Transcripts

Transcripts are generated using a combination of speech recognition software and human transcribers and may contain errors. Please check the corresponding audio before quoting in print.

Joe Supervielle:

Santi, how would you respond to an employee outside of your own IT department who asks, "How will AI make my job easier? Let me do more valuable meaningful work instead of replacing me or even a certain percentage of the people on my team or the type of job I do."

And we're not here today to just kill that myth of like, "Oh, AI is going to replace all of us eventually. It's not really what I'm asking you. But again, just on the practical, like this year, not in the future, if a local government moves forward and they start having some meetings or discussions about it and someone raises that concern, but also opportunity, how do you address that?

Santiago Garces:

Yeah, the first piece is cities change the way that they work all of the time. I remember visiting many older fire stations, and you see that there's the stables where they kept the horses. There's no horses there anymore. We change the way that we run cities all of the time, and it's a natural part of government and a natural part of cities. And it usually means there's some transition, but it's not kind of catastrophic. I would say my encouragement up to some point is like you should find the answer to that question yourself. And in the sense an IT department that is working well is helping you do that correctly and safely and with the right resources. Some things that might make people's job easier and why generative AI could be really powerful is I find it that it is very helpful at getting faster to that first draft of almost anything.

First draft of an analysis, the first version of a data analysis, the first version of a job description, the first draft of an email. Part of the challenge in government... And this is why I'm a little bit maybe more optimistic about... I don't think that this will displace a large number of government employees. Government is fairly understaffed in general. I think when you look at the long-term trends, especially in local and state, there's been a declining in number of employees. The same people are doing more and more work, and that usually means that we end up spending much more time just getting to an answer rather than thinking about is this the right answer. So I think that gen AI can make you shift a little bit the responsibility, especially for people that struggle with writer's block or whatever. It puts you in the position of being an editor rather than that first version writer.

And I would add, because it allows you to iterate quicker and faster generally, hopefully getting to a higher level of quality. You can also do it in a way that is much more collaborative. You can work with others to try to refine the work because we just usually don't have time. We are like, by the time that the memo's due, you've just drafted it because you're busy with other stuff. So I think that the less time that we're spending doing some of that basic piece, some of the part that is tedious and onerous and slow, the better.

I think even for graphic design, again, graphic artists and people that are in creative industries are particularly vulnerable to being displaced by these type of tools, but my hope and the way that I try to use graphic design tools like the image creators, DALL·E, Midjourney, and so on and so forth is they allow me to be a better customer of my creative team. It allows me to express with images something that I would not be able to do that I usually am not able to do, so that they can... In my experience, really good creative people and graphic designers really hard... They're irreplaceable, they're amazing, they add some understanding of the problem that goes beyond just the skill of drawing. So anyways, those are some ways that I think that people can use the tools.

Joe Supervielle:

And that's an example of the AI being additive and not replacing. Whether it's creative specifically or not, I think that is a good example of the humans has their skillset and the AI can augment it or supplement it in at least best case scenario.

Santiago Garces:

And I would just add really quickly, look, I've had conversations with people, especially in smaller cities and towns where it's not even about a choice about like are we displacing someone? It's like they don't have access to some of this kind of talent, policy analysts, data analysts, graphic designers. So this is like for some people and for some organizations that are really small and under-resourced, this is about just creating capacity that just doesn't exist.

Joe Supervielle:

Is there an early opportunity where not just employees but where AI can help the resident? A lot of times before these type of interviews, I asked for some feedback from people not in local government, and the first thing everyone thought of was, can they improve those awful bots that answer the phone and put you through the maze of menus and this and that, and God forbid, even getting to something like permitting? How can it help the relationship between the local government and the residents? Hemant, you want to start with that one?

Hemant Desai:

Yeah, I think, I'll certainly have Santi add to it too, but when I was in a county not too long ago, I do remember old days where our citizens are calling into a regular phone, and sometimes having to wait several hours, several days for response. I think as a low hanging fruit, if I may, that would be my first target is how can I leverage a gen AI kind of tool that can supplement permitting department, a planning department, or even public health social service department where call centers respond in a much faster way. Not having by chatbot, which we had that before too. But now with gen AI, can become more relevant, targeted responses provided by a backend of data that's already existing within those repositories of local government, right? Knowledge basis, if you will. So I see that as being an optimizer versus a replace.

Santiago Garces:

Yeah, I would provoke, maybe at the risk of being a little bit provocative-

Joe Supervielle:

No, do it! Provoke away.

Santiago Garces:

I think if I was a resident, which I am, of a city or a state, I would use some of these tools to try to go and look at some code about something that I want to do. So I'll tell you, I use a lot of the tools to try to process and understand our procurement code, the Massachusetts 30B, which is the state's procurement code. It's hard and dense to read through, but I could ask it to try to give me a version of it that is summarized, summarize the chapter, and try to make it simpler and give it to me plain English or plain Spanish. There's some risks, but again, usually we would have to just go and hire a lawyer to try to... It's like a zoning code, permit, the housing standard code, I would use it to download open data from my government to try to analyze it.

You could go and see if there's differences in the response time for one request between neighborhoods. I think that gen AI could be the ultimate equalizer in the experience for constituents with government, because it, and again, it's not risk-free. But this is also why I think that it's so important for us in government to get better at it is because this is a great equalizing tool. Government is all about really complex and obfuscating processes, and it's really comprehensive, and it can be really hard and daunting to navigate when you're a constituent. These tools can be helpful in doing that. And you don't have to wait for government to go and build the tool. You can try to use some of the resources that exist to try to get the tool to help you. So it's provoking because it's a little bit nerve-wracking for people in government, but I think that it's happening. That's how I use it to help me navigate the complex bureaucracy that I live and operate in.

Joe Supervielle:

Yeah, you said you don't have to wait. I think that's a key phrase. So whether you're the employee or just the resident, and whether it's a local government or a business or whatever else, healthcare, we won't go there today, but the individual can use these tools themselves to try and get a better starting position at least on whatever it is they're trying to get done, and not wait on the entity or the huge corporation to figure it out for them. Is that fair?

Santiago Garces:

Fair. And with all of the caveats that existed, sometimes you might still need a lawyer, and you might still need all these things, but I think that for a lot of people that don't have access to a lawyer, even people with disabilities, people that have a hard time reading, dyslexia, it might be a helpful tool to help support some of these things. So again, it is not without its risk, but absolutely I think that it's such a powerful tool, and we all have to... Regardless of whether you work in government or not, there's times in which you're on the other side of the process, and you're the one who's petitioning and requesting something, and you might need to go and look at the 200-page manual in how a decision is made in government.

Joe Supervielle:

Thanks for the great information so far. We're going to try and cover a few more questions, not quite speed around, but a little faster. So first question, and we covered this a little bit, and Santi explained like, "Hey, experiment," maybe that's already the first step, but what are some initial steps, Boston, or even Hemant at ICMA, we have taken to embrace AI or maybe step two, beyond the experimentation, what next? What can the listeners do next if they've also experimented already?

Santiago Garces:

Yeah, I think a key piece that is helpful is understanding what you value and what are the things that you're have some fear of losing. So I think that in some sense, that is the part that's not about the technology, but how it fits within our value system and our decision making. So in Boston, we valued transparency, so that's partially why we decided to have a guideline that says that we're going to disclose that we use it.

We value community, so we wanted to make sure that this is a tool that's not community and people. This is not a tool about replacing work, it's a tool about enhancing and supplementing. So I think that thinking about what you value and how you think that it would be useful helps you go from experiment to knowing when and how you could use it.

Joe Supervielle:

That's a great answer. Can you elaborate a little bit? Everything you just explained, would you consider that a roadmap, a data governance policy? I guess, again, it doesn't necessarily matter what you term it, but what document or where can you get this in writing so the rest of the organization can start to understand it and begin with some processes to put it in place?

Santiago Garces:

So we, again, put together these guidelines that are on our website about the use of generative AI, and we tried to be purposeful in one instead of going for really comprehensive and complicated answers around policies and scenarios. We tried to make it simple, and then we tried to get resources for people around what are things to do and what are things not to do. We envisioned that it is a living document that will continue to evolve. And we've been based on conversations with people from ICMA and from other organizations, learning about what are areas that people think that should be improved upon. But that's what we tried to do.

And then we've also tried to create a couple of little pieces of infrastructure, a Slack channel to be able for us to talk about it amongst ourselves, to be able to share resources, how people are using it. We've started to do some training just trying to bring awareness to people about the tools and how is it that we think that they could be useful.

Joe Supervielle:

Okay. Without getting too far into the weeds here, can you define the AI architecture? What are the components, specifically the platform, and is it public, private, hybrid?

Santiago Garces:

Yes, so this is complicated question. So I would say, again, for the most part, we've been trying to be opportunistic and try to do like in the spirit of good experimentation, aligning the cost and the resources with where we are in the process. So a lot of things that end up requiring some investment that's in things that we end up looking at. Architecting. For instance, we've been looking at, this is already in the public domain. We've been looking at using large language models to create simple summaries of the types of items that got discussed in a council session, just even to be able to create a better web experience of the record of the voting record of the council. That started with a brief experiment where we took open data that had those records and tried to get the LLMs to give us shorter summaries, and now we're in the process of researching and developing how is it that we could use an API to try to produce these types of summaries.

So the architecture and the solutions depend on the use case, but we're trying to make sure that we're minimizing the risk, minimizing the operational overhead and cost, and then, again, always thinking about what about the AI can be additive or enhance the experience as opposed to just thinking that we should make the AI front and center because people will think about it. If AI is effective in the experience of people, it should feel like it's not there. It just makes it a little bit easier for you to interact with your government.

Joe Supervielle:

Hemant, do you have anything to add on employees, residents, just staff members? Even at ICMA, what is your team doing to help the rest of us embrace AI, and then, therefore, what the members can eventually see down the road? Where's ICMA with this?

Hemant Desai:

For us, as small as we, as you know, we began our journey a few months ago beginning with spreading the awareness through newsletters and other communication medium or SharePoint site, on Slack type of a channel. But for us, our journey began with creating awareness training. So we have several training sessions scheduled for ICMS staff. More importantly, making sure that staff members, because, as you know, whether we as technology departments promote AI or not, people are already using it, right? Because accessibility is there now. So for us, it's more about herding everybody in the same direction, if you will, as best as possible, and providing the tools to ensure that there's a responsible experimentation. So providing training, doing a ground up survey with staff would be one step I would recommend just to get some feedback so that you can understand and read it to find out how many of the staff members are aware of gen AI, and then when are they going use, what are some of the use cases they see?

Because as you asked, Santi, very good response by him about the architecture, understanding what platform the architecture requires so many different moving parts, because each subdomain of any city or county, such as finance, legal, purchasing, HR, social services, public health, will have so many varied needs. So as you create the architecture, make sure you have taken a bottom up approach of validating your assumptions you may have, but then providing a training so staff members even know what to bring to the table when they provided a feedback.

Because I just came to learn from some of our staff members, they said, "Hemant, you know what? You asking us, 'How could we use gen AI,' we don't even know where to begin to ask you questions," right? Because even to ask questions is some awareness of its capabilities, right?

So doing the awareness training through different modalities, whether it's SharePoint, email, newsletters, and conducting hands-on lunch and learn session, round back sessions, which is what we'll be doing shortly, I think will be my recommendation to any city or county to begin their journey, and while experimenting on the backend at different levels with different pilot team members, if you'll.

Joe Supervielle:

Okay, last question. Space Odyssey, Terminator, Matrix, there are a lot of go-tos for the AI technology movies, shows, books, and usually doesn't end up too well. But I actually have a feeling Santi's more positive about these things, so maybe he will throw a curveball in. But what is your favorite... Do you have any kind of hidden gems people might not know about or haven't seen yet? What's your favorite? Santi, go ahead.

Santiago Garces:

That's a great question. If we wanted to be a little bit more optimistic, maybe WALL-E, the Disney movie. But look, I think when it comes to analogies, I think, and there's a lot of people that were thinking about Oppenheimer in the context of gen AI, like the doom of humanity. I think AI has so much other uses, like yeah, that can be dangerous. There's things that we don't know about it, but I think that there's... I would say the only version of AI is not only killing robots, there's a number of things that are much more pleasant than helpful.

Joe Supervielle:

So does it frustrate you when those movies... Obviously they're entertainment, they're not necessarily predicting the future, literally, but is that your response when people ask you about like, "No, it's not actually like that. It does this, this, and this"?

Santiago Garces:

I love them. They're incredibly entertaining and fun. Terminator 2 is an awesome movie, but I think there's a lot of unknown and you don't want to... And I think that that's partially with the anxiety, and we as humans always... Whatever's unknown seems strange and far and scary. Being a little bit more rational about it, again, I think that there's reasons to be scared about it. But for instance, I would say the biggest harm today of AI might be the environmental harm. The amount of CO2 that we're emitting and training these models. So it's like not letting the hype let us get away with what's really happening. And I think, yeah, letting our imagination...

The world can be dreadful. And even in public government, this is why I think experimentation, having a little bit of fun and playfulness... We never get to play, but I think that what is worse than coming across an angry bureaucrat when you need to get something done, maybe if we played a little bit, if we felt that working in government was a little bit of a better thing, the people that we interact with will have a better experience.

So that's my advocacy for optimism and a little bit of playfulness. I think that it makes everything go a little bit better.

Joe Supervielle:

Hemant, do you have a quick answer on what your favorite is?

Hemant Desai:

Actually, so I'm absolutely optimistic as Santi is too. Obviously with the caveats of this risk. My favorite movie actually is dating way back to Short Circuit. "I'm alive," right? Johnny number 5 saying, "I'm alive." If you haven't watched that movie, that's a talk about going back to 1986 or 87, whenever that movie was released. And I think that would be my favorite because that's an optimistic view of how a robot could be helping us. Obviously the risk, as Santi said, bad people will do bad things, but I think good people can do awesome things with AI.

And in my opinion, good does outweigh the risk certainly in the long run, if you put the proper guardrails. So desire for humanity to just put some guardrails around AI will become the augmenting tool that, Joe, you mentioned. And I think so with that, I just end with saying I'm optimistic about what AI will do for all of us.

Joe Supervielle:

Santi, Hemant, thanks for your time today. Audience, remember to register early for a $100 off ICMA's Local Government Reimagined conferences, April 10th through 12 in Boston, and June 5 through 7 in Palm Desert, California. And also sign into the ICMA Learning Lab for more on AI technology and all things local government with training, certifications, and webinars.

 

Episode is sponsored by

Guest Information

Santiago Garces – CIO, Boston

Hemant Desai – CIO, ICMA

Episode Notes

Check out part one of the Immediate Impact of AI on Local Government. In part two, Santy and Hemant continue covering these topics:

• How should IT or HR departments answer when employees ask: "How will AI make my job easier instead of replacing me?"
• How residents can use AI to help themselves and create better interaction with local government.
• What is the next step for local governments and individual staff after experimenting?
• AI architecture.
• Creating internal guidelines.
• Fear and optimism from our favorite fictional AI.

Resources

City of Boston Interim Guidelines for Using GenerativeAI

ICMA's Local Government Reimagined Conferences:

Explore the Future of AI in Local Government

One topic. Two locations. ICMA is headed to Boston, Massachusetts, and Palm Desert, California, for the Local Government Reimagined Conferences. From cutting-edge applications to strategic insights, discover how AI is reshaping the landscape of local government. Register here.

Government AI Coalition

Join over 140 government agencies in creating standards for responsible AI procurement and governance for public agencies. Learn more.
 

Strategic Planning in the Age of AI: Free on-demand webinar with ICMA and strategic partner, Zencity, named a 2024 GovTech100 company.