What the Debate About Government Efficiency is Missing

Marina Nitze: I remember very clearly doing user research with veterans over 10 years ago, and the best experience they wanted from the VA in terms of benefits delivery was no experience. They wanted to separate from the military and have their pension and their disability claim. They did not want to get 50 pieces of mail and get 5 phone calls and come in person. Nobody wanted to do that.

[Theme music fades in]

Amanda Renteria: Welcome back to The Government Fix. I’m your host, Amanda Renteria. I’ve worked on Capitol Hill, in the classroom, on Wall Street, and now I’m the CEO of Code for America, an organization focused on using tech to improve public services and make government work well for everyone.

[Music plays]

There’s an important conversation happening today with declining resources and states struggling to keep up with what people need. Folks are asking, “How do we make government more efficient? How do we streamline processes? How do we innovate to stretch the resources just a bit further?” Today, I’m joined by two guests who are experts in efficiency. Marina Nitze is a crisis engineer who helps organizations solve complex challenges. She’s particularly focused on improving America’s child welfare system. She’s co-author of Hack Your Bureaucracy and the newly released Crisis Engineering. The book draws on the rescue of healthcare.gov, wildfire response, and pandemic logistics to lay out a framework for how complex systems fail and how to fix them under pressure.

David Eaves has been working in the efficiency space since 2009, where he was an advisor to the Office of the Mayor of Vancouver. He helped draft the open motion that created one of the first open data portals in Canada. He went on to advise the Canadian government on open data policy and taught at the Kennedy School at Harvard. Today, he is at the University College of London, co-leading their Institute for Innovation and Public Purpose while advising governments on civic tech all around the world.

Before we jump in with David and Marina, I want to flag that you’ll hear the phrase civic tech a few times in this interview. Civic tech is a term for how we use digital tools and platforms to improve how people interact with government and public institutions, which is what Code for America is all about. Marina and David are civic tech experts, and I want to welcome them to our podcast.

[Music transition]

We are so excited to have you here. I’m excited to be in this conversation with you all because I feel like you are two people I have learned so much from in being in this space, and so it’s exciting to have both of you together. Let me just start by saying thank you and welcome to the show. Both of you are experts in efficiency at a time when that word has become quite a buzzword in Washington. I want to start off with just that central idea and what you actually think that means, how you’ve seen it not work, whether it’s even true. Marina, I’ll start with you.

Marina: My first book, which not a lot of people know that I wrote, was Business Efficiency for Dummies, part of the Dummies series. I think about efficiency a lot, and I think efficiency needs an outcomes goal attached to it. You can obliterate steps in a process, or forms, or policies. If you’re not doing it in service of getting more people healthcare, getting more children education, then efficiency for efficiency’s sake is not the goal. Certainly, often efficiency will be in service of that outcome, but it’s not the end goal.

Amanda: David, you’ve seen this movement, not just what happened here in the United States with DOGE, but in other countries around the world, too. How has the rest of the world interpreted this idea of efficiency?

David Eaves: Yes. I really want to double-click on Marina’s excellent comment, because I think part of it is efficiency for who? A lot of administrative burdens arise because people are making things more efficient for an internal actor, but they’re actually outsourcing or pushing the work on to the end user or to the citizen. They’re on paper, inside the government, the system is more efficient. It’s easier for us, but actually, the total amount of work required has maybe not changed or maybe has even increased, but it’s been externalized onto an outside actor.

The other thing that efficiency does is when you’re becoming more efficient and you’re eliminating things, there’s things that you’re eliminating that you are unaware of. Often, I like to explain to people, like, why do we have the role of product manager? Part of it is websites can be a lot more efficient than coming in, in person. When people came in person, we met our users on a regular basis. We interact with them. We saw what their pains and their trials were. Then the moment they got abstracted into a website, suddenly, all of that got lost. The system was more efficient, but actually, there was this dark matter that was lost that was really important. Then we had to go create the role of product manager to take what was 10% of everybody’s job, so that they’re knowing what users are doing and what the edge cases are, and then turn it to 100% of one person’s job.

Amanda: I love those tangible examples. I think because people are getting a chance in this conversation to see under the hood, it’s helpful because so often we say, “Well, if we do spare you from coming in, that also means we lose that chance to have that emotional and human contact.” Even more important as things move to more AI, more online, more from your phone, in your hand, but not necessarily with a person. Marina, you’ve said you can use bureaucracy to fight bureaucracy rather than trying to find exemptions or exceptions to rules. Can you explain what that means in practice and why skipping the rules usually backfires?

Marina: Gosh, when I was the chief technology officer of the VA, every time I hired a new person into the digital service, their first bright idea was, “What if we just get a waiver that says that none of the rules apply to us?” Spoiler alert, all the way up to President Obama, such a waiver did not exist. We had this really incredible example in foster care where I’ve spent a lot of my time since I was at the VA in 2017. The way that foster parents get financial support, daycare, all these other benefits is through being licensed. That’s the term that we use.

That sounds all well and good if you are someone that is called to be a foster parent and you want to go through the process and your family decides over a course of months, but it is not all well and good when you’re a grandma, and you get a call at two o’clock in the morning to suddenly take in your three or four grandchildren. In that group, only about 20% of them were ever able to make it through this licensing process. We’d leave the kids with you, totally safe to leave them with you for years, but we wouldn’t give you any financial support until you could navigate this Byzantine bureaucratic process.

I think there was a temptation to say, “Can we blow the whole thing up?” What we did instead was we redesigned it from the core. We worked with every single state. We took their current forms, their current policies. We made a new Frankenstein’s monster, if you will, of what a really usable, family-friendly, and also, to David’s point, worker-friendly process. We didn’t add efficiency on one side and take it away from the other. We really tried to have a balance.

This went into effect in December of 2023, and we already have half the country alive with the other half of the country fully publicly committed to adopting it. It’s been totally nonpartisan. The Biden administration was 100% behind it. The current team at ACF is 100% behind it. I think that’s such a great example because rather than trying to sidestep all the complications of foster parent licensing, we redesigned it from within. We used the same terms. We used the same forms, but then made them actually work for people. That made it very, very easy for the agencies to take it up.

Whereas, we’ve seen other innovations where it’s a brand-new thing, and agencies don’t even have the muscle or the line item or the lawyer to figure out how to even implement it. They leave millions of dollars on the table or millions of potential benefits on the table because the bureaucracy doesn’t have a path for that. If you can use an existing bureaucratic method to make the thing that you want to happen, the easiest, lowest-risk thing for everybody to do, that, to me, is the win.

Amanda: Then there’s the question of how do you have those kinds of examples and spread it across the country? David, you’ve been on the forefront of seeing that, of helping different governments figure out how to make change. How important has it been, those quick wins or that sharing? Is it happening enough? What do you see as the– I don’t want to say trick, but the way to get folks to look at best practice and then employ them in their own context, or share those lessons over the course of your time talking and working with governments?

David: I feel like on the tech side, we get over-obsessed with building the capacity without thinking about what is the pump of value we’re actually going to create? Because if you don’t create that, you won’t keep getting told you can do this more and more and more. I think one of the real challenges for governments is actually have a very short window to figure out where they’re going to place bets to try to solve something within a timeframe that’s going to get them enough public support to do the next bet and the next bet and the next bet.

In fact, the most successful teams I’ve seen are working when they’re in opposition. They were not in government. They were the opposition party, but an election was about a year out. It was a year to a year and a half out. They were like, “We think there’s a chance we’re going to win this election.” They’re like, “What are the things we’re going to do?” We had very clear set of criteria about what were the criteria that a project had to satisfy in order to be not just good to do, but one that will build public value, that will increase our authorizing environment.

Amanda: It’s an interesting idea because the public basically sees when government’s there. I think for anyone that’s worked in government, man, everyone thinks it’s slow, but it’s really time-limited. You got to get in, and then you have to figure out the place, and then you got to make these things work. A lot of what you’re talking about is, what are organizations, what are civic tech, what are the thinkers that are out there, what are folks like you guys doing in preparation for it? How do you, in those moments, keep everyone together when it’s a year and a half away, might or might not happen? What have you seen be successful in order to keep people at the table really thinking through it?

Marina: When it comes to building the coalitions, I’m a huge believer in positive peer pressure in the public sector. Can you be like, “Hey, you’re one of the last seven states that’s doing this”? Or, “Did you know that Texas is doing this, California, and you’re still not?” That can be really, really powerful. We all seem to think that we have to reinvent from the beginning.

To your bigger question about how do you keep the team together toward a goal over a year and a half that you may not even hit, something that I think worked well at the VA that I did by accident with my team was we literally had a list on the wall that went from quick wins to moonshots. We tried to make sure we always had something in every box on that grid. You get that sense of forward motion with the “quick wins,” which as we all know in the public sector, could still be a year.

The flip side is I so often see people say, “Well, elections are coming up, so we’re not going to start that project that’s going to take five years.” Well, the project that’s going to take five years always needs to start sometime, and we need to dedicate that time to say, “All right, this is actually not even going to see the light of day until after I’m gone, but I still need to lay this groundwork now, and I need to make space for that.”

David: I find that particularly in the United States, that there’s this notion of real divisiveness. I remember having a colleague I advised working for a government in a country I don’t really want to name, and they were so confident that all the things they’d built would continue when the election was over, and they knew they were going to lose. They were like, “Everything we built is so valuable.” On the very first day of the next regime, they literally pulled all the plugs out of all the servers and shut everything down.

I was like, “You have to be reaching out and connecting with people on the other side, because they’re not going to agree with you on everything. You have to find out where you have common ground so that you can each be building off of each other. There’ll be things that you don’t like about each other and things that you don’t do, but that common ground is the foundation that everybody collectively needs to be moving everything forward. What is your opposition strategy? How are you engaging with them?”

Amanda: It’s so important to recognize what really are the combined or aligned incentives all the way from the beginning to the end. I think when people think about working with government, they forget this exact idea, that time spent in that relationship-building, and why trust is so important, it means you actually have to care to get to know somebody before you talk to them. Everyone has a story and a reason.

One of the things I love about public service so much is that it is a diverse set of reasons why people come in and do this work. To know that just helps you build that trust. I think you’re exactly right. The directness of where are we not aligned gets there faster, because eventually in this complication of trying to make something really work in a bureaucracy, you’re going to bump up against that if you’re not super clear about it.

Marina: I wrote a book called Hack Your Bureaucracy, and the chapter tactic that people remember the most is to stab people in the chest to as opposed to backstab them. Right? It’s like, “Hey, I’m going to disagree with you. We’re going to go in front of the Secretary. I’m going to say that I think this number is wrong. On slide 13, I’m going to propose the opposite of what you’re proposing.” It really builds an unexpected working trust. Then the 99 other times that you are aligned, you’re not worried that they’re going to backstab you, because you’ve been really clear and upfront.

Amanda: These things used to happen behind closed doors. Everything feels pretty public now, [chuckles] wherever you are. It does seem like we’re, as a society, as public servants, or whatever you’re doing, learning how to have those open disagreements, and trying to figure it out together probably makes a lot more sense these days.

David: The other thing I just want to note is you change, and people change. That’s part of what these conversations are about. I was remembering, I used to do a lot of work with 18F and the Presidential Innovation Fellows. It got to this place where it had this rubric that there were three archetype of PIFs, Presidential Innovation Fellows. One was the patriot, which is like, “I’m never going to be a Marine. This is as close as I’m going to come to serving my country.” The other was the careerist who was like, “I didn’t know what else to do with my life, and this was a good choice.” Then the third was the evangelist that was like, “I believe in open source,” or, “I believe in Agile. I actually don’t care if the work I do is successful. As long as everybody learns Agile, I have succeeded.”

Those three archetypes would actually fight with each other, because the people who were there who were the patriots, the professionals, were like, “Who are they? They’re eating up a patriot spot. They don’t even really care to be here. How dare they?” Then the evangelists were like, “We don’t actually care if anything works as long as our ideas succeed.” Everybody’s like, “You’re sabotaging everything.”

What was more important is people change. Sometimes people who were patriots became careerists, like, “Government’s too hard. I got to get out, so I want to have something on my resume, and I’m out of here.” Sometimes careerists were like, “Whoa, this has changed my life. I am now a patriot, and I’m here for forever.” You don’t know how you, too, will change in that. You will evolve. That’s why these conversations matter, not because your job is to persuade everybody else, but sometimes people will persuade you, and sometimes life will evolve, and you will evolve, and you’ll become different.

Amanda: Gosh, I love that visual of that happening, because I think people don’t necessarily see that. The reality of moving something through government requires all these different kinds of people with different incentives. I used to say to some of my business school friends, “You have a shareholder value. You know up or down in a day.” When you don’t have that, I think that’s magic, because it is what I think is the beauty of public service, the beauty of trying to figure it out for a society that is incredibly different. That’s part of the muscle you learn and the excitement, I think, in government and tech coming together and trying to figure this out.

Marina, you talk about crisis engineering in your book, Crisis Engineering. It feels like, right now, we’re continually on a crisis of some kind. You make the case that this is actually a moment of opportunity for lasting change. I struggle with both, like we’re in crisis, we need to address what’s happening right now. How do we do it in a way that’s going to be lasting? By the way, how do we stop having crises? I want you to share a little bit more of your thinking around that aspect in your book, Crisis Engineering in general.

Marina: What prompted us writing the Crisis Engineering book was in the many federal organizations, and then after I left federal government, working in very large private organizations, too, everything is a complex organization, which simply means it’s made of people and computers. Computers are, relatively speaking, pretty straightforward to change. The human part is really, really, really hard to change.

We would occasionally observe that one. I especially felt this in my years at the VA, I would be like, “Hey, hey, hey, hey, there’s a backlog,” Chicken Little running around with the sky is falling, and everyone else would be operating business as usual. Then suddenly, we’d have healthcare.gov, and now it’s like, “Oh, let’s change every procurement, every HR, every approval, everything in two hours, guys. Let’s make all these changes so rapidly.” The humans of that system were not only willing to change, but made change, and then that change stuck. That became the new calcified bureaucracy going forward.

We were really interested in studying how you take advantage of that. It sounds probably very weird for me to say that a crisis is a great opportunity. Whether we like it or not, the human part of a complex system is going to change coming out of a crisis. How are we capitalizing on the next crisis so that you’re emerging from it potentially stronger? Amanda, to your last point, I don’t think you need to be blindsided by crises all the time, and I think a crisis can be very eye-opening for, how did we end up in this moment where the reality that we thought was real is clearly not?

Amanda: Do you see that same being true, David, around the world when you see these moments of political crisis happening? Is that an opening or is that a restart?

David: The one thing I want to add to what Marina says is, I think a crisis is a moment when the delta between what you thought was a reality and what actually is reality suddenly shifts. I think the bigger dynamics around that matter a lot as well. What I mean by that is, is it just your org that is surprised by that and just your team, or all of society? I think when it’s all of society that’s actually quite surprised by that, you actually can get a lot of sympathy, because everybody’s like, “We’re all surprised, and we all want to support you in helping that transition.”

If people outside the organization have been saying for a long time, “The reality that you think is real is not actually the reality that we perceive, and we have seen this problem for quite some time, and you who have opted not to do something because you were comfortable in that place,” then there can be a lot less public engagement or sympathy or support. Now, you’re in a political crisis as much as you are an engineering or service delivery crisis.

I think that dynamic matters a lot. I think one of the challenges we have in the public sector is many crises increasingly are not a surprise to everybody. It’s a surprise to the people who are inside the organization who are comfortable with the way things have been, but actually outside that organization, people have been living in a very different reality. They’ve been getting services differently. They’ve seen organizations organized differently. The public sympathy is much, much lower, which means you don’t get as much support or timeline to make the changes. I think that is what makes the public sector context so difficult.

Amanda: You’re defining what crisis and for whom. [chuckles] It’s interesting because I entered at Code for America at COVID, and it was an unusual moment in our country’s history where it did surprise everyone all at the same time. My first year of being in civic tech was seeing all the potential of people giving folks some grace to figure it out, some leeway to try something new.

I think back to those moments of, “What could or should have we done,” or, “Oh my God, I can’t believe we actually did that,” because now it does seem a little bit harder because the crisis or people are experiencing them in different ways. This leads me to where we are in AI because I do think people are experiencing AI in all kinds of different ways. In that sense, I’m curious, David, how you think about what you just said vis-a-vis how AI is now entering into the space of politics and government and policy, because AI has been in technology for a while. Now, it’s truly now here, both in the public and in the government sector.

David: The challenge we have with AI is there are no good and best practices in the private sector right now. There is nothing to port over. I think equally frustrating is there’s also really unclear demand from the public for the government to be using AI because they don’t also have models. Back again in 2014, 2015, 2016, people knew how to consume online because they had all sorts of services they were consuming online. There were patterns about how to do that that would have been established. Those patterns don’t really exist.

Marina: Such a clear framing, David. I hadn’t thought about it that way. When David said earlier about how there’s a crisis where everybody is surprised, and then there’s a crisis where one team is surprised, and everyone outside of them was like, “That was pretty obviously coming,” I think AI is going to catalyze that a thousand times.

David: The part that makes this even more challenging is this is also not like 1994 when the web first started to emerge because it’s like that moment, except for the web when it first emerged had 30 million to 50 million users globally. It was very small. Whereas the distribution mechanism for AI is 800 million users in three months. We don’t have a great precedent for thinking about where we’re in that. It’s a moment about how do we do something when we don’t have good and best practices from the private sector to lean on, but we have high adoption, but unclear how it maps to things.

One of my favorite problems in the UK is that for the last several years, people have been basically eating up all of the driver’s license test slots in the country. The only way for you to get a driver’s license test, if you live basically in the southern UK, is you have to pay a middle person to get a slot. No one’s been able to solve this problem. That was the same thing that happened with PlayStation 5s when they first came out. Everybody bought all the PlayStation 5s, and you could only get PlayStation 5s through the secondary market. AI is going to make it very easy for people to do that for every government service.

Amanda: One would argue government missed the social media time. Basically, didn’t regulate, didn’t really engage. Government services weren’t using some of the social platforms to talk about how you get services. It really just was the thing out there. Now governments are trying to figure it out. Senators are talking about it. You have working groups and caucuses thinking through this. Are you saying that, given how fast this is happening, people need to be very clear what role or what lens they need to play right now? How does it shape how people then, or how government should show up, how technologists, how nonprofits should show up in this fast-changing space at the moment?

David: I feel personally very challenged by this because in the 2014 model where we had good and best practices from the private sector to lean on, what you did is you said, “Okay, let’s go adopt those and figure out how to navigate them in.” I even wrote a piece saying, “I want a fast follower government.” I actually don’t want my government at the bleeding edge. I want them to figure out what has been adopted in other places and to figure out how to adapt them into the public sector.

Now, I think we’re in a very different, interesting place. Maybe government should be at the edge. Maybe government should be setting the patterns for what good behavior in an agentic world look like. What does a good service look like? What does responsible behavior look like? Imagine if governments had been at the cutting edge of social media and saying, “This is actually what a responsible post looks like. This is what responsible communication looks like in a social media age.”

Marina: Looping back to where we started around a conversation about efficiency and government having outcomes. I was the CTO of the VA for four years. At the end of four years, I had a website that looked good, and you could fill out maybe two forms when I left. It shouldn’t take four years of unbelievable, incredible effort and absolute top-down support. Maybe we could fundamentally get to what is a good– and this is going to be dependent on the particular benefit or service, but what does really good look like?

I remember very clearly doing user research with veterans over 10 years ago, and the best experience they wanted from the VA in terms of benefits delivery was no experience. They wanted to separate from the military and have their pension and their disability claim. They did not want to get 50 pieces of mail and get 5 phone calls and come in person. Nobody wanted to do that. Healthcare is different. They don’t want no healthcare experience, obviously, but for benefits. Maybe if AI is going to deluge our driver’s license application slots and deluge our benefits, could that open the door to fundamentally rethinking how we do benefits delivery at all? I hope so.

David: A little somewhat less optimistic than you, Marina, that AI is going to drive faster change inside the government for the very reason that we just talked about earlier, which is, is the crisis felt inside the organization? The change is only driven if the delta between what the world sees and what you see becomes large enough that it actually affects you. One cannot underestimate the inertia of people being very comfortable in the way that they’re behaving and the way they’re doing.

In fact, AI could have the exact opposite impact, and that now people will be able to create their own interfaces with government websites. It actually no longer matters whether that team moves quickly or behaves quickly. They will simply create their own interface to government. I think the tragedy of that is it will reinforce a notion that the state does not have the capacity to adapt or to deliver effectively. It will reinforce a notion of like, “We’re all better off on our own. We’re all better off as individuals. We cannot trust the state.” That will do enormous damage for when we really need the state to help us because people will not have confidence in it.

Marina: Oh, boy. I’m not usually the optimistic Pollyanna of conversations, but I guess my hope, and I want to phrase this carefully, is that government hiring cannot withstand a million AI-generated resumes. It will collapse, and that collapse is what will open the door to the opportunity for people to reinvent what that process looks like.

Amanda: Yes. I do wonder that it’s happening so fast, and does government’s role become clear because things just break, things just fall, and then you recognize there’s nothing there, so now you have to create it, and government is the only thing that can. What I will say is, in this context of efficiency, none of this with AI sounds efficient because so much of AI, it’s in these early days with the administration we’re in right now. The AI equal deficiency is sort of what the headline was. This conversation is surfacing that, actually, because there aren’t rules, it’ll slow it down, actually, and if you get it right, it’s going to be bumpy until you figure it out, which doesn’t feel efficient at all.

Marina: AI is also a huge category of things. It includes OCRing paperwork or effectively fuzzy word search, machine learning, things that are actually quite established. David, to your point, maybe those are areas actually where there are some best practices to be brought in. I don’t hear anybody really talking about those, and certainly not in service of an outcome.

To bring back to child welfare and outcomes, there are social workers who became social workers to serve families and children, and whose time we desperately need to spend with children and families, who spend 18 hours a week redacting a PDF file for court. Here in King County in Seattle, where I live, our court files are dot-dot files, not dot-doc, dot-dot, that’s how old they are. You spend three times as long filling out a court report because it requires extremely precise entering. Let’s throw AI that, for sure, but measure it in the context of how quickly can this caseworker do their red tape work. How satisfied are they in their job? What does their caseload look like?

In this foster parent licensing work, I can think of 17 states that had specified down to the quarter inch how far apart your window blind slats had to be if you were going to be a licensed foster parent. Now I’ve got to go change that at state legislation? Oof. Instead, the state legislation should say, “Follow the model standards, and then as window blind safety changes, and it does, we can change that way more flexibly.” If you’re in a competitive space, maybe there’s, to David’s point, opposition, there’s two different groups. Fine. Two sets of model standards, and you can debate it out whether you pick A or B.

David: There’s a hugely important thing I want to pick on up in here. One is the only thing that matters in a standard is adoption. The standard gauge won not because it was the best, because more people were using it. The invested capital in it was highest, and so it was easier for everybody to adopt it. That’s the first thing. The second thing is this piece around model standards is so important because I’m pro-open source. I don’t know if I’m an open source advocate. The reason is I think about web servers or your email client for people who are listening. Are you using an open source email client? Do you care?

What’s really interesting is because of the SMTP protocol, you can download Thunderbird or use Mail.app or use Outlook, and it takes you about five minutes to change because the switching cost is insanely low because we have that model standard for how the data is structured. Then you can have a competitive market on top of that for mail clients. I don’t care if they’re open source or proprietary. I just care that switching cost is low. If open source is a better model, then it will win out. If it’s not, the proprietary one will win out.

Amanda: In AI, if you try and create this, you have pretty big players that to the point of adoption who have to follow the rules. We keep hearing right now that AI companies want this, actually. They want government to do something here to make it come together. It’s not really happening. What’s the barrier that you see as to why it’s not? What we should be doing, we as civic tech, we as citizens, as people, we who understand the technology space, what should we be doing right now to drive to those guardrails?

Marina: I’m going to answer that I don’t know that I think of model standards for AI as much as model standards in service of things that have an AI component would be how I would frame it. As civic tech, we can make those. We can make those great examples, and we can make them available. There’s time-sensitive things right now, right, around how do we make Medicaid work requirements not disenroll a bunch of people that don’t deserve it?

What are the model standards for that? We’re not going to be in every state, and whether open source or not, we’re not going to be able to modify 52 systems ourselves, right? What model standards could we be saying, like, “Here’s the pattern to follow, and let’s get it in a state or two. Let’s test it.” I know this work is undergoing, which is why I’m using it as an example and sharing that out.

David: I don’t want us to confuse these two things. I’m a big believer that we should have guardrails, but I don’t know if that necessarily means model standards in the AI space. Right now, we’re so early in AI. A year ago, people talked about agents in a hypothetical sense, and now people talk about them in a very real sense. Even when we talk about them in a real sense, you have to remember the number of people on the planet who are actually running an agent is what, I don’t know, 0.5% of the planet’s population. This is not a widely adopted technology. We just happen to sit in nerd circles where people are doing this, but this is not normal. It’s very, very, very early.

I always like to remind people, the first railway was built in 1825 in the UK, and the UK legislated the standard gauge in 1846, 21 years later. The first railway in the United States, 1836, Pacific Railway Act is 1862, with the amendment in 1863 around the standard gauge, so about 30 years. AWS launched 2006 with S3. We’re now 20 years later. This is a point where I think the innovation for at least storage at cloud has had whatever you want to call beaten out of it. We know how to do this. The benefits of standardization significantly outweigh the benefits of any innovation that’s going to take place at this point. That’s why I think we’re at the sweet moment for cloud standardization. We’re at the same moment we were with electricity, same moment we were with railway, same moment we were with a number of other things that were built that way.

Amanda: It’s interesting to think about what you’re suggesting is whenever we have this major innovation, there’s a time period. In that time period, right now, people are scared. Part of what we’re seeing with our listeners is it’s not necessarily tech people. David, in between the it is here to standardization, there’s a whole heck of a lot of chaos. Right now, particular concern in the public about what are we doing now. What message do you have here given that history that you’ve seen?

David: I think one of the challenges that we have at the moment, again, this comes back to that 2014 comparison I was making, which is the distribution of the internet allows us to get 800 million to 1 billion to 2 billion people in three months. Whereas even 30 years ago, a government could see a change coming and begin to plan a regulatory regime based on what would happen in the United States. They were like, “We’re going to be five years out, and so we can figure out how to do this.”

We don’t have great models for how to regulate things that get adopted this quickly and this much at scale. Candidly, the tools that most governments have at their disposal are not very good. The biggest thing I would argue is if I was in charge of a non-China, non-US government, I would be investing very heavily in an AI safety institute like the Brits have done.

Amanda: That’s great. We always end with the same question. If you could wave a magic wand and fix how one thing works in government, what would you change?

Marina: Everything would have a constituent-facing outcome.

David: My favorite pet peeve is capital budgets versus operations budgets. A lot of stuff gets shoved in capital budgets that’s really actually operations. The fact that we can’t do multi-year cycles of funding means we can’t fund agile teams. We can’t do the product work that many people talk about. The money is broken. We sit around talking about the policy. We sit around talking about the operations and the teams, but the money is broken. The money runs the government.

Amanda: I haven’t heard that answer before. When I think about all the different barriers that we face, whether it’s in government or politics, that also has a lean into government, money shows up often. I appreciate that. If we are driving the outcomes, you end up seeing how those outcomes are affected by the influence of money, too.

I love this conversation. I wish we had more time. In fact, we need to make more time in the future because I learned so much from both of you, continue to, and especially as we’re in this space. I look forward to more conversations, both individually and perhaps even together like this because this is super fun. I really appreciate, Dave and Marina, you guys taking the time to be with us today.

David: Thank you.

[Music plays]

Amanda: That conversation with Marina and David made me recognize that maybe folks have been running toward the wrong thing or the wrong word, efficiency. An efficient government, as though you can reduce what government does to a balance sheet. Now, I know we usually like to simplify in these conversations for our listeners, but we are at a time when we need to remake government to really meet people’s needs today. That means an innovative, smart, caring, dynamic government, one that truly is ready for a future generation.

What I took away from the conversation with David and Marina is that it is actually possible, even when you go through a crisis, as Marina validated in her book, Crisis Engineering. We can take inspiration from what David sees around the world, other governments redesigning their systems to be more effective for all. Here’s the thing. Their views and ideas aren’t for someday out there. They are for now, from experts who are deep in the work with intimate knowledge of what it means to change a complex system into something more effective.

Sometimes it’s so easy to see when you stop and read the words, the Department of Government Efficiency, DOGE. One thing might’ve made all the difference. Instead of efficiency, it could’ve been effectiveness. In order to really unleash the talent we have in this country, we need people who can create effective processes, a better system for all, like Marina, whose expertise is rooted in practical research with the VA, and David, who brings a measured approach to the limits and opportunity of AI, because we are entering this new world of AI in government. As we do that, it’s time to create that system or that Department of Government Effectiveness. That’s all for today on The Government Fix. Thanks for listening.

[Music plays]

Amanda: This podcast is brought to you by Code for America, the country’s leading civic tech nonprofit for over 15 years. We believe that government can work for the people, by the people, in the new digital age. We work with government at all levels across the country to make the delivery of public service better with technology. We partner with community organizations and governments to build digital tools, change policy, and improve programs. Our goal, a resilient government that works well for everyone. Learn more at codeforamerica.org. 

Want to learn more about how government can actually work for everyone? Like or subscribe to The Government Fix wherever you get your podcasts, or visit us at thegovernmentfix.org.  

If you liked what you heard and want to support our mission, visit codeforamerica.org/donate. We’re committed to building a future where government services are simple, accessible, and easy to use, just like you’ve been hearing about this season.

[Theme music fades out]