A Mayor Built the AI Tool Her Own City Needed

Monisha Saldanha (00:01) Welcome to Colorado Tech People, the podcast where we talk with the founders and leaders using technology to solve real world problems. I am your host, Monisha Soldana, an executive with 15 years of product management experience. Today's episode explores how AI can make government data more accessible for both public agencies and the citizens they serve. I'm joined by Nichole Sterling. founder of My Town AI, a company working to turn complex government data into usable, actionable insights. We'll dive into the hard decisions behind building in GovTech, how AI enables this shift, and the real impact it can have on transparency and decision making. Nichole, welcome. Thank you so much for joining me today. Nichole (00:51) Thanks for having me, Monisha. Monisha Saldanha (00:53) So let's dive in. What inspired you to start My Town AI? And what problem did you see in how government data is accessed today? Nichole (01:04) Yeah, so it started because I am a publicly elected official. I sit as mayor of Nederland, Colorado. And it was through my public service that I started to have just a very intimate experience with how governments use their data, where they put it, and you know, all the things associated with that. And when generative AI sort of hit the scene back in 2022, there there was a lot of exploration of all the use cases, and at that time I was sitting on the Board of Trustees and I thought, well, my goodness, why not in local government? That's probably one of the best use cases for AI. and And so my town AI was really kind of birthed out of that pain, the pain of sitting on that board and wanting access to data and just not being able to find it very, very quickly. All of us are very familiar with the clunky local government websites where you go up to the search bar and you try to you know, plug in your keywords and it just doesn't give you what it is that it you need. Or it gives you a hundred different things, but and so you have to search through all of those. And so the the most simplest use case that I started with was let's just pull in these board meetings, these agendas, these minutes, the recordings, so that you know, out of for very selfish reasons, so I can have access to those prior decisions that boards made or you know understanding why certain decisions were made the way that they were. So it was it was really kind of just born out of my own pain. Monisha Saldanha (02:39) Interesting and and as you mentioned, government data exists in abundance. So it's not a lack of data. But why is it still so difficult to use effectively? Nichole (02:51) Because it is in multiple spots, local governments, unlike some of our private industry companies, they haven't necessarily had the discipline or the thought to have a concerted data strategy. I mean, half our documents are still sitting in the community center in a box, right? They're It's still in paper format. And many jurisdictions are like that across the United States. So the not only is the data maybe electronically in all these different places, but also physically in different places that just makes it hard to access. And not until you know generative AI sort of hit the scene was there the awareness that, wait a second, what if we pooled all this data together and allowed like a generative AI application to look over that data so that then I'm not having to just do a keyword search. I can actually ask, hey, where why was this decision made? Right? It's it's way beyond a keyword search. You're asking for the intent. And so to be able to provide that on demand that's that's huge within the local government sphere. Monisha Saldanha (04:08) And What were the hardest early decisions in defining the product and who it's really for, governments, citizens, or both? Nichole (04:18) Yeah, it was never a question for me, that it would be for internal use first. Cause again, I was sort of modeling it off of my own experience. And I knew that internal staff would absolutely use it. But you see the same gaps in data fluency and and information understanding among residents as well. I mean, if I was having a hard time finding information, then I guarantee you our residents and many of them, you know, raise their hands and say, Where do I find this information? It's just it's these These local government websites are just aren't always set up for that. And so it was never a question of who I was going to serve. It was always going to be internal staff first. And then once internal staff felt comfortable with the AI, understood how it worked, what it did, what it didn't do, because you know that's part of our messaging and our training at MyTown AI, is that AI isn't great for everything. And you need to learn that fluency and that understanding in order to really cap capitalize on what it is good for. So once our internal staff folks understand it, then they can choose to go resident facing and and give their residents the same access to the information that they have. And because of course in the local government, it's all about transparency, as much information that you can provide. But now with Generative AI, you can you can provide it in a different way. You It's more accessible than it was before, all the information. Monisha Saldanha (06:00) How do you validate that this was a problem we're solving, besides your personal experience, and that people would actually use the solution besides yourself? Nichole (06:10) Right. Well, I mean, I think that that was that was part of the impetus was that I would use it. And Monisha Saldanha (06:18) Mm-hmm. Nichole (06:19) because again, I was I was solving a pain point that I was already experiencing. And I I mean once when you first go to market, it is all when somebody first raises their hand, like, yeah, I'll pay you for this, you're kind of like, really? Okay, yeah. Yeah, no, that's great. Yes, please. but So I knew that it would be applicable to many other jurisdictions. so that I I I sort of had a you know, just a a gut feel on that. And but the idea is now like in what you've seen across many, many AI companies is that what are the LLMs, you know, these big LLM companies like ChatGPT and Claude, what are they coming out with since it's been a few years since generative AI hit the scene, so we're in year four. What are they coming out with that has displaced some of these early AI companies? And I think that's why it's so important. And this is a tried and true, you know, kind of wisdom, even before generative AI hit the scene, is who do you have at the helm that is developing this product? And you know, for me, I'm a subject matter expert in local government. And that was that's a true thing that you want before generative AI, but it's even more important now. Now, because you did see a lot of these AI companies come up with just this or that to try to make a quick buck. And now you've seen that they have kind of gone by the wayside because these LLMs have come out with replacement technologies or technologies that you know just kind of put them by the wayside. And so to have these subject matter experts that are know both their expertise and also how AI works, that's huge. And you know, especially with the onset of vibe coding i I'm teaching local governments how to vibe code right now and you would think my gosh are you really that like local governments are open to do that absolutely because if you think about the persona that we're dealing with these are resource constrained entities these are folks that have the subject matter expertise what if you gave them the ability to you know these AI coding tools to create the thing that they've always wanted or to create the thing that's in their head. Now they are, they're doing that. And so it is also shifting the relationship that local governments have with their vendors. Because if I can go Monisha Saldanha (08:52) Yeah. Nichole (08:52) and I can create the solution pretty quickly with a couple of AI tools, then what does that relationship and that handoff look like with AI vendors into the future? And this is I'm a member and co-chair the GovAI Coalition's Use Cases Committee. And this these are these are dilemmas and things that we talk a lot. lot about is how is AI shifting the mold within local government. But yeah, to be able to just have that subject matter expertise to to be able to exist today four years after generative AI hit the scene, that's a true testament to those folks who really understand the workflows and the things that AI needs to replace in order to provide better resident services. Monisha Saldanha (09:41) Was there a moment where you had to significantly pivot your approach based on user feedback or real world constraints? Nichole (09:54) yet I'm I feel like I kind of stay a couple of paces ahead of where most of these local governments are simply because you know I'm I'm working through the same realities, the same constraints that they are when I'm in a a city council meeting or I'm talking to staff and they're having to you know supply an analysis of something. So I'm I'm right there in the trenches but because I have the fluency in AI and architecting AI systems, I'm I'm probably able and more quickly able to see the use cases that can pop up and to to to pursue those use cases more quickly than a lot of other folks in local government. So at this point I haven't had to other than you know kind of thinking through restructuring like a go-to-market, but the product itself, I feel pretty I feel pretty honored or I I mean I guess at least lucky, I feel pretty lucky that it sort of stayed true to the vision that was originally in my head. Monisha Saldanha (11:04) What types of data are you working with and what makes them particularly challenging to process or interpret? Nichole (11:12) So local governments have, like you had mentioned, a lot of data and a lot of data over the years. and we deal with both the structured and unstructured, your unstructured data being, you know, like the documents, heavy word text, your structured documents being more are the most structured doc or the most structured data that we deal with is like GIS information. And Just like we see every local jurisdiction has a different municipal code, they also oftentimes will structure their GIS information differently from each other as well. So part of what we've had to create on the back end is a normalization and a cleaning mechanism for some of this GIS structured data so that it regardless of who we're interacting with, we can clean it and structure it to what we our standards are and what we need. And of course, then once we start to dive into some of these documents in local government, they're heavy on tables and heavy in things, just like the structuring of the documents. Everything has a chapter and everything has several articles that you would see in you know, just very heavy legal documents or like land use code. So we've had to be very deliberate about the type of techniques that we use in order to read these tables well. And because you know, some of these tables might span across multiple pages. And if they do, then the large language model that's reading our extraction might miss context because that table is structured, that table spans over multiple pages. And so if I'm trying to, if the LLM is trying to read something on page three of the table, it might lose all the headers that were on page one, right? So we've had to be, like I said, very deliberate about how we interface with those types of documents to make sure that the accuracy stays up. Monisha Saldanha (13:24) And How does AI help bridge the gap between this complicated raw data that you're mentioning and actionable insights for everyday users? Nichole (13:33) Yeah, I mean so once we go through the process of cleaning, of normalizing, and we've got, you know, our ingestion document processing and extraction everything down, what it then provides is the ability for these users to ask just in natural language. And that's really what I see is that you know, generative AI provided was a different interface for users. I remember being I remember doing a presentation to one of Colorado's local governments and the lawyer, the town attorney was in the room and and he asked, so wait, so you're sure like we don't have to do like a Boolean search for this? Like we don't have to, Monisha Saldanha (14:13) Mm-hmm. Nichole (14:13) you know, do the little asterisks and look for the keyword, Monisha Saldanha (14:16) Mm. Nichole (14:16) and it's like, no, you know, you can start to ask just in natural language. Hey, what was that decision around the Big Springs egress back in 2025? Right? Or or whatever it was. By being able to ask in natural language, it just opens up the access to all this data so much more easily. And what we also start to see is that many local governments are starting to implement like AI translation services to even make it the information. information more accessible because they could have hundreds of different languages represented in their municipalities and before it would have been so difficult and or time consuming and or expensive to translate a land use document into all 100 of those languages. But that's one of the things that generative AI is very good at is because it is trained on human language, it can get very good at it. You still have to look for bias, because even language has bias in it, but it at least gets these local governments a lot closer to reaching their constituents than before. Monisha Saldanha (15:26) Yeah, and this next question kind of touches upon bias. What are the biggest limitations or risks of using AI in this context around accuracy, bias, or trust? Nichole (15:39) Sure, I mean just in language, right, there's a lot of that bias. Where a lot of my work happened in the nonprofit that I co-founded called Women Defining AI. I mean, this was a part of our everyday use. We would teach women and non-binary individuals how to use AI, the limitations, the risks, the bias. And it I mean, every day there was an example of, you know, somebody going in and prompting one of the large language models to say, hey, give me an image. of I'm you know one of the classic examples was and she's actually an advisor for my town AI she's a lawyer a very credited lawyer. As she she had prompted several times the large language model to say, Hey, can I have an image of a lawyer? and you know, she's She's Asian American and she said, you know, Asian American, and every time it would provide a a male image instead of it's like, well, wait a second, there Monisha Saldanha (16:40) Hmm. Nichole (16:41) are female lawyers out here in the world, my friend. And so, Monisha Saldanha (16:44) Mm. Nichole (16:46) and so while local governments are less likely right now to be using of the image generation they they are using it and so there is always where we have to have the conversation about where the bias does exist especially when i'm doing presentations or keynotes across the United States because there's the the examples still are are here even four years in the bias is still in there where we also see it in the local government realm and what we have to be careful of is in the data so you know how is the data collected where are the potentials for the bias? A lot of local governments work with consultants, and sometimes this information isn't always addressed or identified on behalf of the consultants either. It's just kind of taken for face value. And so what this is causing a lot of local governments to do is just question more. Because if they want to use these AI systems, not only are they having to get their AI digitized, but then they're also having to address some of these concerns, like because it's garbage in and garbage out. If your if your data is biased or inaccurate going into the AI system, it's going to be biased and inaccurate coming out. So all of these things that we look for just within the large language models in general, we're having to look out for in local governments as well. Monisha Saldanha (18:11) How has building My Town AI in Colorado influenced your journey as a founder? Nichole (18:18) Yeah, so I mean, we're considered a bit of a tech hub, not as much as like Sa San Francisco or Austin, but we do have a lot of talent and I've only ever been in startups in Colorado. and you know you s my my One of my largest investors invests in Colorado rural companies, which is also very cool. and you know, There's always the bias that I think happens when you have founders that are outside of these tech hubs that they must not they must not really care enough to move to San Francisco. And I'm just Monisha Saldanha (19:02) Terrible. Nichole (19:04) I can't live in San Francisco. You know, Colorado Monisha Saldanha (19:05) Yeah. Nichole (19:07) is where my heart is. I obviously decided to pursue civic you know life within a small Colorado town. And so I it's meaningful for me to continue to build out Mytown AI here in Colorado to show that yes, we have the talent. you know, The folks on my team are were from previous startups that I worked with, and so it does but it does sort of sometimes landlock you in terms of access to other funding. because if because there is the bias out there that if you're not in San Francisco or maybe New York, you're not as serious or whatnot. But I do see more of a tech scene and especially the funding coming into Colorado, but it could always use more. And by being a female founder of a tech company, I I hope that I get to be in a position of raising the voice even more so for female founders in Colorado of tech companies. 'Cause there's there's there's some of us out there, but of course, we always need more because the perspective that we bring is differentiated and I don't think that there's somebody out there that could be building what I'm building just because of my unique perspectives. Monisha Saldanha (20:29) And How have you found fundraising in Colorado? Nichole (20:33) There are a few select folks. most like I said, my largest fund My largest investor is here in Colorado, and then the rest of them are outside of Colorado, actually. So it has been a bit of a hybrid approach. where I'm obviously going to pursue and knock on the doors of our Colorado folks, and and also just for the sake of the company, you know, make sure I'm also building partnerships and building those connections to folks outside of Colorado as well. Monisha Saldanha (21:14) What advantages or challenges come with building a GovTech company outside traditional tech hubs? Nichole (21:22) Well Building a GovTech company in general is not a sane thing to do because most VCs, most inventor investors have no idea how local government works. All they know is that it's long sales cycles, and that's about what the word on the street is. And that it's you know it's difficult to be in GovTech. Which is, you know, I have to remind them, I was like, yes, at some of the higher levels. levels of federal government, state government, yes, your sales cycles can be two years long, but at the local government level it's a it's much smaller than that, or much shorter than that. And but Still many investors just don't understand how government works. I remember I was on with an investor and I I mean what sounded like kind of a stupid question to me is the investors asked, well so how do local governments find out about technology? And right there I could just I was like, this is this is not who I want. This is not Monisha Saldanha (22:29) Thank you. Nichole (22:30) going to be you know, a helpful person. And you know, and some are starting to see that local government has a lot to offer in terms of pursuing local government. They can be very sticky, sticky customers. They'll stick around with you. They are willing to pay, and many have obligations to pay for certain software at certain times in their life cycles. So it being an insider into the local government world is a huge advantage to me because I I know the procurement cycles. I know what they need to focus on. I know the the the folks just from outside of Mytown AI and just the government work that we've done on a regional standpoint or whatnot. And so for me it was like so obvious. I'm just like, why? Why isn't everybody Monisha Saldanha (23:25) Yeah. Nichole (23:26) doing SaaS for government? But it is kind of a scary or unknown thing for a lot of folks outside of the industry. Monisha Saldanha (23:36) What advice would you give to founders building products for government or public sector use cases? Nichole (23:44) Go and join one of your local government boards, honestly. I think that is We get a lot of vendors from outside of government who have a solution maybe that they created for some other industry, and then they say, I'm gonna go sell to government, and so here it is. And not necessarily having an understanding of how we actually do things within local government. I see it quite a bit. And so my recommendation, 'cause I on the Gov AI coalition, I co-chair the use cases committee, and every vendor has to come through it has to come through me in order to present to the coalition. And I'm just I'm flabbergasted by how many lack a l some sort of local government experience. I mean, this is something as simple as, you know, being on the planning commission or volunteering for your parks and recreation board. You don't have to go all the way up to the Board of Trustees or to the City Council. But there are opportunities to understand how local government operates, because it is, it can get complex. And if you lack that information, then that does that definitely puts you at a disadvantage. So I think again, the the same saying still applies even before you know generative AI hit the scene is know your customer, and there's no better way to know your customer if you've got some of the insights sitting in that seat of the customer. Monisha Saldanha (25:20) Can you share a real world example of how Mytown AI has helped a government or community make better decisions? Nichole (25:27) Sure. I mean, not only just my own experience in Nederland of being able to surface, hey, why were these decisions made way back then? Before like even I got on the board or whatnot. So that context, that historical context is huge. But one of the use cases that we received that kind of started us more on the path of our simulation. Because I I I always knew I was like, you know what's what's the power move here is if you can collect this local government data and then start to build simulations for them to test their own data, to prototype their own data. Because as much as like the first use case of just let me just scan our data with AI was super powerful for me, I wanted the desire to test and to prototype. So I I received a message from one of our towns, and it's a little town in South Colorado. And sh it was the town manager and she said, Hey, I I have a mandate from city council. I've gotta find one of the parcels within our jurisdiction where we can have an R V park. And it's like, well, that should be easy. Why is that so hard? Well, the reason why it's so hard, and this is what a lot of folks don't understand, is there's all these tools where you can go online and and you can see, what zone is it in? What zone is an RV park allowed within this jurisdiction? Well, yeah, that's the easiest search and the easiest part to do. And maybe you can do some setbacks and understand, you know. Yada yada yada. It's so much more complicated than that, all these different uses that governments have. So for instance, in their particular land use code, an RV park was not just dictated by zone, it was also dictated by the the part the the size of the parcel. There was a minimum requirement that the parcel had to be. It also had to be certain amount of ways away from water, like streams or whatnot. So that had to be factored into it as well. And that was the first use case, because I I I kind of always knew that again, like I said, that we would go into the prototyping and the simulation, that was the use case that jumpstarted our s our simulations work. And it was basically being able to go into every local government land use code or municipal code and find all the constraints. And so I've got this sort of the saying is that we help you find all these invisible constraints of government decision making. And this is one of those. It was very you know, just she just had a simple mandate, find the parcel of land where the RV park can go. But in her land use code, there was just all these invisible constraints that had to be con considered. And this is where many local governments will go and they'll farm this work out to consultants to do and pay anywhere between $250 to $500 a an hour to do it. And I'm saying we have the data, we have the intelligence now that we can kind of save money in those other areas and use consultants in very deliberate ways. and And utilize our own data for our own benefit because you know yesterday she needed the RV park scenario tomorrow she might need to know where childcare can go because she's got a business owner that wants to come in and open up a childcare facility and childcare dispensaries breweries are some some of the most heavily constrained uses within our municipal government sometimes they even have state requirements that have to also be considered. And so we're s we're building the constraints engine to consider all of those things. Monisha Saldanha (29:22) And how does this shift impact smaller municipalities? Do they have enough data or strong enough tech teams to be able to benefit? Nichole (29:32) They have plenty of data. there the data that they if you think about it really it's it's very interesting. Your small municipalities are typically having the same amount of meetings as your larger municipalities. They may have less boards, you know, advisory boards in doing the work, but they still create oftentimes a similar cadence or volume of meetings. Now sometimes I like when I go to read a packet for a board of trustees meeting or like a city council meeting, I've easily had packets where it's 400 pages long just for one meeting and we've got to read those. And and so yes, they have plenty of data. Do they have the tech teams? No, which is one of the reasons why I started to build is like I can pull this data together. I know what these smaller towns are going to need. So what if I pulled it together and it was just this plug-and-play system, right? They could then bring in their own data, such as the board meeting agendas and recordings. We can create the automations on the back end to load those in. And now they have a fully functioning AI clo platform that's a closed system. It's not bouncing out to the internet. which is what a lot of local governments. One. So yeah, I mean that was also part of the reason is just many of our towns on the smaller end don't have the resources, but also even on the larger end, because many local governments are resource constrained to begin with. And yes, they may have instead of you know, Nederland has one community planner. You know, a little bit larger may have two, but they're still they're still running through a lot of work to get this stuff done. Monisha Saldanha (31:22) And are you selling the product as like a SaaS model, like a subscription? Nichole (31:27) Yep. Yep. because that was also part of the thesis is it doesn't have to be a bespoke, you know, situation for each local government. There are patterns and there are sort of foundational data sources and if you just create the back end that can pull all that in in an automated way, you have a SaaS product ready to go. Monisha Saldanha (31:51) Right. And looking ahead, how do you see AI reshaping how governments operate and interact with citizens over the next decade? Nichole (32:01) Yeah, so we have some of these sayings in government. 'cause, you know, everyone says, government's so slow and there are actually very good reasons why some of our processes are slow. It's because we have to take in multiple data points, it's because we are beholden to our residents, and having that public process is really important. And in so There are places in local governments that we want to keep slow and deliberate for those reasons. However, one of the things I challenge a lot of the local governments that I speak to is those those processes aside, where are the other processes in local government that can be collapsed or condensed or you know made faster through AI services? So that we keep the right processes slow and deliberate, and then we collapse and make better s you know resident services in the areas that we can. So one of these might be information access, you know, like we've been kind of been talking about and how MyTown AI started. Another big use case that we see is around permitting. AI is actually very good at reading drawings, architectural drawings, and assessing whether or not something's going to meet code. So either in the pre-development proposal stage or even once it gets to permitting, those are wonderful things to use AI for because then it puts your staff, your community planners, your community development folks out in front of the residents having the conversations. So this is really an opportunity for a lot of local governments to identify where those processes are that could be used for AI and where the processes that we still need to keep slow, deliberate, and are human-centered. And I think it's not just local governments. government who's working through this. It's it's all industries where we're really trying to we're having a reckoning of what can be AI focused and what needs to stay human focused. Monisha Saldanha (34:19) And What has been the response to the introduction of AI in your tool by the governments? Are they like open to it? Are they skeptical of it? Are they fearful? What kind of reception have you had? Nichole (34:34) I think you see still see this across all industries. You've got some folks who are gonna be early adopters, right? That's like on the bell curve, your your first 10% of folks who are just jumping in with feet first because hey, it's innovative and it's cool and they can see the use cases much more quickly. And then you've got 10% that are gonna be over here at the other end of the bell curve who are like, hell no, to to AI. but because you know they've got like some icky feelings about it, whether that's from an environmental standpoint. they don't like the fact that it's replacing jobs and so there's just there they say no. But a majority of folks, the 80%, 70 to 80%, are gonna be in that middle who are curious and or just haven't seen it work for them in the right way yet. And so that is where even when I'm talking to folks, you know, I'm I'm obviously leaning into the the the part of the bell curve that are curious and that are open to it because they have a particular pain point that they're trying to. solve for. And so I generally interact with those folks that are are are leaning into it. But yes, I've met many over the course of the last several years who just don't want to have anything to do with it. And and that's okay. And I think, you know, every organization has to make decisions for themselves of how they want to lean into AI for the right reasons and put their policy and their guardrails in place. And you can't force anybody into it. But the ones who are leaning in are starting are seeing some some incredible benefits. So those are the folks I I tend to continue to interact with. Monisha Saldanha (36:25) Fantastic. Well this has been great. I have a final question for you. What is one book every builder should read and watch? Nichole (36:34) one book every builder should read and why. Well, so I think a lot of folks who have been in AI in the pa or like building AI companies Have have gone in and l Like I had mentioned before, you know, I co-founded a a a nonprofit called Women Defining AI. But the bias piece sticks with me everywhere I go. It's just the lens that I look through things. When I'm designing products, when I'm looking at data, all the things. But not every AI company that got into the AI race has actually taken a step back to understand. like what are the ethical implications of AI? What are the responsibilities that they should have? And so I always recommend the the book right there, Unmasking AI by Dr. Joy. I had the amazing privilege of meeting Dr. Joy in Gloria Steinem's New York apartment a year or so, about it maybe a year and a half ago. And it's just, it's just so much clearer that when you have that lens of looking for bias and discrimination in some of these tools, it's just it makes for a better product for everyone. But not everyone comes with that. So I always I always recommend her book, Unmasking AI by Doctor Joy. Monisha Saldanha (38:15) Thank you for that recommendation. That's a new one for me, so I'll add it to my list. Nichole (38:18) yeah, please do, please do. Monisha Saldanha (38:23) Well, thank you for joining me today, Nichole. Nichole (38:26) Yeah, thank you. Really appreciate it. Monisha Saldanha (38:28) And thanks to our listeners for listening to this episode of Colorado Tech People. If this conversation on using AI to make government data more accessible sparked new ideas, consider sharing it with someone working in tech policy or civic innovation. Don't forget to subscribe so you can hear more conversations with founders building impactful solutions across Colorado and beyond. Until next time, keep exploring how technology can make systems more transparent, accessible, and useful for everyone.