Skip to main content

USDS head Mina Hsiang wants Big Tech’s best minds to help fix the government

Ten years after, the fed’s tech consultancy wants you to help with the next big thing.

Share this story

If you buy something from a Verge link, Vox Media may earn a commission. See our ethics statement.

A photo illustration of USDS administrator Mina Hsiang.
Photo illustration by Alex Parkin / The Verge

Today, I’m talking with Mina Hsiang, the administrator of the United States Digital Service (USDS), which is the tech unit of the executive branch — you might think of it as the White House’s product and design consultancy for the rest of the government. The USDS was started by President Barack Obama in 2014 after the disastrous initial rollout of the website, which now lets millions of people sign up for health insurance but launched as a buggy, unusable mess.

Mina was a member of the “surge” team Obama brought in to salvage, a process that took months. You’ll hear her talk about how she worked from a card table with printouts of the user flow taped to the walls around her. After that, and an extensive career in and out of the private sector, she returned to be the third-ever leader of the USDS in 2021.

The USDS has a fascinating structure — it’s composed of nearly 250 people, all of whom serve two-year stints developing apps, improving websites, and streamlining government services. The idea is to make leaving a Big Tech job for a term at the USDS a reasonable idea — the equivalent of design thinking public service. The concept is working — thousands of people are applying for roles, and the unit is bigger than ever, with big successes like the website and a revamped Veterans Affairs website.

But, as you might expect, fixing how the government’s systems work can involve some thorny political and bureaucratic nightmares. Mina can’t just tell the VA it’s going to have a new app and ship it — that’s not how the government works. A huge part of the job is making big agencies realize that they have to make tech products so people can actually use their services — Mina says a popular catchphrase at the USDS is that “software is never done,” which is meant to remind everyone that simply fixing an app or a website isn’t the end of the road.

It’s a familiar issue for any software company, but the stakes for the government are much higher: these systems are how people get their healthcare and other critical benefits.

I’ve long been fascinated by the idea that the interface to the government is just too hard — that most people don’t know what services are even available to them, and they certainly don’t know how to take advantage of them quickly and easily. So I asked Mina about that and if things like AI chatbots could make it easier to have a unified view of government for the average citizen. Her answer is fascinating and really highlights the gap between the profit and growth motivations of Silicon Valley tech companies and the goals of the government.

Mina has a really refreshing way of looking at all of this, and this conversation made me think about how much more often we should think about software in a context that doesn’t involve a giant corporation trying to sell you something. 

Okay, Mina Hsiang, the administrator of the USDS. Here we go.

This transcript has been lightly edited for length and clarity.

Mina Hsiang, you are the administrator of the United States Digital Service. Welcome to Decoder.

Thank you. It’s great to be here.

I’m really excited to talk to you. It feels like we are in a moment where the government thinking about technology is in vogue again, maybe it hasn’t been for quite some time, maybe ever, but the idea that the government should have some capabilities and build good products and, in particular, have the capability to understand what’s going on with AI. It seems like the nature of the government’s relationship to technology and the tech industry has changed. Do you feel that as well?

I think it always ebbs and flows, and it somewhat depends on society as a whole and how we are thinking about it. I certainly feel that in the 10 years that I’ve been in and out of the Digital Service. Right now [there] is definitely a lot of energy around “How do we tackle these challenges and how do we take on these questions head-on?”

Let’s talk about the Digital Service. It started because of a catastrophe. You were there. The Obama administration tried to launch; the site crashed. I think you were on the team that had to fix it. You had like a six-week sprint to fix it. Now it’s here. We’re 10 years into having a United States Digital Service. What is it? What is it meant to do?

And 10 years into a working site—

[Laughs] Important to note, yeah, it still works.

Which is incredibly important, and the teams that are at centers for Medicare and Medicaid manage it exceptionally at this point.

The US Digital Service is created as a place to bring in exceptional technologists and to work inside government to make the government work better through technology and design. How we do this is we recruit in incredible folks who have experience building the best, the most usable, the most equitable, the most efficient and scalable systems, technology products.

We put those people together with people who have expertise in how to get things done in government. So we build these teams of people, interdisciplinary teams, and we deploy those to agencies who are the execution arm of the government to work on critical programs and to supercharge the teams that are already working on those programs to help them set up interdisciplinary teams to ship excellent products, to deploy them against major problems that have a technology underpinning and a technology component, which, as you know, is basically everything these days, right?

That’s a broad remit, at this point.

And I would add that some of the ways that this has shifted over time is we go into the agencies, we help them serve, recharge, their work on a specific program that is of critical importance, but now we’re also, as you say, government has realized that it wants to build this capability and that this is central to our role. So we really work with the agencies to build out their internal capacity to do this as well. Helping them understand: What do product managers do? How do you hire engineers who are competitive from market and sort of standing up the types of management structures that will make all of this effective inside those agencies? So we are an execution arm but also trying to be a catalyst for change and evolution within government as well.

I feel like I’ve been doing this show for two years, and the only question we’ve been trying to answer is, “What do product managers do?” What do product managers do?

Product managers take a look at the core problem that you’re trying to solve and the resources and tradeoffs available and really try and facilitate everybody getting to a path forward on how we should try and best solve that problem given the constraints, the resources available, what we consider a solution, whether it’s financial, whether it’s [the] number of users served. So I think their job is to build a path forward using the people and the resources available and to help everyone air their tradeoffs and make sure that we’re really solving the problem we set out to solve.

I asked that question partially to be silly because I think you could get 500 different answers at 500 different companies, but partially because it’s really important in this context. The thing you’re describing is almost like a consultancy with some internal capabilities.


But then you’re going to build a team for someone else and walk away, and that team will have its own culture. How do you think about that tradeoff? As a consultancy, is it something different? There’s a danger in mapping how the private sector works to the government, but that feels like a pretty close parallel.

I think that is a good model, with the caveat that consultancies operate under their own incentives. US Digital Service does not get paid more if we create more different tasks at an agency that requires our engagement. I would just say I think a consultant model is quite accurate, but there are some core differences. So we’re not incentivized about how much business we can drive there. We are helping the agency.

We have extremely aligned incentives with the agency in a lot of ways in that our goal is to deliver services to the public, and we’re in a position where — and consultants do this, too, but it plays out in different ways — we’re partly hired by the agency, but we’re partly hired by the American people. We’re a part of the White House. The White House provides its degree of coordination and strategic direction for agencies, so they’re also one of our key stakeholders. We’re also in a position where we can tell the agency hard truths or help them redirect or help them think about different approaches, different objectives, in a way that I think can be hard if you’re trying to sell business.

Government’s crazy, right? There are a lot of systems. What’s the biggest hack that you’ve had to undo in your time in the government? 

I’ll go back to just because it encapsulates so many of these things so perfectly. There were 30-something contractors — all of them were building different parts. There was somebody who was managing the infrastructure and somebody else who was managing the front end and someone else who was doing the user research. And there wasn’t a central standup or operations control or anything in advance. I arrived several weeks after the site was already out, and I got brought into this situation, and I’m like, “Okay, so what? The site is down, so I can’t go through the site myself.” And so, I’m like, “What questions are we asking people?” And everyone looks around, and nobody knows what information we’re collecting from users to determine if they can have health insurance. I called the call center because you can sign up on the call center, and they asked me a bunch of questions, and I wrote down the questions and I sent it around.

I was like, “These are the questions they asked. This is the data we collect.” And so starting point: be a user. I think I applied for health insurance — don’t tell anyone because we weren’t supposed to do that — but I applied for health insurance 30 times. I just kept making up new email addresses. And over time, we built... Actually not over that much time. We had to put in instrumentation. And so we got every screenshot. We had a bunch of users go through and take screenshots so we could figure out where monitoring was and figure out user flows. And my desk was like this folding table in the hallway, and I’m in charge of metrics and a few other things and product management and trying to put together key metrics. And on the wall, I have pinned all of the user flows. I have this great photo. It’s literally all these pieces of printer paper with stickies on them where we want to put in Mixpanel monitoring.

And people from every different contractor are coming over and being like, “Oh, where did you get that? That’s really helpful. That helps me debug this. That helps me debug that.” And you’re like, “No, I went to the website, and I printed it out.” And so, I think, you asked about the product mindset, just making sure that everyone feels like a user and takes that mindset. That was just so eye-opening and really helpful for the rest of the time because the frequency with which all I have to do is turn around my computer and be like, “Let’s do this together,” is tremendous. So many amazing stories like that from that time.

The idea that you’re going to bring a product culture into an agency, I’m guessing 10 years ago, that was a much harder road than it is now. Has that changed? I don’t think you’re explaining on a fundamental level what a product manager does anymore. I hope not. But the idea that you’re going to bring that culture to the agency and that they’re going to adopt that culture, that changes the agency itself, right?

It definitely changes the agency. I would say we’ve made a lot of progress, and things are really different. I don’t show up in an agency anymore and have them say, “What is user research?” Well, sometimes “what is human-centered design?” But “what is user research?” is not a question — like “what is the cloud and it’s completely unsecure and I couldn’t use it,” definitely conversations I had to have 10 years ago. You were joking when you said, “What is a product manager?” But it’s a squishy concept. Even in the companies where I used to lead product, it was a squishy concept. So I think explaining what specific roles should do and how they should fit together actually ends up being a lot of our work. And a lot of our work is not just explaining it but showing [it]. So we do the first project together, and it’s like, “What does a product manager do?”

You say it the way I sort of said it, and that’s kind of the same as a project manager. It’s kind of the same as a policy expert. We actually just live the program and live a project and all do it together for six months, a year, 18 months. And that really helps them see, “Oh, these are the ways that it’s really materially different from what came before.” If you go on USAJobs right now and look through all the federal jobs, there are not enough product manager jobs. And if you look at all the people across the government, how many product managers should we have? There are not as many as you would expect given the number of programs and products there are. So there’s still a lot to be done there. Our colleague Jen Pahlka likes to talk a lot about how product managers are kind of a key piece of this and that, I think, still, there’s room to grow.

The tension there that seems really apparent to me is that if you’re the person in charge of the Social Security Administration, the way people experience it is through It might be that the administrator of the Social Security Administration needs to be the product manager because they’re ultimately accountable for the success of the thing that people only experience through a website or an app or something. Is that just me inventing a tension, or is that a real tension? Do those people get that they understand that the user interface of their product is the product?

We like to talk about product CEOs in tech a lot, and I think a lot of exceptional products are built by CEOs who have a real product mindset, who live inside the product, who take customer calls. The way you get to very senior positions in government is quite frequently through a policy process, through policy expertise, which has not a lot to do with execution in government for lots of reasons. Policy is deciding what we’re going to do, and implementation is doing it, and never the two shall spend much time together. So no, I don’t think most agency heads think of themselves… Some of them do, and we’ve come a long way in helping instill that perspective, but that is not how, when they get in seat, they think about things.

I think that is one of the opportunities that we have as USDS. We go into an agency in a very senior position to work directly with the agency leadership on a specific program. So we have done a lot of work with the Social Security Administration, and some of it is bringing them along with, “What should you be asking your team to make good product choices?” — like making sure that they get demos, making sure that they see videos from users of where do users get confused, because not all of us are the target demographic for the services.

Then, taking a further step back and saying, “Your product is not just your website; it’s your website, as you said, and your call center and your in-person service.” And how do we stitch those together? Because, for a variety of reasons, over time, those are all managed by completely different business units with completely different technology systems, so there’s a real need to sort of pull those together and have product mindset from the top down.

I want to come back to just some big-picture Decoder questions for a minute. This is a lot of tasks. The government is vast, it’s sprawling. There are a lot of things to do. I’m sure you have a lot of inbound. How big is the USDS?

USDS is 230 people right now.

Is that big enough?

Definitely not. It’s the biggest we’ve ever been, and I get unfulfillable requests every single week, as you would expect. There is such an incredible need for these capabilities across the government and whether they should all sit in, as you said, a consultancy attached to the White House, is a great question. But the total number of engineers across the civilian government is significantly smaller than at one big tech company. And the number of services and products that we are managing and developing is dramatically larger. So no, there’s a lot of need and tremendous opportunity. We are hiring — agencies are hiring right now. There are thousands of open positions for incredible technical talent across the federal government.

So you’re 230, it’s the biggest you’ve ever been, you’re hiring, and you’re getting bigger. How is the USDS structured?

We are structured a lot like a consultancy also. So we have communities of practice, so verticals, that are all of our engineers, all of our product managers, all of our designers, all of our procurement experts. And then we have people deployed on project teams. Those have their own internal structure but are necessarily temporary, but some of those are three months temporary and some of those are three years temporary. So that varies. And then we also have a significant operational side of the house. It takes a lot of work to… so everybody is on a two-year term or less. I know you spoke with President [Barack] Obama about this. We bring people in, and working in government can be an incredibly hard mental shift.

Over time, it has evolved, and we have a lot more folks who have been here or have heard about it more directly, but we bring people in for a short period of time. Sometimes they decide to stay the entire two years, but that exercise, hiring in government, is challenging. Convincing people is challenging. We have a big talent team that really works to both do recruiting and help us process all of that hiring and manage I think we are on track to do almost 2,000 interviews this year.


So it’s just like a significant operation to get to where we need to be.

Is the management layer of those 230 people, is that on two-year terms as well, or is that more fixed?

Everybody’s on a two-year term except, I guess, technically me.

How do you maintain a culture if everyone’s coming and going, or is that even important to you?

Oh, I think it’s incredibly important. We spend a lot of time together. I mean, I think it’s challenging, and it’s even more challenging in sort of a hybrid world, but people are deeply united by what they came here to do, right? People come here because they are deeply motivated by the mission, want to have a massive impact. USDS, we take people who have a lot of experience, and part of the reason that we’re doing so many interviews is we’re on track to get almost 6,000 applications this year.

We will hire 1 to 2 percent of those people. So this type of being able to drop into an agency, help them from a technical perspective, but also have the exceptional EQ to really help bring the agency along and teach them how to do that. It’s a rare skill, and it requires experience. I think we have folks who have maturity and are deeply united by their love of hard problems, their love of the mission, and their commitment to impact. That is incredibly helpful in keeping everyone sort of working together against that North Star.

There have been a lot of layoffs in tech this year, just a lot of turnover in general. There are more at Spotify this week, as we’re talking. Is that a new pool of applicants or talent for you to go mine? Are you seeing that there’s more interest in government work now?

We have definitely had more applicants this year than in any prior year. I think that’s a function of several factors, frankly, some of what you led with, which is that this is becoming more of a thing and it’s grown every year, but I’m sure that factors into it. There are a lot of incredible people who are free agents now, so that’s an unfortunate fact but also an opportunity for the government as well.

When you’re sitting with your talent team, do you say, “Okay, Meta did it. There are a bunch of Meta people on the street. Let’s go talk to them”? Does that come up, or is that just more ambient?

There is one agency in the federal government, the Office of Personnel Management, and their job is to sort of help the entire government understand how to do hiring in the government way. They make up the rules, and we have worked with them at the end of last year and the beginning of this year to stand up sort of a function that is called Tech to Gov that has pulled together lots of different agencies and has worked to build a common engagement model. We have job fairs where lots of agencies show up, and we have speakers, and people can go meet with people leaving. Tech companies can come and meet with agencies, learn about what they do, understand how it would work.

I think kicking off that initiative, for example, was motivated in part by just the significant availability in the market and the need and coupling those together. It just really felt like there was a moment to kick-start those types of efforts. I wouldn’t say on a day-to-day basis. Like I said, we have 6,000 applications of people who are desperately wanting to come and work. So we do have recruiters, and we keep our eye on what’s going on in the industry, but that is not our main focus.

Other agencies often get criticized for having revolving doors, right? You go work for the thing that might regulate you or might compete with you in the case of the USDS, and then you go back to work for the companies. You’ve had your own investments come under pretty harsh scrutiny because you had some waivers for conflicts of interest along the way. How do you manage those conflicts? Is that a concern of yours?

We are very careful to ensure that folks who come in who have specific interests don’t get into a position that would even have the appearance of impropriety, let alone actually have any influence. I do think there’s a real tension to grapple with, as you know, the tech industry and a lot of companies now pay with equity, and the model of compensation in the world has shifted, and the government ethics regulations, I think, need to catch up and figure out how to think about that because you want people who have expertise in a seat. Figuring out what is truly competitive and what is not, I think there’s an opportunity there, but we’re extremely careful about it because, as you say, I think it would be totally not worth it for anyone to even perceive that that was an issue. This is just such important work to be undertaking.

Let’s talk about that work. You mentioned how big the government is, how many needs there are. This is the classic Decoder question. You have to prioritize where your teams are going to go. How do you make decisions? What’s your framework?

There’s the “how do you decide what projects” framework, and then there’s their general framework. There are so many problems to take on. There’s so much work inbound. I am lucky in that it is not exclusively me. We have other stakeholders who engage in that. I work very closely with the leadership of the Office of Management and Budget, which is the part of the White House in which we sit, and also with the chief of staff’s office at the White House to triage and evaluate the impact of opportunities. Ultimately, we’re extremely focused on, “Where can we go that will have the greatest impact as determined by what will have the greatest impact for the most people in the most significant need?”

That is a combination of: Is this the right type of project for USDS? Does it have a technology component? Is there a need for our skillset? It is those impact metrics, And there is also a component of long-term change. Our impact can be greater if the agency is near a tipping point, where they can move to a place where, over the long term, they will start taking on more of these problems in a different way. We’re also evaluating how ready the agency is to do things differently. Sometimes, there’s no chance. Sometimes, even if the agency is not quite ready to think about things differently, you have to go do the work because it’s so critical. But we definitely… that is an important factor for us as well.

Does Joe Biden ever send you a screenshot of some horrible agency website and just be like, “Fix it”?

Not him personally, but the chief of staff definitely.

Jeff Bezos famously just sends question mark emails to his deputies. Does that come down from the chief of staff for the White House? Like, “This is broken, and we need to fix it.”

Yeah, sometimes.

Give me an example.

It’s less frequent that we would get a screenshot and more frequent that we would get a letter from a constituent or a letter from a user. It’s more like you get tickets, and the way that people submit tickets to leadership in this context is you write a letter to the White House or to the president. We have engaged in significant work at the Department of Veterans Affairs over the last nine years. And it’s some combination of letters and engagement from constituents, letters that go to the Hill that then come to the White House, and frankly, media. I will also sometimes get a news story that says, “Is this really this big of a problem, and how do we fix it?”

We have been engaged at the Department of Veterans Affairs for the last nine years. At the beginning of that time, the VA website helped you understand who was the secretary of the VA and what is the structure of the VA. And who cares? Not any consumer who’s like, “Where is my claim? How do I make an appointment with my doctor? Which facility are they actually in when I’m going to see my doctor?”  — real transactional questions that any normal consumer would have. This was at the same time as there were a variety… the VA was sort of in crisis. They were having trouble scheduling veterans. There were a lot of challenges with data coming over from the DOD and people figuring out where their benefits claims were and their benefit appeals were. And so we got a lot of inbound on those topics, and it helped us really identify areas to prioritize what the biggest pain points were.

Because it was clear that we needed to have a wholesale transformation of how that agency thought about integrating all these services. But you’ve got to start somewhere. In part, those are really helpful for prioritizing, both that we needed to focus on issues at the VA and then also which of those issues to take on first, so we started that work. Initially, we launched a compendium site, a beta site, that was called It was different from the VA website. It only handled a couple of the things that you needed to do, but it was the place that we could direct you to do your benefit appeals. And then over time built out the functionalities of that, eventually moved that to be the main site. Now, if you go to, it’s very transactional-oriented. It’s like, “How can we serve you? What do you need to do? What are you trying to do?”

And then, more recently, [we] have additionally supported the VA in building out their own mobile app, which, as your demographic will understand, building an app is the thing that executives are always like, “Should we have an app?” And you’re like, “Oh.” Your bar has to be incredibly high before you decide. But similarly, we got a lot of user feedback. We looked at the functionality really hard. The CTO there is a former USDS-er, and he’s incredible. And ultimately, they decided it was worth the investment and critically important. Now, it has over half a million monthly actives. It has over a million users. It’s like 4.8 stars in the app store. I think [it has] something like 100,000 five-star reviews. Looking back at where everything was nine years ago, it is impossible to imagine that that is where we are now. But that is what agency transformation looks like.

And it shows. In that time, trust metrics across the VA — which is not exclusively due to their digital experience, but by the way, in the process of integrating those on the front end, you also have to build out your business processes and figure out how to make those more seamless — has gone up by 20 points. And that’s huge for a critical demographic.

This is an amazing example because there are layers upon layers here, and we can just go through them. The first one is, “Boy, there’s a bunch of old computer systems and databases that need to get integrated so that you can use a website to process a claim with a hospital and talk to the DOD.” That just seems intractable. At any company, you’re like, “We have five different databases. One of them is run on COBOL. Fix it.” It’s already intractable. Our company is trying to migrate to WordPress, it’s intractable. Even that piece of it is hard.

Then there’s, “We need to build some front-end user experiences to replace our, like, ‘Here’s a picture of the secretary and a phone number for the main contact line’ into an application that you can transact with and that is reliable.” And then there’s, “We’re going to care about ratings and reviews in the app store,” which is all the way down the line.

The first part of it, just that we’re going to bring all the databases together. Did you just set a team on that and tell the VA, “Get out of our way. We’re going to fix it”? Or did they have a team? How did that work?

Oh, I love this question. No, I mean, we don’t do anything that is not in deep partnership with a bunch of different partners. We are 230 people; at that time, I think we were 60 or 80 people. But the VA has a huge IT department of their own and a huge operations and customer service department of their own. And most of the build work across government, almost all of the build work across government, is done by contractors. There’s also a procurement shop at the VA that is letting contracts to vendors to build all these capabilities, to update the databases, to do the user research, and build the front end. We deploy small teams; you would not believe how small on some level the team can be. But a dozen people to go and help the agency really strategize “What is our roadmap?”

You said unifying all the databases. I’m sure all of your listeners are like, “Here’s what we should never do.” Just like a database unification project that has no… those never go well. You have to figure out what use cases are you going to focus on fixing first. A lot of our initial work is getting everyone around the table to look at the data on what users are doing, where we’re getting stuck, what are the bottlenecks, and decide what problems we want to take on first. It takes… you eat the elephant one bite at a time. And all through that, we are working closely with the existing tech teams.

So some of this is bringing in the executives, the existing tech teams that are focused on those parts of the experience, the procurement folks, the contractors, and helping everybody shift what their objectives are: how they work together; how we make this more closed loop so that let’s adjust our methodology to say that the outcomes and some types of user ratings should factor in to how contractors are evaluated, which will change how they decide to do their user research and change how they decide to do their monitoring. Let’s build a process where we start to look at the Google Analytics from utilization, and that builds into how we prioritize next things. Oh, first we have to put in Google Analytics so we can do that. It’s a very centralized… How do we help catalyze all of these partners to work together in concert? Our goal is not… we very rarely hands-on keyboard build the app. Well, really any of it. We spend most of our time with a combination. There are very real technical problems, but a lot of nontechnical problems, too. And it’s really, “How do you catalyze all these partners to work together?”

The work is done by the contractors, and we’re here to help the agency figure out how to most effectively draw the roadmap, prioritize the work, manage the technical roadmap, and also the product roadmap. How do you build that into the contracts and all of that? It’s a very highly leveraged model in that way.

There’s a real tendency to take technology for granted, to believe that you can just sort of deploy it, or that it will continue running, or that you don’t need to update it once you have it. I can see that in the government maybe most of all, particularly when I listen to our elected officials, that you can just wave your hand at technology and make it go. Something like the VA. You get the VA wrong, people don’t win reelection. There’s a lot of political pressure around a process like that to make it better instantly. How do you manage the pressure that comes with people taking it for granted or demanding a quick fix against, “Okay, we have to do this invisible, very difficult database migration in order to get to the place that you want”? 

That’s one of the tremendous challenges. You’re hitting on the head a couple of the fundamental issues that we try and work with agencies, but stakeholders across the board, to sort through, but some of them are some of the immovable objects. So one, you asked, “What do product managers do?” The desire for new quick hits and to do something new in government rather than fixing what’s old is huge. And as any product-driven company would tell you, if you just keep launching new things, that’s not the right way to satisfy your users in a lot of different ways. Figuring out how you’re going to knit those together, it gets confusing. You get proliferation, and as you said, it means you’re not paying enough attention to your legacy products.

I think we’re having the conversation now, which is incredibly important. I helped write a paper a few years ago for the Defense Innovation Board, the title of which ended up being “Software Is Never Done,” which is basically like, “Stop thinking that you build it and then you put it on an operations and maintenance contract, and you never have to modify it again.” That’s a real good way to waste a lot of money over time. The conversation is there, and I think the executives who are moving into leadership positions in agencies are beginning to have more experience in managing ongoing investment in operations and IT. But the way government budgeting works, it is challenging, and this is one of the key things that we are constantly working on, that we’re working with our colleagues at the Office of Management and Budget about, right? How do we structure this budget to make sure that we can have the right degree of investment over time? It’s a real challenge.

There’s also just the general politics of it. The Obama administration created this agency. Trump cut the budget, Biden raised the budget. The Congressional Republicans, if they accomplish anything, they will cut the budget. There’s an election coming. Who knows what will happen? How do you manage through that? Do you assume that there’ll be some fixed number you can work with, and then you’ll have an increase or decrease, or are you just waiting to see what happens?

You mean for USDS, or do you mean for an agency that is trying to deliver services in a consistent fashion despite fluctuating budgets?

It sounds like the same problem across the board, but for USDS?

I think USDS has, to a certain extent, more flexibility. We have a lot of demand, and I think we’re highly leveraged when we go into an agency. We help them become more efficient and more effective. At the end of the day, I think USDS can scale up and down and just take on more and less programs depending on the priorities across the board.

Do you see your job as political? Do you go meet with various congressional leaders and whoever else and prove the point that this is valuable, or are you insulated from that?

We definitely do. Ultimately, our budget comes from Congress, and just like any startup CEO, part of my job is to go explain what our ROI is and show—

Do you do the graphs that are up and to the right forever? You’re like, “Here’s the hockey stick” to [Utah Senator] Mike Lee?

No. [Laughs] I was trying to think of a funny thing to say about that, but I don’t even like... I guess the thing I could do that with is the number of inbound requests we get, which does have some of that feature. 

So what’s your sell? You go meet with some Democratic or Republican senator, [and] you’re like, “Hey, it’s budget time. Here’s our number of inbound requests from across the government. Here’s the new, here’s the new, here’s the new FAFSA form. Here’s the stuff we make. Look at how good it is and how much more of it people want.” Is that the sell to the rest of the government?

It’s, “Look at how good it is.” It’s, frankly, continuing to invest in your technology has a bunch of benefits, and doing it intelligently. So one, we help agencies more effectively manage how they spend on technology. I’ll be honest, as you know, in my private sector life, I had lots of contracts, too. It takes a high degree of expertise to effectively manage contracts to get what you want, so we help agencies figure out how to most effectively spend their money and use their money. And a tech roadmap that is thoughtful is obviously much more cost-effective than one that has a little bit less expertise in it. So I think there’s a cost-effectiveness piece.

There’s also a broader cost-effectiveness opportunity, which is, part of technology is to make us all more productive. One of the huge drivers of cost at a lot of agencies is a very large number of calls to the call center, for example, or people processing things where we have a backlog. And I think those are political issues. You have a backlog of claims, that’s a problem. And one of the ways that you accelerate that is by making all of those claims processors more effective and helping them be more efficient about processing those claims. So I think there are multiple pieces where technology is helpful, and we are one of the most effective instruments of making technology more effective in government, both in a user-facing way and a back-end way.

So I want to bring all these threads together, so this is a perfect segue. A thing that I think about a lot is we’ve talked about product managers and user research and all this stuff. If you go to a young person, and you’re like, “The fastest way to get what you want is open this app, and the app is good and it’ll be effective and you can just push some buttons.” Most young people are very digitally native now, they’re like, “Great, I don’t want to talk to a person. I don’t even want to make a phone call. I will just figure out the metaphors embedded in this user interface, and I’ll get what I want done.” And the kids know how to use Snapchat. The greatest example of “I will navigate this bizarre user experience to get what I want,” and that thing is a success.

I go to my parents or older folks and say, “You need to figure out this computer to get what you want from the government,” that is just crashing into a brick wall of antipathy because they’ve had bad experiences with that stuff or they think it won’t work or they think it’ll be too hard. That is a huge discrepancy in your user base in a way that I think Facebook is like, “Whatever. We want the old people to go away. We’re focused relentlessly on younger people forever.” They have the luxury of being able to do that. And then you have a bunch of politicians that write your budget, who, in this country, are almost exclusively older people. How do you manage that? If the interface for getting good government services becomes more and more like phones or apps, we’re going to leave a lot of people behind. And then most importantly, the people who write the checks for it are also going to be angry at it.

I think you are mixing together a few different factors, and so maybe [we need] to tease those apart. I spent a lot of time building tools for elderly people to use, and frankly, they are psyched. They’re on Facebook.

Oh, are they?

I was entertained when you said that because I said, “The kids actually aren’t on Facebook anymore, and it’s only the older demographics that are on Facebook, and a lot of them are on Facebook—”

Right but Facebook… Meta thinks this is an existential problem for Meta’s business. 

Well, they don’t spend money, and they’re not advertising... So it’s a problem for their business model. But what it proves is that that generation is totally willing to be digitally engaged. And so, if your parents say, “I don’t want to deal with the government in that way,” I guess I would say I think that’s an expectations problem because they have had bad experiences in the past, and it is the case that maybe someone who is younger and more digitally native would be able to figure out a bad experience. But the truth is that we need to raise the bar on the actual experiences such that all these demographics can engage online.

I have seen enough — I ran a Medicare business, we were texting constantly with our members. They’re excited to be engaged in technical ways that are frankly less expensive than walking into a Social Security office. Obviously, some people really need to go to a Social Security office, and we should have those options available for those people, but they also find it inconvenient often to go do that. You have to find transportation. For people who are still working, you have to take time off work. I think there’s a tremendous opportunity to meet those people where they are, and I think a significant proportion of them are willing to be met digitally, but we need to raise the bar on what our digital experience is, how simple it is, so that they can engage in that way.

And to the last point that you were saying, we have to slowly… I don’t think you can change people’s culture and expectations overnight, but we have to have a slow success rate, a relentless set of successful programs and successful applications so that people aren’t deterred just by the idea of it. So I think the opportunity to engage with older populations digitally is still extremely salient. And we see that. Medicare has been rolling out new tools for shopping for Medicare plans, new tools for looking at and comparing Part D plans — very high engagement rates. So I think it’s just about making the tools simple and appropriate.

Yeah, let me ask that question a different way. I appreciate the pushback there, and I don’t disagree with it, but I think for a lot of folks, the online environments they find themselves in are actively hostile or actively trying to scam them or actively pushing them toward advertising or a transaction of some kind. You just see it all over the place, and then you have to say, “Well, I need you to trust that often the same sort of format on your phone in an app, the same kind of design language, is going to lead you to good outcomes from the government,” where maybe you have already a distrust of what might happen there. That seems like a real problem. If people open their email, and it’s full of spam, and then you’re like, “Click this other button, and this will get you your health benefits,” you’re operating in that paradigm whether you want to or not.

And I think, in general, and this is a huge generalization — I’m sure I’m going to get emails about it, but I’m just going to do it for the sake of the argument. In general, the audiences that have come up in the paradigm are used to navigating it, and the audiences that have not are struggling with it. And obviously people are different on different parts of the scale. I think Gen Z is actually pretty bad at technology, just like they’re bad at file systems. They haven’t grown up with them. So there’s my out, don’t send me the emails, but that seems like a real problem. The paradigm of ultra-capitalistic user interfaces on the internet and then the government are in conflict. How do you manage that conflict?

That is a very astute question, and it’s a complicated question to answer. We have rolled out quite recently… the office of the federal CIO, by the way, Clare Martorana, the current federal CIO, also used to be a USDS-er, which is part of our plan to longer-term change government is have people stay in key roles. So the federal CIO’s office recently rolled out digital experience guidance, and part of the goal of that is to build in a degree of uniformity, consistency of navigation and language, and overall branding that allows us to be extremely clear about, “What is a government website? What are the things that you should trust? Why?” to be very clear and still accessible.

It also shouldn’t be the case, to your point, that it has to have complicated legalese for you to believe that it’s the government. We should build other ways for you to say, “This is a trusted service,” and it is still simple for you to use. And so to build brand guidelines and usage guidelines that make it more clear when you are on a trusted website of the United States government and when you are doing key transactions. So there’s a web design system that we are standardizing on for the whole government. There are branding materials and a navigational framework and other things that we’re working on to make it more clear that this is a government thing that you should trust.

I also say the other half of what you’re saying is that there are challenges in product development because we have to hold ourselves to much higher standards. All of the shady sharing your data between apps and applications and websites, we don’t do that. And we hold our vendors to higher standards when they’re delivering services for our products, our federal government’s products, which says that any data that comes through this, you can’t use it for other purposes, either. And that leads to a different experience sometimes than what people have come to expect. But it is what people vote for and in their hearts know that they want, but we are actually living that walk, walking the walk that they want rather than doing a whole bunch of other things. So I think there is more complexity. And this is where some of the really interesting and hard problems come in: how do you deliver an exceptional user experience but do it in a way that meets your moral standards and the privacy and equity expectations that the public has? So I think it’s a great question.

When you mentioned the monthly actives for the VA app, my mind immediately goes to how tech companies would boost their monthly actives. Have you ever thought about sending just a super thirsty post notification? You have notifications waiting for you if you open the app. Is that a game that you would play?

To what end? 

To boost the monthly actives numbers so that when you go to Congress, you can say, “Look at the monthly actives.” That’s what a tech company CEO would do. They would boost their monthly actives so they can tell their investors in the next earning call that the monthly actives went up.

To go back to what you were saying earlier about lawmakers, they’re not.

It’s a radio show. [Laughs] I just want to point out Mina definitely just acknowledged that is true. She nodded her head.

[Laughs] As a recovering venture capitalist, I have a lot of commentary on that. To take it back to a question you asked earlier and a point you made earlier, lawmakers, at the end of the day, their number one metric that they’re deciding on things is not monthly actives. It is a good piece of a pitch, and it helps us tell a story, but at the end of the day, they care about what they’re hearing from their constituents. So I guess that’s my actual answer is that I think, in a good way, mostly we’re all sort of focused on, “Okay, what is the actual experience rather than what proxy metrics have we cooked up so that we can have a quick pitch deck and look at 42 companies a week.” So I guess my answer would be, we try to be much more holistic about it and look at sort of the overall picture of what’s actually happening and whether our metrics are reflective of that is one of the big questions that we’re... What are the most appropriate metrics for us to be worried about?

Let me shift gears a little bit. I’m really focused on the idea that the government should have an interface and that you can get a lot of value from the government if you could only figure out how to use it. And that’s a lot of what your agency is trying to do. You’re trying to unlock it so people can use the government. And right now, that is expressed through websites and apps, and I think that that has a long roadmap to go.

But you mentioned, President Obama was on the show, we were talking about AI. There’s a lot of government interest in AI, not just to regulate it and bring it under control and make sure it doesn’t kill us all but also because if you could just talk to the government and it could provide you services, that would be much better than having to navigate some library of apps. Is that on your mind? Is that on your roadmap that we should have some sort of LLM-powered interface to the government? Because that seems like the next turn for a lot of things. Maybe not LLM, just some sort of AI system. I’m just thinking about the LLM just lying to you about your Social Security benefits.

I do love this question. Going back to your last question, we do hold ourselves to a higher standard. We’re not just going to try random stuff, and as you say, give people incorrect information about their Social Security benefits. So is it on the roadmap to understand how to incorporate AI to more effectively deliver all these services? Absolutely. But do we need to get to a place where we have a degree of reliability, consistency, frankly, equity in all of these tools and our ability to control what they do and understand and make sure that they’re going to operate exactly correctly? Yes.

If you go to, yes, there’s a recruiting thing, but there’s also an inventory of use cases, which lays out all of, well, the majority of the ways that agencies are currently exploring using AI, already using AI. And that’s growing. But a lot of it is pilot programs to figure out, “How does this influence how we communicate with the public?” etc. So I absolutely think that that is, as you say, the future of lots of companies and the future of technology and government, but we’re going to have a really high bar for when and how that gets used. And frankly, that’s one of the reasons that the recruiting question about this is so important. In order to do a great job of utilizing those tools but also evaluating how we’re doing on all of that, it requires us to have the people inside government who have that expertise, who know how to implement systems, who know how to tune them and test them and build them out.

This brings to mind just a split. You have the ultimate Microsoft Office problem: you have a core set of users who need to do one thing, and then I’m certain you have people who expect the features to always be there across all these products no matter what, and you can never unship a feature. Do you think about how to balance between the power user of the Social Security website and the person who’s just showing up on the first day and is a little confused and needs help?

I don’t know that we have quite that problem. There’s definitely a long tail of programs that are a smaller number of people, and we definitely have to... These programs are critically important to people. That question made me laugh. I don’t know who’s the power user of the Social Security website because there’s a finite number of things that you need to do.

But there’s one person who’s like, “I know there’s a form on this website. I know how to go find it. If that moves, it’s all over. I’m mad at my elected official.” That’s a weird outcome. Now, you can’t reorganize the website. Changing a website taxonomy, as I well know, is a very complicated idea. [Laughs]  That’s what I mean by power users. There are people who have expectations of how it will work, and they rely on those expectations, and they’re not going to be motivated to figure out a new way. And then you have people who maybe you can bring on more people or younger people or different people or attract a broader audience to the services if you make it simpler. And those things seem... That’s what I mean by a Microsoft Office problem. I know the Office PM would love to get rid of most of those icons, but they can’t.

I think we’re always going to have an extremely long tail of use cases. We don’t have the problem in that I don’t think that we absolutely need to get rid of all of the long tail. I think that there’s some product mindset where you just decide that you’re going to fire a certain percentage of your customers because maintenance for all of those features is too much of a pain, and we can’t and don’t want to do that. So we will maintain those features, and they may not be as perfectly updated as they should be, but we don’t have the same drive also to manage expenses to get rid of those programs.

Those programs are critical programs that have people that they serve and budgets that are allocated to them, so we don’t have the same motivation to prune. I do think simplifying and making sure that we’re not dropping critical functionality while we’re rolling things together is very important. Ensuring if we build something that is “No Wrong Door” strategy or a simplified front door for multiple programs. We do have to make sure that if there are people who would enter the program through a different channel… There are people who go to get SBA disaster loans from federal emergencies and also from state emergencies, so making sure that we keep available access points for people who come from different scenarios is definitely critically important.

I asked that question in the context of AI because you could see how the AI interface just solves that problem, especially for the government where you’re saying, “Actually your main interface should be a chatbot or your main interface should be just asking Sir, and we will begin the process of walking you through what other app services agencies you need to talk to.” And it’s almost like an ombudsman role, to use that term. I don’t know if that’s exactly right, but it’s almost just a guide to the government that sits next to you. Am I just making that up, or do you see that as an opportunity to just simplify people’s relationship to the government? Because the idea that the government has an interface, I think is pretty radical. And the idea that the interface would actually be openly helpful is even more radical than that.

A different piece of the digital experience guidance is around search and even just improving information architecture on current tools and then improving our ability to ensure that search results, and not just central Google Search, but each agency has its own Google box. What you’re talking about is the continued natural evolution of that to a certain extent. And absolutely at the point that those tools are ready for that, I think it’s an exciting vision, and I agree with you.

Because that’s where search is going. That’s where Google... Literally, the day before we are talking, Google announced Gemini, which is going to be the future of Google Search. I don’t know if it’s any good, but they say it’s going to be the future of Google Search. And you can see how that maps directly to “You ask it a question, and it tells you the answer.” That is basically all people want from the government. I ask a question, and it tells me what services I have access to. I am just curious, is that a thing that has to come from the top down? Is that, President Kennedy needs to announce in the State of the Union that we’re going to have AI government in the next 15 years? Or is it, “Okay, we’ve got a bunch of product manager-type folks at a bunch of agencies, and we can baby step them toward that outcome”?

I think the truth is a little bit more complicated of what people need and how they engage with government. I think it’s a very cool and exciting vision, and I agree with you that that is where search and a lot of things are going, and I think that we can continue making progress on that.

But a lot of people need to transact with the government, which is a pretty different engagement model than just finding out things. And a lot of people are already... Most people are already in a cadence of transaction with the government, and what they really want is to continue to transact with the government but to do as little paperwork as possible for that to continue to be valid. I think for just finding out things, there’s a real, I think what you’re saying makes sense, but I think there’s also a lot of things to be done even now. Another program that we’ve been working on is automated renewals for Medicaid. There are some other opportunities. Medicate is a state-administered program. States have to determine every month if the person is still eligible for Medicaid. By law as part of the ACA [Affordable Care Act], states have to first look at the data that they have available from digital sources and other places to determine if they already have enough information to know if you’re eligible. There are services where we can verify your income, and it can tell. Some of our work has just been to improve how that automated check works.

You don’t have to be recent... Instead of a state sending you an envelope that requires you to resubmit a pay stub every month, how can we just not bug you just the way your health insurance doesn’t ask you every month if you’re still working at that company because there’s an automated way for them to evaluate that? It’s just like having the normal transactions and services that you rely on work in the background instead of bothering you on paper for you to go into an office or submit something. I think, yes, that is a piece of it, but there’s so much more operations that the government is engaged into, and we can really do so much to decrease the burden on that and take it off of the public and just make things happen in the background.

I’ve really enjoyed this conversation because it is both: the problems seem very challenging, and then it is also just such a different perspective on how to think about technology problems and user experiences than, I don’t know, trying to sell ads to people. What kind of people are you looking for? You said you’re recruiting. You said there’s a big influx. What kind of people are you looking for as you think about, “Okay, we have all these problems to solve”?

Absolutely. USDS is hiring, and in addition, lots of my colleagues across government are hiring. I want to put in a plug for the Presidential Innovation Fellows, the US Digital Corps, but also every single agency that I work with is currently hiring for excellent technologists who want to work on important problems across all of those boards but, in particular, at USDS. We’re looking for product managers, designers, engineers, data scientists who have experience in production in real-life environments and complicated environments, who are excited to take on hard problems that have a technical component but also nontechnical, like a people and an organizational component, who have exceptional EQ so that they can work closely with all of our federal partners, and who just really want to work on the most important and hardest problems for massive impact.

I was just talking about that auto-renewal thing on Medicaid. This is healthcare so that people can go see doctors. And in the last three months, we have made it possible for over 3 million people to have auto-renewed Medicaid that they would’ve fallen off of otherwise. We have helped prevent over 500,000 people from accidentally getting removed due to a different bug. The problem and the impact should be deeply motivating, and then it’s just an incredible group of people to work with — people who are excited about increasing trust in government, who are excited to solve these problems in a long-term way.

Mina, thank you so much for joining us today. This was really fun. I feel like I learned a lot about computers and government.

Likewise. Thank you so much. These have been excellent questions. Really appreciate it.

Decoder with Nilay Patel /

A podcast about big ideas and other problems.