are these different people from who would work or flourish in the more hierarchical setup place? Where do you go? Do you move away from the Air Force? Do you locate close to a university? Do you locate in Silicon Valley? Where do you go? Because you want to be close to that new information. If you're trying to learn all about AI, you want to go where that's happening, but you also want to be close because you want to be able to bring the knowledge back. And so it's a very fine line and a trade off thinking about how you do that. And I have other research that really touches upon that, because technology adoption among startups, for example, sometimes it's only 20 meters that matter in terms of learning.
- [20:08]
Maria Roesch
From each Other we've done several cases now on AI and how organizations are trying to plan for it. And the kinds of things that you just described are what many large firms are trying to figure out how to do, particularly on the talent side, because we know everybody's going to need a new set of skills going forward. Not all the jobs that exist today are going to exist five years from now or maybe even two years from now. So I think a lot of firms are trying to think about how to do this and do it well and do it in a way that I think is respectful of the employees that you already have. How can we retrain people, get them to think differently about what they do? So very interesting. Salsa back to you for a second. You're very good natured about this whole thing. It sounds like it was fun. I'm going to guess that it wasn't always fun. That there's some challenges that you've encountered along the way. Maybe you can describe a little bit about the road bumps.
- [20:50]
Victor Salsa Lopez
I think really in this role, I'm not selling anything. At the AI accelerator, the goal was to do research, but we did. I at least felt as though I was a bit of a salesman sometimes. I needed to explain the good and the bad that was going to come with adopting this new technology and the technical debt that was associated with it. Let's go back to that pilot scheduling problem. There required investment money, not just for the research, but then to actually put it into production. To actually go and say, hey, I need money. Which means I had to go back to the large organization and sell this idea. But I'm not selling anything. I'm not. I don't make any money. That I think for me was particularly hard of finding that fine line of trying to explain what needed to be done, how much it was going to cost, why it was going to be better. And it wasn't always tangible.
- [21:32]
Maria Roesch
It sounds like there's a big educational component though. The burden is on you and your colleagues to figure out how to kind of educate the rest of the organization or at least the people who are helping to pay for things.
- [21:43]
Victor Salsa Lopez
Absolutely.
- [21:43]
Maria Roesch
About why this matters.
- [21:44]
Victor Salsa Lopez
There's a great cartoon that we always went back to. Yeah. Maria knows. She's already giggling. It's cavemen. And they have this cart with square wheels. And there's a scientist in the back saying, with paperclip saying, I have a solution for you. Look. And he's pointing at these round wheels. Right. And the inertia, if you will, of the other cavemen on their Square wheel wagon is. No, I don't have time for that right now. I'm too busy. That is how we do it.
- [22:07]
Maria Roesch
Pushing the square wheels.
- [22:08]
Victor Salsa Lopez
Pushing the square wheels. Right. But they've never seen round wheels. They have no idea. Right. And to the organization's credit, there are going to be things in here that likely will not transition. There are things that are going to fail. Not everything is going to be a resounding success. So we need to be able to be very honest about where our falls lie, where we think some of this technical debt is going to be and be willing to accept those trade offs that our senior leaders give us of. This is something that we can incorporate today. This one we're going to have to wait.
- [22:35]
Maria Roesch
Yeah. And you can learn a lot from failure too. As long as there's a tolerance within the organization to recognize that. Maria, do you think it's harder to do this in a government organization than it would be in a private?
- [22:47]
Victor Salsa Lopez
I wouldn't say it's harder to keep it going. Let's put it that way. I think it may be harder to get it going. But once you've started the initiative, I think the government is very supportive and does have deep pockets and is very patient. But there is this issue, and this is like goes back to Ken Arrow thinking about an underinvestment, especially in more tough technology when there's high market uncertainty, also high tech uncertainty. That what we see is that there's an underinvestment, especially from the private companies. And this is where government can really help. But government can also really help even if there's market certainty. Technology maybe is kind of certain, but requires a lot of fixed cost investment that they can help as a coordinating mechanism. So if you think about Sematech or the RAD lab, there again it was more coordinating the effort because there are many people who want to work on it, many people who want to help. There's so many smart brains and minds out there that can make it happen. But how do you bring them together so that they're not doing all of these tiny isolated initiatives? And that's where the government can be really great.
- [23:55]
Maria Roesch
Yeah, yeah. This has been a fabulous conversation. I knew it would be. I expected as much. And there's a lot at stake. Right. I mean, we are talking about the country's national defense. I'm going to ask you each one more question, but Salsa, I'll start with you. Can you just tell us what the most important lesson is that you can take away from your experience at the innovation Accelerator and at afworks, what have you taken away from that?
- [24:16]
Victor Salsa Lopez
Having these very small, focused teams that can, to Maria's point, bring all of this together and putting us in the right location, both physically, so that we can learn that 20 meter separation and then connect back to the top of the organization. So don't bury your innovation at the bottom. You won't make it through the bureaucracy, through the top. And so getting directly from your senior leadership, your C suite, the intent, so that your innovators can go and make that change. Getting the buy in is really, really important. So don't bury us, give us the movement and the ability to understand and put us in the spot where we can really connect with the people that are going to make it happen.
- [24:54]
Maria Roesch
Yeah, that's great.
- [24:55]
Victor Salsa Lopez
Yeah. So I really want to stress that the Air Force innovates that sometimes gets lost and I think it's really important to bring that to the forefront. So I. That's something students and instructors take away because this is also. Hopefully others will take the case and teach this as well in their MBA programs.
- [25:12]
Maria Roesch
Yeah, yeah. Maria, I'll give the last word to you. If there's one thing you want people to remember about this case, what would it be?
- [25:19]
Victor Salsa Lopez
Often we think that the dominant paradigm to do innovation is internally, that you have to do it all by yourself. But there are actually other ways to organize it. But it comes to the consideration of really, like, do you need speed? Is openness important versus maybe secrecy is more important. Then you want to keep it inside. And so really thinking through these trade offs and that there's not just one way of organizing innovation is what I really hope students will take away. And then when you're setting up these vehicles for innovation, really thinking about people, process, product and place and how they reinforce the goal of the organization, Sometimes people decide, oh, we want these people or we want this kind of product, but then it doesn't fit together and then that could have all been in vain. So really thinking about how everything aligns is incredibly important.
- [26:07]
Maria Roesch
The four P's Maria Salsa, thank you for joining me.
- [26:10]
Victor Salsa Lopez
Thank you so much.
- [26:16]
Brian Kenny
That was HBS Professor Maria Roesch and Major Victor Salsa Lopez in conversation with Brian Kenny on Cold Call. We'll be back next Wednesday with another handpicked conversation about business strategy from Harvard Business Review. If you found this episode helpful, share it with your friends and colleagues and follow our show on Apple Podcasts, Spotify or wherever you get your podcasts. While you're there, be sure to leave us a review. And when you're ready for more podcasts, articles, case studies, books and videos with the world's top business and management experts. Find it all@hbr.org this episode was produced by me, Hannah Bates. Kurt Nickish is our editor. Special thanks to Ian Fox, Maureen Hoch, Erica Truxler, Ramsey Kabaz, Nicole Smith, Ann Bartholomew and you, our listener. See you next.