![Boston City Hall lit up at night](/sites/g/files/omnuum10826/files/styles/hwp_21_9__1920x825/public/datasmart/files/leon-bredella-bvjtr3yqune-unsplash.jpg?itok=jyWHPwq2) # Redefining City Governance with Generative AI Episode Sixty-Seven December 04, 2024 [ Betsy Gardner ](/betsy-gardner) In this episode host Stephen Goldsmith and CIO Santi Garces discuss the potential to revolutionize urban governance with generative AI. Garces, the chief information officer for the city of Boston, joins Pr. Goldsmith to talk about the transformative power of GenAI in urban governance. In the first episode of this recurring conversation, they share how GenAI is already revolutionizing the way cities collect and use data, interact with residents, and empower city employees. Embed *Listen here, or wherever you get your podcasts. The following is a transcript of their conversation.* **Stephen Goldsmith:** Welcome back to the Data-Smart City Pod. This is your host, Stephen Goldsmith. I'm a professor at the Harvard Kennedy School, and Director of Data-Smart City Solutions at the Bloomberg Center for Cities at Harvard. Today, I'm introducing a friend and a person you've heard much about, and you'll be hearing more about in the future, Santi Garces. Santi is currently the CIO for the city of Boston and a senior fellow at our Center at Harvard. And in addition to all these formal introductions, Santi, let's have you tell our listeners a little bit more about your career…where you started as a student at Notre Dame, found your way into then Mayor Pete's office in South Bend, ended up in Pittsburgh as CIO and now you're in Boston in a position of major responsibilities and in one of the best data managed cities in the country. So, tell us a little bit more about your route. How did you get from Notre Dame to be a Big Deal CIO in Boston? **Santi Garces:** Yeah. Thank you so much again for having me. Again, for me, there's always been a passion for trying to solve interesting problems, and I think that when I came to the United States to study, I grew up in Bogota, Colombia, I was keen in being able to be closer to the action, being close to where some of these great discoveries and great inventions were being made. And after studying electrical engineering and political science, then doing my master's at Notre Dame, I came across an interesting mayor trying to do interesting things. And I had the pleasure of working with Mayor Pete Buttigieg at the time who I think that had been elected to office when he was 28, and tried to figure out how is it that we use technical skills to try to make people's lives a little bit better? How so that we could build better trust in government? And I figured that if I did it in the US for a few years, I would be able to go back to Colombia. And instead, as you mentioned, I just got deeper and deeper going to different cities and being able to work on interesting problems with other interesting mayors. And I have the pleasure of working with Mayor Michelle Wu at the city of Boston. **Stephen Goldsmith:** Over the next several months, we're going to do many podcasts together on AI and generative AI. But before I get there, let's just think a little bit about the following. Boston City Hall is a pretty sophisticated place, right? You've got a highly talented mayor. You've got your predecessor, Jasha is there in another role. There's a number of Harvard fellows hanging around trying to help you think great thoughts. What distinguishes a data literate city, meaning city hall, not the city generally, from one that's a little bit behind? What would be the characteristics of advancement from the data standpoint you would look to? **Santi Garces:** So I think, again, over the past 13 years that I've been in government, I think that there's a little bit of an evolution in how we think about data and our maturity. I think about 10 years ago, there's this sense that if we only had all the data, we would be able to solve all of the problems. And I think that up to some point in an age of COVID, George Floyd's murder and all sorts of social turmoil happening, I think that I myself have felt a little humbled around what's the ability of data and being able to solve these complex issues. And I think that where we are now is thinking data. First of all, data can look in different ways. It's not just the things that sit in our databases, on our tables, it's the collection of experiences that people have is how people feel about the city. It's a combination of things that are quantitative and qualitative. And in some sense, what we can do is not only think about trying to optimize government to run in this very efficient way, which it can be more efficient, but rather being able to use this way of collecting information to enable us to have common dialogue to try to solve problems in a way that is iterative. And knowing that often in government, part of the challenge is that the objectives are at times at odds with each other. But what we can do when we have data is that it allows us to be able to look at the same things, even if we're coming from different standpoints to try to drive a conversation and drive better decision making. So I think at this point, when we think about, again, data literacy, we think about an understanding of different kinds of data, qualitative and quantitative to a better understanding of what data describe in terms of being able to really get into the experience. Like what are the phenomena that underlie that representation that we have? What are the biases that we have in the way that we collect the information? What are its limitations? What might be missing? So three, we can think about new ways in which we can collect the pieces that are missing or make informed guesses about what are the parts that are missing. And then be able to deploy that in a way that is iterative, in a way that brings people along. Because ultimately, it's not just about building a highly technocratic organization, it's about building trust and building community. And I think that that's at the spirit of what the mayor wants us to do in Boston. The way that we solve problems should make us better off as a team and as a community. **Stephen Goldsmith:** Well, I can tell you are now a data fellow at the Harvard Bloomberg Center because you use a lot of fancy words. So let's connect your two roles here for a second. As we transition into AI, which is going to be subject of our podcast, let's take what your last really helpful answer and translate it this way. Let's take, I call him John who's the senior supervisor in the North Transportation District in Boston, or pick any other persona. How will he or his peer, I call her Mary in the West District, how will they take that data? How will they use that data? What difference does AI make to them? **Santi Garces:** Absolutely. I think that this is the piece that really has me fascinated, and again, in some sense, fulfilling this notion that being at the right place at the right moment, that allows you to work on very interesting problems. So ultimately what John and Mary are faced every day is trying to make the best decision they can, right? They're getting a bunch of requests from their supervisors. They're probably having to find some mix of responding to constituent requests, being able to do some preventive work that ensures that the things that they're responsible for in good working condition, and then keep solving these complicated questions around how to do the best that they can with limited resources. So in the current world, they are faced with systems that are pre-designed that we have to spend a lot of time making sure that the app shows the right information about what's in the queue. Hopefully we haven't had any systems that are broken. Someone hasn't sent an email to the wrong email address with the right request. So we're counting on a lot of systems working very well. But every time that the question changes. So for instance, let's say that Mary and John were expected to pick up mattresses because it's just one of the large items that get thrown out in the curbside. What if the state legislature makes a change in the way that we pick up mattresses? And all of a sudden we have to have an entire new process of how to pick up mattresses? So John and Mary's life are complicated. They're under a lot of stress. I think that the opportunity of generative AI, first, there's a lot of things that need to be true, and I think that we're just in the early days. So, there's...not to bypass all the skepticism, but what if John and Mary became empowered to be the best data analysts...like one of the best data analysts that we had in city government, but they would never have had to learn how to write code. They never had to learn about Python. What if they could produce better visualizations than any one of these experts that do dashboards? What if they all of a sudden could start doing simulations? What if they could say, "Hey, if I pick up mattresses first and then pick other large item pickups, what would be the average amount of time that it would take for me to pick up all the trash that I have to pick up in my district?" And I think that the reality is that we're getting close to that being true where we have the ability of putting tools that allow people who are subject matter experts to interact with technology and data and statistics in ways that they can put the subject matter expertise at the forefront. And we've taken out a lot of the obstacles for them to interact with the data to make better decisions in their job. And ultimately, they know what looks right and what doesn't look right because they're on the hook. Again, I think that with generative AI, we have these incredible capabilities where they're going to be able to do more. But not only them, everybody in the organization, their employees, the constituents. So John and Mary are going to start facing the choice that other people are going to be asking questions. "Why haven't you done it this other way? Look, I've tried to optimize the pickup of mattresses. Why are you not optimizing your mattress pickup in your region?" **Stephen Goldsmith:** You [wrote in Fast Company](https://www.fastcompany.com/90983427/chatgpt-generative-ai-government-reform-biden-garces-boston-goldsmith-harvard) several months back that you asked OpenAI to suggest interesting analyses after you uploaded 311 data. Tell us a little bit about that, and what do you mean by you asked it to suggest interesting analyses? **Santi Garces:** Yes. So again, this is the piece that is really, I think, exciting. So some of these generative AI tools, we have the ability of uploading open data. This is something that anyone in the city of Boston, whether they're an employee or a resident or just someone who's interested, and if you live in your community, you could use your own 311 data. We can upload that CSV, put it into the generative AI model, and ask questions in plain English or in Spanish or in other languages. And what will happen is that the generative AI model, these large language models will start generating computer code that allows some computer to make these statistical, these analyses with the data that we have. So you're able to ask really simple questions. In some sense, you don't even need to have a starting point. As I pointed out in the article, you could ask, “what are interesting trends that we could find in 311 data?” And the three things that it started suggesting was you could do time series analysis. You could try to see how things have changed over time. You could make geospatial analysis. You could see what is the difference between different regions of the city. What are different calls that happen in different regions? You could run statistical tests. You can do more advanced analysis. And even without any sense of how to be able to do these things on a technical level, you can have a tool that's helping you think about even what are the questions that you could be asking. **Stephen Goldsmith:** You and I have been working with five or six large cities, Dallas, LA, New York, Chicago, Boston, DC, Kansas City on performance data and how generative AI can open up the use of performance data to change the way cities operate, and also change the way government listens to its residents. If the goal is performance, is to improve the responsiveness of government in a way that residents appreciate, how can generative AI and the city's open data, how can you bring those things together for better communications for a more virtuous cycle of listening, understanding, iterating, and participating. **Santi Garces:** So I think communication and trust, I think will happen in two directions, right? From the government out to the constituents and then from the constituents back into the government. So let's start with the latter one. The experience of the city, there's usually much fewer city employees to constituents. In South Bend, there is about a thousand city employees to a hundred thousand residents. And the reality that means that people know more about what's going on in the city than government ever will because they're experiencing more of it all of the time. In government, there's also, we have different silos. We have different departments that work in different things. I think that the first piece is because constituents have the ability, they're consuming information in this way that we simply could not gather. There's, I think, new opportunities for us to be able to process even with things that the constituents are already sharing with us the common sections of 311. 311, we tend to think about what is the structured data? What is the service request type? What is the workflow? But often people are complaining or sharing perspectives about what they're experiencing that we're leaving blank, these unstructured data that is just really difficult for us to do anything with. And I think that, again, large language models are the best calculators of these incredible machines to process words. And I think partially what will start happening is us thinking about new ways of capturing and interfacing with constituents, new ways in which we can capture data that we haven't before. And in performance management programs like Kansas City and other cities like South Bend, we've been using survey data to try to compensate with the fact that when we receive 311 complaints, we're getting a biased dataset. We're getting only people that want to take the time and feel comfortable taking the time complain to us. So it's good information, but it's not a complete information. And again, I think the generative AI gives us the possibility of being able to consume information and unstructured data that we're receiving through these traditional channels and then through some of these other channels like surveys. Then let's think about direct information going in the other direction, information from the city, going back to the constituents. So one of the things that I've also learned is we do a lot of work, but it is hard for a resident to know all of the things that we do. And I remember one of the things, the project that I worked on earlier in my career with some really brilliant people, they envision us creating these neighborhood... I think that might have been one of the first interactions we had. We submitted it for Ash government innovation awards, is this neighborhood level reports where we were interviewing neighborhood associations about their priorities and then trying to create reports around the areas that they were interested in. And I remember in this particular neighborhood, they were concerned about illegal trash disposal like people just going around the neighborhood and throwing trash. And I remember, I think that we showed them that there was over two tons of trash that had been collected in the previous year. And they were very angry. They were very upset about the city, about this illegal trash disposal. But I remember the initial reaction when we showed them the data is like they're actually very grateful. They're like, "Oh, that's a lot." We show them how much that neighborhood accounted for something like 40% of all the volume of trash. So does it solve the problem of throwing the trash? No. But I think that they felt hardened to know that we knew there is this trust in us knowing how much trash we have that was being illegally thrown, created trust. And again, I think that in the spirit of generative AI, the more that people can know what we know, the more that they can go through the noise of having these large data sets that really have all this information, but nothing about what they actually want to know. If we can reduce that for them to be able to get the context of what they want to know and be able to better understand how is it that government is operating in their context, I think that it'll increase their trust. Anyway, sorry, big answer. But I do think that generative AI is an opportunity of helping flow information in both ways. **Stephen Goldsmith:** So we're going to have time to continue our conversation over the next months with other guests on generative AI. What would you hope that others in your position around the country and around the world, what topics do you think we should cause them to pay attention to? How could we catalyze more of the examples that you just gave? **Santi Garces:** I think my sense is that there is a fair amount of skepticism. I was telling someone that generative AI reminded me in the current state, reminded me a little bit of cellphones in the early days. They were big, they were clunky, and they were expensive. And it took decades from going from this big, clunky expensive thing that was not very nice looking to have a phone in everyone's hands, kind of enabling everybody to live their lives through those devices. So I think we should probably spend some time thinking about the things that are complicated around risk, around ethics. But I think that there's also a great opportunity to think about the possibility. So I thinking about how can we visualize data in new novel ways, in ways that are much more interactive. In some sense could asking questions out of your data set replace the dashboard? Is generative AI going to destroy the modern dashboard, which is like the workhorse of performance management progress? Two, can we start enabling new ways of thinking about data, like the creation of synthetic data sets? Can we use generative AI to create data that represents something when we don't have data about that thing? Which I think is interesting, and there's ways of doing this in ways that are productive. Thirdly, I think ways of enabling almost participatory data. This is something that I learned from Sarah Williams. Can these tools be vehicles for people to bring their own data? Data that expresses their experience that they typically don't share with us, but in a way that helps us create a more comprehensive understanding of a problem? Anyways, I think that there's many more, but those are three that I think that are fascinating questions that I've been asking myself. **Stephen Goldsmith:** This is Steve Goldsmith. I'm a professor at the Harvard Bloomberg Center for Cities, and we're talking to Santi Garces today. Santi is the CIO of Boston, one of the country's thought leaders and top practitioners in the use of AI to improve the way cities work. We look forward to our continuing conversations with him. Thank you, Santi. **Santi Garces:** Absolutely. Thank you for having me. I'm excited for the future conversations as well. ## 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. See also:- [ Artificial Intelligence ](/topics/artificial-intelligence) - [ Civic Analytics Network ](/topics/civic-analytics-network) - [ Performance Measurement ](/topics/performance-measurement) Share on:- [ Facebook ](#) - [ Twitter ](#) - [ Linkedin ](#)