![Podcast microphone against a blank background](/sites/g/files/omnuum10826/files/styles/hwp_21_9__1920x825/public/datasmart/files/podcast_microphone.jpg?itok=3Q9qYvou) # The 100th Episode Mailbag! Episode 100 September 02, 2026 [ Betsy Gardner ](/people/betsy-gardner) It's our 100th episode! And we're flipping the script: Stephen Goldsmith is the guest, not the host. And he shares stories that completely reframe how cities should think about technology — from a 1970s bot that shattered assumptions about poor people using tech, to why AI alone should never drive government decisions, to the importance of bench marking the status quo. **In this episode, you'll learn:** - How to separate the AI hype from actually helpful AI uses (hint: it starts with a question). - Why Goldsmith rejected patronizing attitudes around technology adoption, and proved detractors wrong. - The distinction between "residents" and "customers," and how that changes everything about how you budget and design. - How to give public servants back time and respect, rather than saddling them with meaningless tasks and paper pushing. - Why your old, top-down performance management system is dead and why you need a new, democratized model. Embed *Listen here, or wherever you get your podcasts. The following is a transcript of the conversation.* **Betsy Gardner:** Welcome to the Data-Smart City Pod. I'm Betsy Gardner, the show's producer. And today we're doing something a little bit different. Stephen Goldsmith is the guest this episode rather than the host, because we're doing a mailbag episode. Steve, how are you feeling about this? **Stephen Goldsmith:** Not good. I like controlling the questions and not being responsible for the answers. **Betsy Gardner:** Well, we had too much interest, so you had to get into the guest spot. We do have a lot of questions, kind of ranging from your time at City Hall in Indianapolis to your work in New York to some of our more recent research. But let's start with just hearing a bit about your background. Where and why did you start in public service, and then how did that bring you to where you are today? **Stephen Goldsmith:** Well, that covers a half century. I don't know that we can do that in three minutes, but...well, in short, I was always interested in being a mayor. I went to law school. I couldn't go to mayors school, there was no mayor's school, so I went to law school. I eventually managed to get elected as District Attorney in the county that included the city I grew up in, Indianapolis. I served actually three terms as District Attorney, probably one term too many. And then I ran for mayor. I was influenced early in my career by a Rhodes Scholar, Phi Beta Kappa, who became the mayor of Indianapolis and occasionally communicated with me...a guy named Dick Lugar. And I wanted to be like him. And I thought being mayor was the hardest job in America, so therefore, it would be a good job to have. **Betsy Gardner:** What's kind of the thread that runs through all of your work? **Stephen Goldsmith:** Well, maybe the easiest way to say this is there is always a better way to produce a high quality service. There's always a way to rethink it, to reimagine it, to change it, to listen to people. So the thread is - whether you're prosecuting cases and representing victims or you're a mayor looking at place-based restoration or helping people own their own homes or the work that I did for Mike Bloomberg in New York City, trying to help a very large bureaucracy and improve its responsiveness - there's always a way, always a way. **Betsy Gardner:** You recently talked to us about the way that data has been running through your work and the way that you've sort of been advocating for including data and technology in government, kind of from the beginning. Could you tell us a little bit about, I know in particular when you were in Indianapolis, there was some pushback against incorporating technology, and you chose to go ahead with it and showed that there was actually a really easy, widespread adoption of the technology that improved the experience of the people and the experience of the employees. **Stephen Goldsmith:** There's actually been four moments like that in my long history, career, one in child support collections and the use of technology. One when Indianapolis was one of the first cities in the country to do e-Gov online transactions, maybe the first. Another was trying to advocate for a data analytics center while working for Mike Bloomberg. And the fourth is the role of AI and its potential for disruption in a positive way. I don't want to spend the whole time talking about Steve Goldsmith stories, because they're only interesting to me. But I have to tell one story. So, I was a prosecutor, and in the Midwest, prosecutors collect child support. So they help collect child support for custodial moms - custodial parents, almost always moms. And those parents are disproportionately parents on some social service benefit AFDC, TANF. So, I worked with trying to help these moms, and we injected technology and improved our collections – there's much more to the story that's really kind of fun, but we improved our collections from $900,000 a year to $38 million without hiring anybody. And received an avalanche of inquiries from moms like, when's my court date? Is the check in the mail? Has the guy paid his? Well, and so what happened is because we went from $900,000 to $38 million and didn't hire any people, we didn't have anybody to answer the phones. So, the phone queues were really long. So, some guy came to us and said - now remember, this is like 1978 or something - and said, how about we have a, we have a bot we can put on your call center. Like, a bot? Like, in the late 70s? And everybody said, A, it's not going to work. And B, if it does work, these parents are poor. They're not going to know how to use this technology. So, this is my first awareness of kind of patronizing attitude that bureaucracy has, that people would rather have bad services - and you're going to force them to put up with bad services - rather than give an opportunity to use new technology. Day one: 6,000 phone calls. And the bot said, put in your Social Security, your case number and we'll answer these questions for you. The ones I mentioned. All right. So that taught me that technology can provide solutions. That the restriction is the way we think about our customers/residents/folks we're going to help. And thinking about their capacity and what they need more creatively. Injecting innovation through technology can be transformative. **Betsy Gardner:** I really love that story. It sounds kind of like one of the throughlines then, is that back and forth between cities and residents? And there's honestly, from your story, there's kind of trust on each side. And maybe that trust isn't always extended equally. But something that we have been talking about is what we're calling ‘responsiveness.’ And one of the first questions that we got for this episode was what does responsiveness mean? So, could you define responsiveness for the audience? **Stephen Goldsmith:** Yes. Perhaps. First of all, you know, many of us, many people including occasionally I, use the word customer. But of course, they're not really customers. They're residents or citizens, they're not customers. And there is a distinction there because you can't spend an unlimited amount of money resolving one person's problem because that comes at the cost of someone else. So, there are public responsibilities. We've been thinking about responsiveness as how the government listens. Listens to community groups, listens to the needs of the residents. Fulfills their expectations about the quality-of-life issues - how they care about their parks, what they're saying about their streets, their sidewalks, their trash, their job opportunities. So, responsiveness to us means listening better, synthesizing what you hear, acting quicker – often in advance of the problem – like preempting the problem and then reporting back to people what you've done for them. That responsiveness loop will create trust. Trust creates additional levels of engagement. Additional levels of engagement create a better civic infrastructure for your community. **Betsy Gardner:** So, another question that we've got is about the same topic because we just put out a [new paper called Next Level Responsiveness](/next-level-responsiveness "Next Level Responsiveness "). And it's talking about how [the Responsive City Cycle that we wrote about a few years ago](/responsive-city-cycle "The Responsive City Cycle") has now changed. And we're saying has supercharged the cycle. And so we've got a question about what are the benefits of AI in the cycle. Is it just speed that AI would make the cycle go faster? Or is there another benefit, or more additional benefits, that we get from plugging AI into it? **Stephen Goldsmith:** Well, there are many benefits. We just wrote an article on Data-Smart that you helped me publish that discusses how Cleveland and Dallas are using cameras mounted on trash trucks and other city vehicles to take pictures of neighborhood conditions for purposes of nudging people to fix code violations, graffiti, other violations of city code that degrade the quality of life in that neighborhood. So, let's think about AI. AI can read those images. You don't have to have somebody watch thousands of hours of video. It can compare those images to what the code says about, are you allowed to have a commercial facility here? Are you allowed to have an abandoned car there? Are you allowed to have graffiti here? Can turn those into either violation notices or better yet, reminder notices like please remove your car, or whatever the case may be. So, in this case we've got AI facilitating the collection of the information. We don't have to have people driving that street every day - separately from the trash trucks that are already there - you don't have to have people listening to it. It turns it into automatic recognition. So those quality-of-life issues. Jump to the other side. You know, I was deputy mayor of New York City. Where we had about 20, 25 million calls a year to the 911 center. 60% of those calls were informational at the time. Do I have to move my car to the other side of the street today? Right. And what that meant is all of the capacity of the individuals answering those calls was consumed - not all of it, but much of it - was consumed by answering those questions. So, if you answer those questions mechanically, if it can be answered mechanically, that frees up all that extra time on the part of those city officials to solve the difficult problems and help people. That's responsive. I'm going to use my discretion, my concern, the reason I became a public employee was to help people. It wasn't just to juggle pieces of paper and make mindless phone calls. It was to help people. So, we're going to redirect that information. Many, many more stories about how AI has the potential to so dramatically change this equation. **Betsy Gardner:** Yeah, it's interesting. It kind of makes me think that there is the speed that is improved, but then the saving of time is not just a speed question, but it almost sounds like you're describing it provides more...meaningfulness or, like, dignity to some of the work that we're asking public servants to do. **Stephen Goldsmith:** Right. Well, you could ask - your question is interesting. You could have... you could think about time on in two different dimensions, if you will. One is if somebody wants to get a permit to remodel their kitchen, they should be able to get it in two hours, not two months. There's no reason why they should have to fight through 5 or 10 different agencies. It ought to be a simple front end that's customer facing, that's enabled by AI, that speeds it up. That's one way to think about time, right? The opportunity cost of making people wait to improve their homes or whatever the case may be, get their permit to do their plumbing or get their certification to be a plumber. Then there's another aspect of time, which is, you know, every city can take the number of employees it has and times the number of hours or minutes they work per week and figure out how many of those minutes are productive, in being responsive to a resident, and how many are just kind of mindless commodity action. So, another idea of time is let's redirect as much time as possible to value-added activities on the part of conscientious public employees. **Betsy Gardner:** So, we have another question here that I think kind of goes a little bit more broad from that point, which is: what is the main thing a city should consider when it comes to incorporating new technology? What do you think about that? **Stephen Goldsmith:** Well, I don't think in the abstract it should consider incorporating new technology. That's backwards. It should say what are the problems we have and what are the technologies that will help us solve those problems. And then evaluate the technology in the context of the problems that it can solve. I think too often many people, including myself, get enamored by the technology, which is really easy with AI technology, and forget that that's not the purpose of the exercise. The purpose of the exercise is to help people. We also recently have worked on this [paper on interoperability](/asphalt-algo-interoperability "From Asphalt to Algorithms"), right. Which is, kind of, what does interoperability have to do with your question? Well, if you're going to evaluate technology, you shouldn't evaluate its ability just to solve a particular problem. You should look at how it plugs in and becomes interoperable with the system, right? It's not just how you handle the parking meters. It's not just how you write tickets for extended parking. It's how you manage the mobility and the curb and sidewalk. So, the question for purchasing technology is; what problem does it solve, what component does it do in a system, and will it play nicely with other vendors who have complementary technologies as well? **Betsy Gardner:** That's a good point, and we can share that paper in here as well. A question that has come up a lot from practitioners, actually, is that a lot of residents are either kind of skeptical or maybe a little burnt out because they have filled out surveys before and nothing's really happened, or they've given input, but they don't feel like that input has really been addressed. So, what would you say to a practitioner, someone who's working in city government, when they want to build trust with these resident interactions, and they don't want to just gather data that that sits in a dashboard? **Stephen Goldsmith:** You know, we had this article - what was it, Betsy, a few months ago? In Vital City about, you know, how too often the [rules of government make the "bureaucrat," in quotation marks, leave their discretion and judgment at home](https://www.vitalcitynyc.org/nyc-government-ai-accountability-discretion/). Like they have to go out and fill out these forms or write citations for X when they know it's unfair and are not necessary. Let's think about AI. So, a field employee goes out to a scene, that field employee on their iPad knows everything about that, say, restaurant if they're a health inspector. How many citations, what type of food, how many customers it has, when's the last time it was inspected, were there sightings of rodents in there? I mean, you can come up with plenty of things that are immediately available, right? And they walk into that restaurant, do they see a minor issue, do they see a major issue? Do they write a warning ticket, do they write a real ticket? Do they coach the restaurant owner on, "I know this is a well-intentioned thing you do, but you're violating the rules for the following reason." We need to give our frontline workers the discretion to help the folks solve their problems - except when they're not willing to solve them, then they get a ticket. The inspectors need to document what they do, so their supervisors, instead of looking through 300 files, can use the same AI techniques to evaluate how one inspector user discretion compared to another, how one inspector used her discretion for folks of different ethnicities and different neighborhoods or different foods, right? So, our ability to evaluate fairness and equity creates accountability, but our ability to give folks at the frontline discretion creates responsiveness. **Betsy Gardner:** We have another question: based on your experience with Stat programs, why did they need to change? What did the old model get wrong or what did it miss? **Stephen Goldsmith:** When I worked for Mike Bloomberg, inside my portfolio was the Mayor's Management Performance Scorecards. These were very comprehensive. Each agency had dozens of things that it scored. Our team sent out evaluators to look at the streets to determine whether the scores seemed to be accurate or not. It was very comprehensive. There were a couple limitations, though, that now no longer need to be necessary. One is those agencies, those departments, did not have the analytic capabilities to determine why some scores are high and some were low, and why were they higher in one neighborhood than another neighborhood. And what were the outliers? And they just, they had no capacity to do that data. And for the few agencies that did, it was just like one person attached to the commissioner. So, it was a very top-down process. The, you know, the mayor and the deputy mayor, agency directors trying to find out what was going right or wrong. Today, we suggested in [StatGPT](/statgpt "StatGPT"), we have the potential for the democratization of that process. There's no reason why you need a coder to find out why 10% of your locations are producing 90% of your potholes. Why are they doing it over and over again? What questions can you ask about the underlying conditions? What kinds of questions can you ask about the maintenance schedules or the truck schedules? Or the water drainage schedules or fill in the blank. That now is possible with natural language inquiries by the supervisor of the work crew. So, what's happened is the ability to provide tools to a broad array of supervisors to let them be more responsive, again, preemptive, and proactive in identifying neighborhood and resolving them. **Betsy Gardner:** In practice, would that look totally different from the old meetings of having everyone in one room and going through those documents? Like, do you imagine this is something that would be happening much more dispersed? **Stephen Goldsmith:** Yes. I mean, you go back to the original origins of stat. It was from Bill Bratton when he was the enormously successful chief in New York City. And the model was the commanders would be in front of him. He would interrogate them on, well, your robberies are higher than do your robberies and that sort of thing, right. And he would be backed up by his data geeks and green-eye-shade guys in the back room. That was a very top-down system. It worked because he was so strong, and was top down, but it had a couple of problems. One, the turnaround times were longer because commanders had to go back and kind of examine the data themselves and look at the situation and talk to people. There was no speed and iteration. They could have - you could imagine, a really smart commander and a really smart chief.... police commissioner they're called in New York City - having this exchange of information, questions back and forth, where the commissioner, the police commissioner himself, is probing in a way that unlocks the imagination of a person talking with him. So, the speed of iteration. But even more importantly than that, you want that precinct leader, precinct captain to go back, bring his team together, his lieutenants, and do the same thing and tell the lieutenants: here are the tools you have. It's a natural language tool, it's called AI. You have layered data. We have...we have layers and layers and layers of data. Ask questions of the data to see if you can figure out what's going on, what the problem is, and what the outliers are. So, the differences are speed of the iteration and distribution of the capacity to solve the problems. **Betsy Gardner:** We have another question that I think could be related to some of this. And it's about how can you tell the overblown AI hype from the genuinely helpful tools and agents? And I could imagine introducing a new model of performance management that integrates AI could maybe pull up some of those questions. **Stephen Goldsmith:** I think we go back to an earlier question you asked me, which is how do you decide what to buy? I think it's the same sort of answer. What's the problem you're trying to solve? And how has that problem been solved more quickly or better with whatever tool you incorporate, right? How do you identify graffiti more quickly? How do you look at the pothole before it occurs, because you can examine the underground conditions. How do you use, you've done a lot of work with ZenCity, on social media. How do you identify problems in communities where folks don't call 311 because they...have language barriers, they don't know the system, they're scared to interact with City Hall. But they have opinions that they express, anonymized on social media. Or through polling. Just asking them questions. So how do you listen better? How do you use IoT devices as sensors better, to understand air quality in those communities. And then what do you do about it? So, I think what we have to do is understand the system problem and the ways to approach it. **Betsy Gardner:** I also thought it was interesting when you [talked with Brita Andercheck from Dallas about how they give 12 months for an AI tool](/scale-innovation-data-change "How Cities Scale Data Innovation Through Cultural Change"), and if it isn't giving a return, then they shut it off. Which is an interesting way to sort of separate the hype from the helpfulness. **Stephen Goldsmith:** That's a really good point. You know, Brita in Dallas and Elizabeth in Cleveland…there's a handful of other cities that are really very competent in evaluating technology and using it. There's a bunch of challenges here. I mean, one is that the life cycle of software and hardware is much shorter. And that creates its own set of problems, particularly in the awkward way that cities buy things, right, the procurement cycle’s longer than the life cycle of the product they're trying to buy so that, you know, that's confusing. The other is that, many of these bigger problems are system problems. They're not agency problems. They're not transaction problems, they're systems problems. Which means you need to require that a technology is interoperable with the rest of the system, not just so that you have data access, but so you can get rid of one of those companies without messing up the rest of the system. So, I think the ability to fail fast that people talk about a lot, I think is really critical in this area. **Betsy Gardner:** Is there anything that you are worried about as cities adopt AI? This person is wondering sort of, if this went poorly, like what would you worry about the most? What implications would that have for future adoption of technology or innovations, like, if we mess up here, what is going to happen down the road? **Stephen Goldsmith:** Well, we will make mistakes. Some of them will be front page mistakes and some will be page 50 mistakes. There's a whole set of things to worry about that I don't specialize in, which is privacy, security, algorithmic bias. I mean, these are things that are really serious issues. They're just issues that others are more knowledge about. I spend my time on applying the benefits of AI, but raising my hand each time I do that, saying, don't forget this. Folks are talking about privacy and cyber and those things. So, that's one way to think about it. Another is, I just had this conversation, referencing back to when we were in New York City Fire Department trying to reduce the number of ambulance and EMS calls in order to speed response to those who needed it most on the theory that there's a number of folks who need a defibrillator within 3 minutes or 5 minutes, I can't remember what is was, right. So, when you do that and you use the data to do that, you're going to help a lot of people and you're going to hurt somebody. You're going to make a mistake. Perfection is impossible. So, I think that measuring the future versus clearly benchmarking the status quo, the problem that many make is they do in good faith something new without really publicly benchmarking the status quo first. Well, yeah, we did make a mistake, and it was a pretty serious make mistake, and we apologize. But because we've changed the system, 400 people are alive today that would not have been alive otherwise. Right. And so, it's the ability to categorize your improvement that's most important. **Betsy Gardner:** That's a really good point that because we're used to the status quo, it feels like we're not necessarily going forward if there is an issue, it feels like, oh, we went so far back. Okay, we have a couple of closing questions. And one question is, if you were mayor today, what would you do differently? So, what would Mayor Goldsmith do in 2026 that either you know, you couldn't do or you didn't do in 1996? **Stephen Goldsmith:** The tools that are available today are so dramatically better. I can't think of anything better than being a mayor of a big city today. The greater the problems the greater the opportunities to solve those problems. You know, with GIS, we can spatially layer the data. We can have natural language inquiry of the conditions in the city or the causes. We can look at fairness and equity. We can look at layered data and just ask questions of a city. The ability to do that to change the system, to make it more responsive, again, are so dramatic, A. B, our definition of efficiency back in the old days, right, was how much better we did something. Well, that seems like it made sense, but it's not. The definition should be how much better is it for our users, our citizens? How much more quickly can they accomplish their goals? How much easier is it for them to fill out the information? So, today's technology allows us to start from the human-centered design, right? Start from the person who needs it and go, go the other direction. So, I would just think totally different about the work I did, from trying to make things efficient from the agency standpoint, to trying to make them better designed for the person who needs the services. **Betsy Gardner:** That's interesting too, because we've been talking about MCPs and working with cities on building and hosting MCPs, and that being that big front door where a resident has a question and the MCP can connect all the different departments that traditionally would be siloed. So, I could see you having a Mayor Goldsmith MCP. **Stephen Goldsmith:** Right. I mean, look, if I were mayor, what I would have would be, I don't know how you would name the skill, like Mayor Skill Guy that I talked to every day, like, okay, tell me a little bit about kind of, here are the issues I had today. Where can I go find out which questions should I be asking for these people? Okay, well, I'm about to host a meeting here are the people at the table. But would you help me pretend that you're sitting at the table? And who would you be representing who's not at that table? What questions should I ask that would represent them? I mean, the ability to be a thousand times better is enormous. I mean, just interacting with community groups. I mean, who's using AI to listen to every transcript, every community meeting, every city council meeting, every committee meeting of the city council, every agency board meeting. Just tell this little bot guy to go listen to them, create summaries of the transcripts, and distribute them to everybody, including community folks who should know what happened in that meeting. Sort them geographically. I mean, the ability to change how we think about service, it's just like over the top, which is just amazing. **Betsy Gardner:** I don't think you set this up for me, but, you know, listeners, pay attention because we'll have something about that idea coming out...hopefully soon. What message do you have for people who are in public service today? **Stephen Goldsmith:** Every minute that a person is in public service today, they have an opportunity to make the life of someone else better. That's an opportunity that nobody else has. And so, they should think every moment, how am I spending this time? And does it produce a better outcome for the people who live in my community? That is wonderfully motivating, and I would recommend that, technology aside, focusing on that produces the best outcome. **Betsy Gardner:** All right. Well, thank you so much for being the guest today on the podcast. **Stephen Goldsmith:** Thanks for your questions. I'll be eager to see how you edit them. **Betsy Gardner:** Thanks for having me on, I guess. **Stephen Goldsmith:** All right, see you later. **Betsy Gardner:** Bye. ## About the Author ### Betsy Gardner ![Headshot of Betsy Gardner](/sites/g/files/omnuum10826/files/styles/hwp_1_1__100x100_scale/public/2025-05/Betsy%20Headshot%20resize.jpg?itok=k2OsSp1g) Betsy Gardner is the editor of Data-Smart City Solutions and the producer of the Data-Smart City Pod. Prior to this, Betsy worked in a variety of roles in higher education, focusing on deconstructing racial and gender inequality through research, writing, and facilitation. She also researched government spending and transparency at the Lincoln Institute of Land Policy. Betsy holds a master’s degree in Urban and Regional Policy from Northeastern University, a bachelor’s degree in Art History from Boston University, and a graduate certificate in Digital Storytelling from the Harvard Extension School. Share on:- [ Facebook ](#) - [ Twitter ](#) - [ Linkedin ](#)