Government benefits are broken. Is AI the fix?

0:00Hey, folks, Jeff Berman here, co-host of Masters of Scale. I am thrilled to share some of the new names who will be joining us at this year's Masters of Scale Summit. They are the leaders driving the most pressing conversations in AI. Replit founder and CEO Amjad Massad, Cloudflare's Matthew Prince, Signal president Meredith Whitaker, and many, many more who will take the stage this October 20th through 22nd in San Francisco. We want you there with us, too. Join us at mastersofscale.com slash pioneers. That's mastersofscale.com slash pioneers. On Masters of Scale, iconic leaders reveal how they've beaten the odds.

0:44Asking really strong questions is a superpower. You want to show up with something radically different and how they've grown companies to incredible heights. The greatest rewards always come from the greatest risks. That's hit the gas. Airbnb, Zillow, Microsoft, Liquid Death, and more. Hear from the founders who've changed the game. It's anything but business as usual. Find Masters of Scale on Apple Podcasts, Spotify, YouTube, or wherever else you get podcasts.

1:21I think the new rules of our society are going to be written by technologists. How the society operates will be written by code. It will not be written by people who pass laws. I think that AI can be transformative, and I'm seeing programs work quicker, programs work better for people that rely on them the most. We spend more in this country on health and social services than we do on defense. Medicare alone is$1.1 trillion. So we should want these programs to both work well and be managed well. AI is either going to happen to you or it is going to happen with you or it's going to happen for you.

2:01And our goal is to have it happen for you. When people stand in line for hours to get their driver's license, We hear about hundreds of thousands of people who were wrongfully denied government food assistance. They rarely think, I can fix this. But Phaedra Ellis-Lampkins thinks, I can fix this. I can make this better. She's the CEO of Promise, a software company whose mission, in her own words, is to move money in and out of government to the people who rely on it. Promise is already running programs in Mississippi, Pennsylvania, Florida, and more, making them faster and cheaper to administer without losing sight of the low-income families the system is supposed to serve.

2:47In my conversation with Phaedra, we get into how she's making all of that possible with AI. I'm Rana El-Khalyubi, and this is Pioneers of AI, a podcast taking you behind the scenes of the AI revolution.

3:09Hi, Phaedra. Welcome to Pioneers of AI. Thank you so much. I'm happy to be here. I'm so excited for our conversation. So I saw you presented Masters of Scale Summit last year, which was great. How was that experience like for you? It was great. I mean, it's hard not to enjoy when some of your favorite people are in a place talking about some of their biggest ideas. So it's just an amazing, amazing experience to be there. I would love for you to ground us in the work that Promise is solving. It does seem like navigating the government system is a real hot mess. Yes. And so can you describe for us America's system of public benefits and how does that work?

3:51And what is it like as a user or a citizen navigating that system? Yeah. So Promise focuses on essentially making government work for the people who rely on it. And that means it should work well. It should be inexpensive. We should be able to measure its impact because my fundamental belief, and I think our company's belief, is that a society is only as strong as its ability to govern itself and operate. Everyone has a nightmare story of going to the DMV and it taking hours and hours or having an appointment and not being seen. And most systems, I think that we have seen, we're clear that they are not operating in the most effective way possible.

4:31At the same time, it is not just the public sector. We see whole cottage industries of consulting firms getting paid$300 to$500 million to do a modernization that barely is off of cobalt, which is an aging system for people use for programming. My parents programmed in cobalt. Yeah, did they? Does it still exist? Yes, it still exists in many places. And so the fact that we need folks who understand COBOL or that information is still stored under someone's desk on a server, that is not the best that we should be able to offer our society in 2026. So my parents are over 70 and they got their green card a couple of years ago and we've been trying to get them health insurance.

5:16And it is an absolute nightmare. And it's so frustrating and it's creating so much tension within the family because it's like I'm responsible for getting this figured out, but I'm really struggling. Like where to go on the website, who to talk to. Yeah. I don't know if you have any thoughts for us. I have a lot of thoughts. And so many thoughts. One is a lot of why we exist is because of what's happening with your parents, which is, one, the law changes a lot. And so right now, for example, your parents would get probably Medicare or Medicaid depending on their income. And when you look at what is required, there are things like work requirements.

5:57So even if you're over 70, you have to have work requirements. They haven't worked in the U.S. ever. Right. So the way it works in the United States right now is they would be required to report volunteer hours or that they had tried to volunteer. And so part of what's hard about these programs is you have state laws, you have federal laws, and you have changing laws and changing policies and procedures. If your parents were in the state of Florida, I could help solve this for you in two seconds because they're our clients. And the reason is, is because we have a system that they use through AI that actually can manage policy change and that can audit and say, hey, this doesn't work.

6:34This is right. And then actually change the system in real time. And so part of what we're talking about is you might go to a website and the website, the law changed in January, but the website didn't update the new policy change. So you're still going to be allowed to apply, even though your parents aren't actually eligible, with three things that need to be met. And that is the fundamental problem. We have not updated systems to recognize that laws change, policy changes, and that the system should move as quickly as we pass those policies. And it should get easier to use, not harder to use.

7:06Can you give us a sense of how many people are using these systems and how much money is flowing through? Yeah, we spend more in this country on health and social services than we do on defense. Medicare alone is$1.1 trillion. So when you hear the defense department saying we're trying to become a trillion dollars, just think, oh, that's really just Medicare. So we spend a lot of money. So we should want these programs to both work well and be managed well. A lot of the funding happens at a federal level, but then goes to the states to implement. And so part of what I feel so excited about is we work with the state of Florida, as an example.

7:42We've integrated with the utilities and with other social programs. And so if you, for example, apply in California, you come into an office, you bring a copy of your bill, you might do a form online, you bring your income. And in 2026, we should not run programs like that. So what you would do in Florida is we're integrated with the utilities. So we know what the bill is. And you have the information. You have the information. And as we think about it, it's more effective for us to integrate. It's also more accurate and it decreases the risk for fraud. And that is what we think about. If the bill exists, your employer reported your quarterly wages.

8:20Why do we want a human to bring us one? It doesn't make sense. I think part of what's hard about health and social services is it's largely dominated not by software companies, but by consulting firms. What is the difference between a tech company doing this work versus a consulting company? Yeah, it's a good question. I think the difference is a tech company is measured based on outcomes and a consulting firm is measured based on hours and a number of people that work on it. So if I'm a consulting firm, I want a big contract. I get a big contract because I'm going to have a lot of hours and a lot of people working on it.

8:59If I'm a tech company, I want high margins, which means I want to get it done as quickly as possible, as efficiently as possible, because I'm measured based on outcome. And so as a society and government, we should want outcome measurements. We shouldn't want people to be paid by the hour. These aren't lawyers, right? This is a project manager. Why would we want to pay those people by hours? It's the wrong incentives. And so it doesn't align the incentives with getting the work done and being most efficient. Yeah, it doesn't make sense. OK, so I want to talk about what exactly Promise does. But before that, I want to kind of wind the clock back to how you got started.

9:33And so I started my company out of MIT and one of the applications of the technology was helping kids on the autism spectrum. So we spent a lot of time debating whether we should do a for-profit or a nonprofit. And we ultimately went for a for-profit organization. We felt like that would be more sustainable. So I'm curious about kind of when you were starting Promise, did you have to think about that? And also what led you to start Promise? Like a little bit maybe about your backstory. Sure. I had a really different life experience than most people who founded companies, which is I run a nonprofit.

10:06I ran a labor federation. I was an elected leader of a labor federation representing unions. I worked in music. I worked with the musician Prince and then understood the impact of technology. I was just like, wow, this technology wasn't good for workers. It wasn't good for the environment. It isn't good for musicians. So I wanted to understand it. Thought about going to business school, got offered a job working with an investor. And then I went to work at a company called Honor, which is now the largest home care agency in the country, a technology firm. And so for me, Starting Promise was really about how do you build a company that in the same way we've done for the defense sector, that you expect innovation to be centered in health and social services since it's the most amount of money we spend as a society.

10:50And it is the place that your parents, that our children will depend on kind of what we think the future looks like, especially in an AI world. Yeah. I have to ask you about your experience with Prince. I know you get this question a lot, but how did you end up with that kind of relationship and connection? And what did you learn the most? It was really interesting. I have a friend, Van Jones, who if you've ever watched CNN, is on CNN. Okay. And he introduced me to Prince. And I was pregnant, and he asked me to work on a project with him, and that project went well. And then Prince called me and said, can I be your client?

11:25And I was like, I don't have clients. Like, what do you mean, client? client. And I ended up working with him. And it was really incredible for a couple of reasons. One, I spent a lot of my time working on kind of justice issues. And what became very clear to me is that people liked doing fun things more than they like doing hard things. Like when you're having a rally, you're like, come do this. And what I discovered is when you invite people to a concert, they want to come. Oh, if it's fun, people want to do it. And I was like, oh. So I I learned something like, oh, we're asking people to do hard all the time.

12:01The other thing that was so interesting to me about Prince is that there is a boldness that I had experienced, but not at the level of his boldness. Like ambition? It isn't just ambition. Maybe I'll give you an example, which is he was frustrated one day. And so he kicked everyone out and he sound checked every instrument himself. Wow. And so you realize when someone has taken that time and that talent where they are good at everything so they can control and understand and create, it was a discipline that I had not seen before. And it made him bold because he could tell you what to do because he could probably do it better.

12:37And so I just was like, there is an incredible commitment to excellence. That's why when people tell me I have a daughter, she's like, I don't want to practice. I was like, Prince practiced every day. For sure you can practice. Okay. For sure. You can do it. You can do it. Don't go anywhere. We'll be right back after this short break.

13:14Hi, I'm Molly Graham, host of Work Life, a podcast from TED. The most important lessons about work usually aren't the ones anyone teaches you. They come from experience, from uncertainty, from figuring things out as you go. On this show, my expert guests and I explore how careers really unfold. Through change, through doubt, through the decisions that shape who we become over time. Because those moments aren't the exception. They are the work. Listen to Work Life wherever you get your podcasts.

13:51The business world is moving faster than ever. And when change hits, we need to learn in real time. On Rapid Response, you'll hear candid conversations with CEOs and leaders making tough calls about AI, human talent, responsibility, and the bottom line. How they navigate uncertainty, pressure, and high-stakes moments. I'm Bob Safian, former editor-in-chief of Fast Company, and I'll be your host as each episode breaks down what you need to know right now. You can find Rapid Response wherever you get your podcasts.

14:37I sold my company and I'm now an investor in early stage startups. And we see a fair amount of like what we would call gov tech companies, right? Like companies serving the government sector. And I often have two big questions slash concerns. One is how big is this opportunity? Yeah. And you're kind of addressing that this is actually pretty big. And two, it just sounds like a nightmare to even get into the system as a supplier. Especially like my sense is a lot of government, they don't want to do the wrong thing. Yeah. So curious about your experience there. I think the market's huge. It is a hard entry point, right?

15:18We've been doing this for a while and we just signed our first federal contract this year. And I think what's harder than other markets is you have forces consistently working against that have a lot of infrastructure. So an example might be, I have a big consulting firm. I probably have a third of my staff who came out of the administration that you're trying to work with as a startup. And so they can call someone, they can talk to someone. We already have a contracting vehicle. There's actually a firm that has a law that means you don't have to go through procurement if you work with them. I think the other part is, I think that the Defense Department has largely made a decision that it is okay to innovate, no one would expect the Defense Department to build its own plane.

15:55It wouldn't even occur to you. And I think it's largely because in defense, people have made the recognition that intelligence is valuable, that the way that you manage information is really critical. And I think on the health and social service side, it's a little harder because everyone is scared and remembers a story where someone hired a startup and there's a saying, no one gets fired for hiring Deloitte, right? Like, even if it doesn't work, it's like, that's Deloitte. It's not your responsibility. Government is not a system that rewards innovation in a lot of places. Can you walk us through kind of a couple of examples of what life was like before Promise existed and then you came along and what does it look like now?

16:36Maybe I'll give an example, the state of Mississippi, just based on thinking about your parents. And we work on work requirements there, which for folks who don't know is some states had some work requirements. There was a new law passed that has very specific work requirements. And in a lot of places, in the state of Arizona, as an example, there's been almost like a half cut in food stamps, the number of people on food stamps, a third of which are children. And so we know less children will have access to food stamps. In the state of Mississippi, they want to support what the Trump administration is doing, but they want to make sure in line with that that they are not wasting money, that there is not fraud, and the system is working well.

17:15And so we actually executed a program there that makes – think of it as like an automated work requirements reporting. And so what we'll do is we will reach out to you to first make sure you're aware of work requirements. What are they? What does it require? And we'll do that by text. Then we're going to allow you to do that reporting through us by text. And then what we're going to do is have a subscription service where, for example, if you have wages, we can pull them once we get permission so that you don't have to keep providing the information, but you give us the ability to access it. We ask you ahead of time, and then we're able to pull that information.

17:53And so in Mississippi, Mississippi had almost a 2 % decrease in their fraud rate or their SNAP error rate. And why that matters a lot is because basically what the administration said is we're not giving you states money to pay for your food stamp administration if you don't decrease the fraud or error rate. And so what we saw is Mississippi did great when we looked at other states. They brought their error rate down. They're killing it. And our basic premise, we were talking to someone this week, is if you fix the problem before it enters the system, that's the goal, right? That should be the goal.

18:30You want to train the person, give them access to information, and stop it from being an after-fraud problem. Can you clarify a little bit, definitely for me, but also for our listeners who are not familiar with how these systems work, what is the relationship between work requirements and health insurance or work requirements and SNAP benefits? Like, why are these things linked? It's a great question. So for most of these programs, you have an income requirement, right? Think of it as someone who's making less than$13 ,000 a year, right? So you're looking at an income requirement. Health care are the same things.

19:04Food stamps, as an example, there's a work requirement, which means that in addition to having that income less than that, you have to be able to prove that you are trying to find work, that you have work, or that you're volunteering. There are rules to be able to receive that benefit. So these are all like the qualifying criteria in a way. And they continue. So it is not just that application. You have to continue. So the work we're doing on work requirements is because you have to keep doing it to keep the benefit. Yeah. So you basically you apply and there's all these criteria, including work requirements.

19:37And then what is really magical about what Promise does is that you're able to almost like not auto populate, but connect to the systems that has the information. So you're able to continue to pull this information. Yes. It's an easy way to think about it is first it's education, because, for example, your kid might go to college and they've turned 18. so they no longer should be under your food stamps. And it might not occur to you like, it's June, my 18-year-old is graduating. You don't think immediately I have to go announce to someone that now the number of people in my household has shifted.

20:09So the first thing you do is you want to keep reminding people, these are the rules. Did anything change? And then they can do a change with you. Okay, oh, yep, it changed. So that's the first thing. Okay, we're going to tell them. The second thing is we're going to say, can we pull your income so that we can ask you, has anything shifted? Okay, great. can you take a photo? Is there something we need to do? It makes it easier for the person who relies on those benefits, but it increases your source material because now I'm pulling it directly from the employer. We're always trying to think about how to get the highest quality data because that's ultimately how you reduce the most fraud.

20:42And so the other systems are designed like once you find fraud, let's audit it, let's do something about it, where promise is like, how do you make sure people have information? How do you get higher source material? And how do you stop it before the payment goes out. That is our model. So let's talk about the role of technology and AI specifically in Promise. When did you decide to incorporate AI into the product? And actually, what kind of AI are you using? So I think for us, it's important to probably make a distinction where is their AI, where is their machine learning, just to like nerd out for a second.

21:16Love it. Love nerding out. A little nerding out. And I would say we're more machine learning than Anything else? And the difference might be for the folks who aren't nerds is that what you want to do is you really want to train something to be as smart as possible about the specific things it's going to encounter. And then basically training agents. We have so much access to data. What we're trying to do is train it to be able to understand and recognize patterns and know what to do when those patterns exist. Because for us, artificial intelligence is really about outcomes. The analogy we try to think about internally is Waymo, right?

21:54First, you started with humans driving. Then it drove with the humans in the car. Then it like you need those kind of quality controls. But for us, we feel confident in our agents making decisions. But we launch with humans running alongside agents. At any time, there might be 70 agents working on one case, one agent to do text messaging, one agent's looking at blurry photos, another agent is responding because you didn't send something, but we wouldn't expect there to be one kind of generic agent that's able to do all of those. Right. Because I would imagine every agent slash machine learning algorithm is trained on a very specific, for a specific task and on its relevant data.

22:38Exactly. Exactly. So maybe you're applying for something that requires payroll records, but it can only be these specific dates. So we have an agent that only looks to make sure if that wage information is in with a very specific range, because if you're doing it without promise, you submit it, you wait for a person to send you a letter or call you, whereas we can do it in 22 minutes because the agent says wrong dates. And then another agent sends you a text message that says reply here. So it's like the layers of agents, I think is much more likely to continue to succeed. We can make a decision in 22 minutes.

23:14Now, I'm going to play a devil's advocate for a second. I can imagine somebody listening to this and they're thinking, oh my God, now AI is going to make a decision on whether I'm going to get the SNAP benefit or this social service. What would you respond to that? Well, I wish that were true because I think you would probably get a better decision. Let's talk about bias in AI versus human bias, I guess? Totally. Let's talk about it. We see human bias based on their own experiences. We see human bias based on deserving or not deserving. When you look at the impact of like high caseloads, I guess what I would say to the person who would be concerned, and which most of America is now apprehensive of AI, is I would say it is coming.

23:58It is like protesting the automation of cars. It is like, it is coming. So the real question is, does it happen to you or do you make it work for you? And does it make a better government or does it destroy the humans? And so it's coming. And the reason I think people should root for companies like Promise is because we don't launch without QA and we've pulled agents back. So I think people are right to be concerned, but the world is competing for the future of artificial intelligence and we should want it to work well. We should have rigor around it. I appreciate you saying this because I think it's so important to have a high bar of what can get shipped, right?

24:42And kind of doing quality assurance around data and algorithmic bias is really important. But I will also say, to your point, I meet a lot of founders who are not thinking about this at all. So I appreciate that you guys are kind of taking that seriously. We really are. And part of it is, maybe I'll give you a couple of examples, language, right? When people use Google Translate, I'm always like, oh. And especially for things like social services, because what I worry about is someone does fraud because there's been bad translation and it asks for something different. Right. And so you you need to actually translate something.