So, so about three and a half years ago, we're just going to jump in. About three, excuse me, three and a half months ago, President Biden issued an executive order laying out his administration's vision for harnessing artificial intelligence for the benefit of the American people, and so I'm super pleased to be here on stage with two of the people who are charged with making that vision real. It's all on their shoulders. So, we have Clare Martorana, Joe mentioned she's Federal Chief Information Officer, and Mina Hsiang is the administrator of the US Digital Service. Their jobs differ, but they share the idea that they're charged with figuring out how to use technology to connect citizens with the many services the federal government provides. That's a big, gnarly, enduring challenge. A lot of folks in this room have been working on it for years, decades, and we have been talking about a long time. The question now is, can AI help solve that challenge? Can it actually help provide citizens with the services that they deserve and pay for? And, if so, if AI can help, how do you line up people and policies to make that happen?
So, Clare... Actually Mina, I'd like to start with you now that you're here. You've been running the US Digital Service since 2021. Looking at that October executive order from President Biden, how did it change how you do your job?
I apologize. I mean, in many ways, very little. I think our job has been, as fast as possible, to recruit exceptional technologists into the government and put them in positions where they can help transform how agencies deliver services, help more effectively deliver individual services, and provide implementation consultation to policymakers so that as we make new policies, they are well informed by what will truly happen out in the field, out in agencies, in response to those policies. This accelerates that. So, one of the things that was most exciting about the EO for me, I know there was something there for everyone, but one of the things that was incredibly exciting was the attention and focus it put on talent, and on saying the only way that we can meet this moment, when technology is poised to so fundamentally change how we deliver everything, is for us to have talent inside government who truly understands what that means, how it works, and how to leverage that.
And, you know, for all of us wonky folks here, it's pretty hard to get policymakers super excited on questions of, like, HR and talent, and so this, for me, was incredibly powerful, and I think demonstrates the degree of commitment and understanding inherent in that EO, that this is such a fundamental part of making this successful, so that was exciting. The rest of it is incredibly exciting stuff for us to work on, and a huge additional pool of work for us to do, but that is the thing that will most change and supercharge and scale up the work that we do, which is, how do we bring in the right folks to do the work here?
So, in some ways, just a follow up on that, it's a recruiting tool to say, we've been calling for people to come work in government, work on...
It's not just a tool, it's a mandate. It's not only a tool, but it's a statement of purpose, which is we recognize that, in order to do that, there will need to be a surge of talent, and we are committed to making that happen.
Clare, can you make this concrete, if we can? You'll receive hundreds of billions of dollars in IT investments on the part of the federal government. What is the perfect use case for AI, that you see?
That's kind of an easy one.
Okay.
In climate. There's a lot of climate research going on. I read from one of our use cases about a problematic algae bloom that happens in a lake in Guatemala, and they were trying to figure out the algae, and why it's blooming, and how that relates to temperatures, and all kinds of really interesting things. They use predictive AI, and we're starting to be able to build some models that we're giving them some really keen insights in order to save some of the fish that are being impacted negatively by this algae bloom. That's one example. Firefighting, there's an incredible amount of predictive. I won't have all the language perfect, but in our forests, there's a fair amount of tinder that can be actually seen from space, so you can predict where your wildfires are going to be the most gnarly. And so, you can get in front of that by actually tending to some of that ecosystem a little bit differently, or you can also just deploy your people more strategically, and understand where the fires are going to break out, how they're potentially going to travel. So, we're using this in a lot of really interesting ways across government that are giving us some really keen insights.
But, I think the thing that is most interesting about that is, to Mina's point about talent, there's also enabling talent. It's not all researchers that we are going to be hiring. We need IT teams on the ground in agencies that are have modern skills, they are using modern software, they're working on a good tech stack, and they're able to actually help the ecosystem in your agency actually go on a journey of innovation, setting up the governance structure, doing all of the governmental things that we do to make sure that we are testing things safely and securely.
But on this question of citizen services -- those are great examples, concrete examples -- things that actually, I'm a citizen interacting with government? Is AI useful in those sort of interactions?
Yeah, I don't think we're there yet, candidly.
Okay.
I think we're beginning this journey, and there are probably very specific use cases that have even evolved since we did our last use case inventory. For example, a study just came out demonstrating that, in customer service call centers, utilizing AI in a complementary fashion, not to replace humans, but to help bring new people, to help surface the right answers to the people answering the phones, and to help make them more productive, has dramatically increased productivity, particularly for sort of baseline customer service reps. So, it brings new folks up to a certain level, and it allows you to have a much higher level of performance in a call center. We've got millions of phones -- the Federal Government writ large gets millions of phone calls every single month -- and one of the challenges that we consistently have is having the right productivity and the right levels of staffing in the workforce, and so the opportunity to get everyone up to the right level there, utilizing new technologies, is tremendous. It doesn't require us to invent something new, but it does require us to put the right evaluation criteria, or the right safeguards, in place, and then to do normal implementation of new technologies in a product environment that is in use in other places in industry as well.
The US digital service was created in many ways as a response to the failure of healthcare.gov back in 2013. If we had all our AI tools up and running back then, would that have helped at all?
That's an interesting question. I think no. Some of that is why having the right talent is essential. Having cool technology is no substitute for having the right talent. Clare likes to joke that most of our jokes are completely serious, most of our problems are not technology problems, and most challenges in technology organizations are not technology challenges. Having different technologies would not have changed the fundamental challenges that we had with management and multiple contractors and inappropriate management and accountability. So, I think we still would have had a complicated situation. We're not going to eliminate the need for people who know how to manage complex technology situations, and for an understanding of how those interface with the policy and the end objectives. And, in fact, because AI is going to reshape how organizations think about their productivity and their work, you have an even deeper need for having really strategic thoughtful people who understand how technology influences mission in the room.
Clare, did you ever have a thought?
Yeah, really, the way that we cascade information in government is unique, and a little bit different. Congress passes a law, or the President signs an executive order, and thinking that, just from those materials, teams can be effective and impactful, delivering the intent of it, is preposterous. These are written at 100,000 foot view, and we have to translate them for every single agency use case to the best of our ability. So, we've been trying to focus on something we call human -centered policy design, where we're actually getting the policies out to the audience. Making sure that we're working in AI use cases in, specifically civil society, academics, labor unions, an entire cross section, federal employees, federal agencies, and really trying to make sure that we're hearing from them about the obstacles on the ground that are challenging them, whether it's funding, whether it's the color of money, whether it's staffing, whether it's lack of interest from their program partners in not wanting to take a risk, like thinking something's too risky to attempt. So, we're really trying to look at this kind of from all of those different dimensions. But, just foundationally, it's about talent, having the right people in the right rooms, and at the table, to be able to translate.
Just to expand on that, it's not just about new talent. Leaders need to have an understanding about how technology affects the mission of the organization that they are leading. We see this in organizations across the board, and that is equally true in in government. Some of it is recruiting new people, and some of it is making sure that leaders, and people across the board, truly understand how this will influence their work.
In that vein, Clare, you talk about the challenge of the President issues an executive order, that doesn't make it real, necessarily, out in the world. There was an executive order right after President Biden was sworn in, in December, I think, of 2021, about transforming the customer experience.The language of government, one of the very essential things they exist to do is, there's these services that it provides, it has to actually make those services accessible and meaningful to people.
Is there a way, just to play devil's advocate -- that's a lot of work that's still being done, meaning you're working on that day in and day out -- the idea that AI is going to take folks' eye off the ball. It's like we're still working on building usable, accessible websites and apps, and now we're adding this AI layer and everything needs to be a chatbot. We still haven't solved the sort of basic technological questions that we've been talking about for a long time. Is that cynical, to think that way, that the AI is just like the shiny object now that's going to distract us?
I think that they address different parts of the challenge. The customer experience EO is really saying, this is how we should evaluate ourselves. This is what we strive to accomplish. This is how we, as agencies, should seek to evaluate whether or not we're meeting people's needs, and how we should manage our programs. So, that is like a mission and an outcomes piece. I think a lot of what's in the AI EO is much more about tools and accountability. So, in service of providing better customer experience, what are the tools that are available to you? What people does it take to accomplish that? And how do you need to put safeguards and other things in place to make sure that you are doing that in a fair, equitable, reasonable, not creating harm, way?
So the thinking being, you don't have to use AI? But if you do... That's how you read the executive order?
Yes, As with anything, it's another technical set of technology tools, it's not a mandate to use AI in every possible scenario where one could use AI. It's use the best possible tools, and if those include AI, we have to think pretty hard about we have to make sure that we understand what we're doing.
I think, on the customer experience Executive Order, it was also challenging agencies to know their customer, like really know their customer, know the journeys that they go on, the moments that matter, the interactions between agencies, where someone might fall out of seeking to get an answer from the government, and how we might actually design our interactions to serve that individual. It really kind of started at a higher level, it wasn't about the tools and the technology, it was really about, you are senior person at an agency, you have a set of customers, you also have a set of employees who are also customers, and you need to create the discipline, and hire people that have familiarity with a more modern way of working, so that you're not just writing one RFP and deciding that you already have the solution, you have an openness to listening to your customers, and potentially finding gaps in the service that you're delivering, and recognizing that you have to do some work. It could be cross-functional work, it could actually be technology work, to help the person get the benefits or get the answer that they need in the environment.
I think that's right. I enjoyed your question. I think technology for technology's sake is never the right idea, and AI in one form or another has been around for a long time. We have an explosion of interest, in part because it is solving people's needs now, and it is addressing things. So, in the same vein, I think all of this is about what technology actually solves the problems in front of us, and how do we make sure that we do that thoughtfully, and use the tools that are available, not just implement the coolest new thing even if it has no practical application for folks. So, just think it's very important to like make sure that we're we're using the right tool for fixing the problem in front of us.
One of the learnings in the AI world from the last several months, last couple of years, has been that it takes a whole lot of money to build AI well, that's why it tends to be these bigger corporations, these very well funded corporations, that have done it too far. The funding model in the federal government, at least traditionally, has been very fits and spurts. It's a little bit of funding here, the program didn't immediately show that it changes the world, so you get no more funding. How do you sort of marry those two, if you stipulate the federal government should be working in implementing AI tools into their work, how do you marry that with the funding model that says, this is six months, at the most five years, that you're gonna get money for?
Budget, my favorite subject? This is hard. We are going on a transformational journey. But, I will say, we have consistently been investing in a lot of capabilities that we have not -- just as a member of the public, I maybe would not be quite as familiar with. For example, the National Labs. Our National Labs are an absolute treasure. We run out of Oak Ridge the fastest supercomputer in the world, the supercomputer Frontier that replaced Summit 2, which was Summit 1's... We are doing investments on an annual basis for many, many years, building out some baseline capabilities. I think what we don't do as good a job of, and there's a really interesting opportunity, is looking across the ecosystem to find the areas where we do have capabilities, and try and match some of the agency needs to those capabilities. Some of our agencies actually run programs on these supercomputers, the VA has a Million Veteran Program where they're doing a DNA project for the last 10 years, and they are actually using some of the compute power at Oakridge to try and accelerate some of their key learnings. Those are examples of where we are making investments over a long period of time, sometimes generational investments, and we aren't necessarily connecting some of the operating needs in government to the capabilities that we have.
It's going to be hard for every agency to go through this journey, but we have places like NASA, we have places that have been leading bleeding edge across the board. What we're trying to figure out, through some of our communities, is how do we connect the dots between the places that are doing this well already, the key learning and the talent that they have in those environments, how do we match that up with an agency that's just really struggling? Potentially, the thought of them bringing Copilot into their environment is going to be a lot more complex than it is in some other environments. So, we're also trying through some of these councils, and the convening powers that we have, to bring people together to try and unearth some of the assets that we have in government, to try and figure out how we can leverage them to get to where we need to go faster, and learn and share, because it's in a trusted government environment that we can actually share information with each other a little bit more expeditiously.
One of the hallmarks of the Digital Service has always been, let's break big technology projects into smaller parts, iterate, test it out, see how it works. AI, again, requires a lot of money, it can take a lot of years to test out a model, see what you learn. It is not iterative in the same way that a smaller software project might be. You can disagree with the premise of the question, but how do you marry that USDS approach, of sort of scrappy move quickly, with AI projects that can be much bigger in scope?
I think the AI companies would say we've built it piece by piece, and year on year. That said, the federal government does tons of huge inspirational things. We also do some basic stuff. We answer the phones, we answer people's questions. We do incredibly basic things over and over and over again, at massive scale. Those basic things are incredibly important to people. We are having a conversation with millions of families about their food insecurity. We are having critically important conversations. But, they don't require a supercomputer, they require fundamentally doing a good job of engineering a straightforward product, using basic iterative principles that allow us to work with the humans who are affected with the agency, and the policymakers, and with the technology that's available. One of the things that AI brings us is a new assortment of technology that is available. It requires us to be thoughtful, as with all technology. A decade ago, we had to make sure that we were thoughtful about the implications of using the cloud in a bunch of contexts, and now we have to be thoughtful about the implications. But, similarly, we could say it costs a lot of money to build a whole public cloud, but using it is a pretty straightforward thing. Using a lot of these AI capabilities, that others have spent a lot of money to develop, is something that can and should be part of sort of baseline normal software projects and service delivery.
Just one thing to add really quickly. We also have a lot of paper in government, like a lot of paper, and we have fax machines. We can get past this, that is the hope of AI, and just modern software evolving, not just putting everything in an AI bucket, but that we have big opportunities to do really basic things well, and that's inspirational.
The last minute, you both talked about the importance of attracting people to government to do this work, people like yourselves that have private sector experience. Other than saying there's a lot of paper, come help us solve that, what is your current best pitch very quickly for... you talk to folks, please come do this work. Very quickly.
I'll jump in quickly because Mina recruits every single day all day long. It is, you can have an impact in government that you cannot have, that I have not ever been able to have, in the private sector. You can work on missions that span everything from banking, to farming, to space. There's something across government for every single person's interest. And, you will be going someplace where you're rare, your talents are rare. There aren't a ton of technologists, that are my normal colleagues from private sector companies, that I interact with here in government, so going where you're rare, where you can drive a great impact, pretty much gets everybody that I talked to.
Your 20 second pitch?
Clare said it perfectly.This is, I don't know, my eighth or ninth job. There is no other place where, every single day, you can go to bed and know that you have made such a large impact on the lives of real people in complicated, desperate, or important circumstances. You work with amazing colleagues. You're rare, and that means you can have outsized impact, and so it's just a tremendously rewarding place to work, I learn new things every single day. And, it's just an incredible opportunity to serve your country.
It seems like one takeaway is, the work's the same, AI is maybe a new tool in the toolbox. Does that work?
Good summary.
All right. Thank you both. Thank you, Clare. Thank you, Mina. Thank you all for being here.
