Dell Technologies World 2024 | Jon Siegal & Rich Gagnon | City of Amarillo

Clips
News
More from Dell Technologies World 2024

Jon Siegal

SVP, Product Marketing

Dell Technologies

Richard Gagnon

Assistant City Manager & Chief Information Officer

City of Amarillo

Dell supports development of AI avatar for the City of Amarillo

Artificial intelligence democratization can sometimes feel very far away from the current reality, but in the City of Amarillo, Texas, AI will soon be impacting thousands of residents through a unique collaboration with Dell Technologies Inc.Dell has partnered with Rich Gagnon (pictured, left), chief information officer of the City of Amarillo, to create a website with a digital human, named Emma, who can help communicate with residents and translate requests, such as how to pay their water bill or what happened during the last city council

play_circle_outline Dell Coverage in Las Vegas: Exploring Savanah Peterson and Dave Vellante's Vegas Adventures at the Circle Bar

play_circle_outline Dell and Amarillo: A Transformational Partnership Utilizing Technology to Enhance City Residents' Lives

play_circle_outline 10. Use of Dell and NVIDIA infrastructure and software in the project.

play_circle_outline Enhancing Emma through Cultural Appropriateness and User Feedback for Dynamic System Improvement

play_circle_outline 14. Development of the project in the cloud with plans to migrate on-site.

Info

Jon Siegal & Rich Gagnon | City of Amarillo

Jon Siegal

SVP, Product Marketing Dell Technologies

Richard Gagnon

Assistant City Manager & Chief Information Officer City of Amarillo

search

>> Good morning Dell fans, and welcome back to Fabulous Las Vegas, Nevada. We're midway through day two of our three days of coverage here on theCube. My name is Savannah Peterson, joined by my co-host and laptop fumbler, Dave Vellante.

>> Hey.

>> Hey, how's it going? Steady hands after our martinis .

>> It's good. I don't know. Yeah, right. Late night.

>> Yeah. Hey, we were pretty behaved last night.

>> We were.

>> For Vegas in particular.

>> Although I went by the ... we used to call it the Circle Bar.

>> Oh, yeah. Yeah, yeah.

>> And of course, back in the day, EMC World, it was the Circle Bar, and of course, all the former EMC sales guys were there. So I was trying to sneak to my room, but I couldn't get by them. It was fun. Good little reunion. Yeah.

>> It's always worth having a big reunion. Speaking of reunions, we've got CUBE veterans on the show. John Siegal, thank you so much for being here.

>> Pleasure as always.

>> It's a joy. You've got your big smiles. You always bring us a fantastic customer story, and I'm very excited for CUBE first-timer, I believe.

>> Second.

>> Second time? Oh, repeat visitor. Sorry. First time with me then. We'll cut that in. Rich Gagnon, thank you so much for being here.

>> Happy to.

>> You have a very unique role. We don't always get to have government pros on the show with us. Tell us a little bit about your role at the city of Amarillo, just to get us started.

>> Sure. So I'm the assistant city manager and CIO responsible for all technology in our city.

>> Casual, just all the technology within a municipality. Some pretty exciting stuff. So John, talk to me. I know you were really excited to bring Rich on. What's going on here? What's the partnership?

>> Look, about a year ago actually, we started working together, Dell and the city of Amarillo. And Rich will share, but he has got a transformative vision for how to take technology and apply it to good. And so what you're going to see here is ... and I'll let him talk more about it, but it's really about making life better for the citizens of his city. And not only that, but he's also advocating to do the same, and we're now working with him to do the same and make it more repeatable for other cities around the country.

>> Right.

>> So how did you end up meeting each other? Did you reach out to Dell? Yeah, what does this look like if I'm a city and I want to partner with a power player like Dell?

>> Sure, sure. It started actually after our broadband project. We had about half of our city that was missing-

>> This makes sense.

>> Right. So we closed that gap, and in doing that, what I discovered honestly about my own city is we have more refugees per capita than any city in Texas. 24% of our residents don't speak English at home. We have one middle school speaking 62 languages and dialects in the same school.

>> Oh, my gosh.

>> I'm a nerd. I started with Hugging Face.

>> You're safe here. You're amongst nerds.

>> I started playing with models, and I was really impressed with the abilities to translate and thought, "Hey, there's something here." Reached out to Dell and said, "Hey, I have this crazy idea. What if we use generative AI on our website and made that an interface into services?" So instead of having to know English or typing, you could just have a conversation and get access to information and municipal services.

>> That is so important. We talk a lot, and we joke about it a little bit. There's a lot of buzz around democratizing AI and what that really means, and I think there's a lot of lip service in that space a little bit sometimes. But the real practical application is how does it democratize access to even resources and tools and information, which is exactly what you're doing. You have a demo for us, don't you?

>> I do. I could show you.

>> Would you mind pulling that up? We'd love to see it.

>> Not at all. So this is actually our website in development right now. It'll go live in October. And it is a digital human that resides on our website that allows residents to connect to the website. She'll come up and ask, "What services do you need access to?" We do it in text for those that are hearing challenged. And then you just ask a question, so, "How do I pay my water bill?" And then she will tell you how to do that and then navigate you to the site, to the place to pay your water bill.

>> And this is Emma.

>> This is Emma.

>> And how'd you come up with a name? CrowdStrike has Charlotte. IBM in the sixties had a thing called ELIZA, which was the original AI chatbot. How'd you come up with Emma?

>> And then you've got Watson. Yeah.

>> To be transparent, it was the result of a dad joke. We were looking at the names and Emma was one of them. And I said, "Well, if she's Emma, her last name has to be Rillo." And everybody laughed. We were like, "Okay, that's a thing." It's Emma Rillo.

>> .

>> Wow. That is just-

>> Very good. I have two daughters.

>> That was great. How old are your daughters?

>> 30. I have three grandchildren now. So they're grown.

>> Awesome. What do your grandchildren or your children think of Emma?

>> They adore her. Yes.

>> What a lovely thing to say about an AI.

>> They actually do. And it was really funny, when we first started testing her, my head of innovation, Kristen showed her kids and her daughter said, "That's the sixth grade teacher everybody wants to have." And I knew we were into gold. That's the response that we wanted.

>> It's interesting, your reaction. I mentioned ELIZA.

>> Yeah.

>> ELIZA in the sixties ... I think it was late sixties, they would put ELIZA in front of folks that they felt had depression because ... Yeah, and so-

>> Companionship. Yeah.

>> Yes. And so you think about AI, it's much better than sitting on a phone and a call center, right?

>> Actually, that's something that we're having to deal with. So when we did our initial testing, what we discovered with our senior population is exactly that. Our initial testers, they won't stop talking to her. So I came back, and two and a half hours later, they're talking about their grandchildren. So instead of fighting that, we're trying to embrace that and work with some of the nonprofits, and actually work with our seniors community so that we could have a community companion, if you will, that addresses issues.

>> It's a weird dynamic. When you use ChatGPT, do you say please and thank you? I do.

>> I do sometimes.

>> Me too.

>> Just in case the machines come after us.

>> I feel like you need to. Emma Lu is one of the versions, right? Emma Lu is another version, if you will, iteration of Emma for the senior community.

>> Yeah. So we're working with non-profits to figure out, "Okay, how do we fund a companion for our seniors and let the seniors define what issues do they want to talk about? What personality will she have?" And I think in the long run-

>> Oh gosh, that's so precious....

>> we're going to start to see that. We're going to see it with our homeless community. We'll see it in education. So we're really making a study, we call it Conversational City. We're doing a study around what does generative AI for good look like in a city, in a community, and in society at large? How do we use that to make impact?

>> I'm real glad we're having this conversation. This is a great one. When you're communicating with your ... We all live in a tech bubble. We're talking to fellow nerds all the time. We all have our great opinions and hopes and dreams for this new technological revolution. How are the citizens of the city of Amarillo embracing this? What's the response been like as you do some of this user feedback?

>> It's been really interesting. In the beginning ... Anytime you innovate, whether that's in a corporate setting, you have to create a safe space to innovate in.

>> .

>> When you're in government, that safe space has to extend out to your residents. So we really didn't hide from that. We hit that on the front end. We passed what we call our Digital Dignity Rights and Privacy agenda. So it's a framework by which we can ingest new technology and protect the rights and privacy of our residents. We're very transparent. So we wanted to address the fear on the front end. That's been incredibly helpful. And now that there's a level of comfort that, "Okay, the city's doing the right things, and I can see that." Now they've embraced the technology, and it's really fun actually to introduce her to different groups. Because until you see her, it's just a concept. But I have employees that tell me that she's their favorite coworker.

>> So the vision here is to bridge that communications gap-

>> That's it....

>> amongst the citizens. But this obviously wasn't eight years in development. Just very rapid time to development, time to market. What's the tech underneath this whole thing?

>> So first of all, this is a collaboration between Dell and NVIDIA and some of our third party partners as well, and part of our open ecosystem. In fact, this is a great example of the Dell AI factory with NVIDIA in action, really more than anything else. We've been talking a lot about the factory. And underneath it, if you will, is Dell infrastructure. So you can imagine whether it's our servers and our storage, but also working with NVIDIA around their AI enterprise software and their infrastructure as well. And frankly, a lot of this is also our services. This was really critical here in the early going, and frankly, still ongoing basis is helping the community figure out what are the right data sources, how to integrate those data sources into the model, picking the model, picking the use cases, implementing the platform, integrating it with some other third party software that we're using in addition to NVIDIA and Dell to create that AI avatar interface that's so unique and so conversational, and really I think what's changing the game here. So underneath the covers, to your point, is Dell and NVIDIA infrastructure as well as some software. But what we're really doing from a Dell perspective is providing what the AI factory is, it's that easy button.

>> That's what everybody wants these days.

>> That's what they want. But yeah, under the covers, look, there's a lot of complexity for sure. But we're hoping that we're providing that great simplifier that allows them to turn all their great data, train it, optimize it, and create these outcomes that Richard's talking about.

>> It's a RAG. Are you vectorizing the data?

>> Yes, exactly.

>> Exactly.

>> And the vector database is a well-known vector database or an open source, or is it an integrated?

>> It is. It's actually Prion that's providing the vector database services for us. We made the decision early. You've got to think about where your data sources are. Actually, all of her data comes from our website, which has been really interesting to me because it's forced us to relook at the way we interface with our residents, period. So it didn't take very long for departments to realize, "Hey, we don't do this very well." It doesn't facilitate a conversation. And if it doesn't facilitate a conversation for Emma, it's clearly not answering the questions that our residents really have. So we've gone department by department and revamped our entire website. You could just bolt a digital human on your existing website, it won't work. You'll have garbage conversations. So department by department, we've engaged with our residents and asked, "What do you want to know about water? What do you want to know about finance?" And we've defined those conversations, that builds the website. And now as we learn more, we just update the website that gets ingested into the vector database and we use RAG and it keeps us updated. If you notice on the demo, there's a thumbs up, thumbs down. We actually track those. We track every conversation. Now we're anonymized. I don't know who's having it. But I can see if it was good or bad. I can tell if Emma could or couldn't answer it. And now we have SLAs for our directors. So they see, "Okay, here's questions we need to get answered in the next 48 hours." So the conversation improves every day as we go forward.

>> Yeah, I'm sure. We saw this with our CUBE AI. We had way more thumbs downs, and you have to respond to them and you have to continue to iterate. Are you using multiple LLMs? Are you playing around with them? What are you finding there?

>> Well, website's just our first use case. I have a two-year road map. And that is one LLM, but we're building a new city hall. When you walk in, Emma will greet you in the front door and tell you where to go. That can be a much smaller language model. So the persona will be the same, but the use cases will define what LLM that we use.

>> So she's obviously a humanoid looking robot or digital human.

>> Right.

>> How did you decide on her form factor?

>> That was like nine weeks worth of research. We worked with UneeQ who actually built her. We had our internal call center staff actually involved, and then we had residents involved, and we went through iteration after iteration. Emma is the brand for the city, but our long-term vision is now with speech to text, you can actually recognize languages and dialects. We want to get to the point where when you interface with us, Emma can recognize, "Okay, that may be a Somalian man." And then introduce a different avatar that is culturally appropriate to that person. So ideally, when we're all said and done, no matter who you are, when you interface with our city, you'll speak to someone that looks and feels like you. It seems like a little thing-

>> It's not....

>> but it's a really big thing. It makes people feel like, "Okay, they really care about us."

>> This is running in your data center?

>> It's not right now. So we took an old-school approach. We're developing in the cloud. I did not want to wait.

>> Yeah, great. Okay.

>> This gave me a year of development while Dell got their infrastructure pieces ready, and now we'll migrate it on-site.

>> Okay, so why don't you just keep it in the cloud? Why are you going to migrate it back on-site?

>> Because of the number of use cases that I've already formed, and it's cheaper to move the AI to my data than my data to the AI.

>> And why is it cheaper? It's just what it cost to just run tokens in the cloud?

>> Yeah. Well, and the datasets. So we have some use cases where we're working on something now ... You think about a conversation that's two-way. Right now, a resident asks a question and Emma gives an answer, but we're making some progress on follow-up questions from Emma. So now we can ask a question. Emma could ask a question and residents could give an answer. So a simple example is we're giving updates for council for the first time this week from Emma. So she'll tell residents, "Here's what happened in council."

>> That's so cool. Yeah.

>> I can ask, "What did you think about that?" And now my residents give answers. We're using insight software that typically you see in commercial operations to gather that, now I have a sentiment engine for my entire city that's updating on every topic. So you can see where that would get expensive as datasets start to grow.

>> How'd you pay for this? Was it incremental budget?

>> Good question.

>> Did you have to steal from other budgets?

>> No, actually I've been plotting this for a few years. So we had some funds for redeveloping our website, and we sat on those and added to them, and I kept telling our communications department, "We're not spending it yet because this generative AI thing is going to be something." And then when we ran the numbers for expanding our call center, in order to meet the need, we would have to spend about 1.8 million a year in additional salary to have all the languages and translators, and we're doing it for a fraction of that. So not only did we ... We had the money through other projects, but it actually saves us in -

>> Yeah, so the value add, which is exciting.

>> Yeah, exactly.

>> John, I'm curious, are other cities like this reaching out to you?

>> Absolutely, absolutely. So as I mentioned at the outset, Rich has been advocating as well, and he's actually been speaking to other customers with us on our behalf. There's a tremendous amount of interest. So really what we announced yesterday, it was the Dell gen AI solution for digital assistants, which is actually being demoed right behind us right now. And it really is ... We've taken all the learnings that Rich just shared with you and we've simplified it and made it a repeatable solution so that other cities can actually streamline the process. What we did with Rich, as I said, started a year ago, strong partnership. He's a much a partner as a customer, really at the end of the day. But we've taken those learnings and now I think we've really streamlined it and it's now the ability to ... We still need the services component to help customers figure out the data sources and everything else. But we've packaged the solution in such a way where now it's going to be easy to deploy, including on-prem or at the edge, because based on what Rich described, there's a lot of advantage to that, whether that's privacy or security or cost. So we're making that now simpler than ever to deploy on-prem.

>> You've obviously built up an application portfolio over the years. As you get experience with technologies like this, how are you thinking about your portfolio? In other words, what areas of the portfolio do you feel like, "What I was doing there, I can now do differently for one-tenth the cost."

>> All of them.

>> Everything?

>> Everything. So we actually restructured, even. We've reorganized around the change that's coming. We all see it, this wave of change coming in. And I had two concerns. I was really concerned with our change management process and how we innovate. So we restructured. I moved communication, innovation, and our PMO into the same office. So now we are actually going through our entire application set. I'll give you a simple example. Illegal dumping was an issue that we really struggled with. We put that in the innovation department. They did an engagement with our residents, and we found out that we didn't never solved the problem because it's actually six different problems. Engaging with residents, they told us exactly what the problems were. We're doing the same thing with applications. Well, the outcome of that is we use machine learning, we've got dump sites, we've dropped our fuel costs by $480,000 this year, and that's trash. That is applying just a little bit of AI just to solid waste. And we're seeing that across every department. So I think fundamentally, we're changing the way that we look at how we do business.

>> Oh, your budget director must be happy.

>> Oh, very.

>> Well, and you're getting teams out of silos. I think that's a big thing that we hear a lot. And even within the partner ecosystem that Dell has, and everyone else. It used to be this Silicon Valley secret, whisper, whisper, whisper, "We're doing these things or whatever's happening, or this team is fighting for this budget." And I really do feel ... We've talked about a lot on the show that this water level's rising together as we all collaborate to design the future of our lives and our family's lives and everything else. All right, gentlemen, last question for you. Since you're both CUBE Veterans and I'm sure we'll have you back, what do you hope to be able to say on this stage with Dave and I next time you're here that you can't say today?

>> That we're having two-way conversations, and that Emma has been extended beyond the walls of the city and is helping our little Guatemalan children in Pampa learn English. So I think we're going to see her applied more broadly across our community.

>> Love that. We can't wait to celebrate that with you. What about you, John?

>> I hope we have another 100 examples of what Rich just went through.

>> Yes, yes.

>> And look, not just in city government and in smart cities as he's really building here, but one of the things we didn't talk about is the same exact capability, this digital human capability and interface is useful in several industries. We're seeing interest not just in the cities, which I mentioned earlier because Rich has been helping, but in retail such as online as well as physical stores, actually even more so. We're getting interest in healthcare to really help bridge the care teams with the patients, whether it's when they first visit or when they leave, or actually during their stay, and financial service is another one. So I also hope that in addition to the 100 of those, we have 100 of other industries examples as well. So this has taken off. And again, we are so grateful to Rich and his collaboration here, and we hope to help to continue to grow his community as well.

>> Well, we look forward to hearing as the story evolves, and we'll have all 100 of those city municipalities up on this stage.

>> That's right.

>> One big, cozy, little happy family on our next podcast.

>> There you go. Deal.

>> Yeah, sounds great. Dell Tech World 2025. Can't wait to see it.

>> All right.

>> Dave, thanks so much for another fabulous interview. Thank both of you for coming to hang out. This is a real fun one that I think applies to a lot of people. We're making AI real here on theCUBE, folks. And thank you for tuning in from home. We're midway through day two of Dell Tech World here in beautiful Las Vegas, Nevada. My name's Savannah Peterson. You're watching theCUBE, the leading source for enterprise tech news.

Jon Siegal & Rich Gagnon | City of Amarillo

search

>> Good morning Dell fans, and welcome back to Fabulous Las Vegas, Nevada. We're midway through day two of our three days of coverage here on theCube. My name is Savannah Peterson, joined by my co-host and laptop fumbler, Dave Vellante.

>> Hey.

>> Hey, how's it going? Steady hands after our martinis .

>> It's good. I don't know. Yeah, right. Late night.

>> Yeah. Hey, we were pretty behaved last night.

>> We were.

>> For Vegas in particular.

>> Although I went by the ... we used to call it the Circle Bar.

>> Oh, yeah. Yeah, yeah.

>> And of course, back in the day, EMC World, it was the Circle Bar, and of course, all the former EMC sales guys were there. So I was trying to sneak to my room, but I couldn't get by them. It was fun. Good little reunion. Yeah.

>> It's always worth having a big reunion. Speaking of reunions, we've got CUBE veterans on the show. John Siegal, thank you so much for being here.

>> Pleasure as always.

>> It's a joy. You've got your big smiles. You always bring us a fantastic customer story, and I'm very excited for CUBE first-timer, I believe.

>> Second.

>> Second time? Oh, repeat visitor. Sorry. First time with me then. We'll cut that in. Rich Gagnon, thank you so much for being here.

>> Happy to.

>> You have a very unique role. We don't always get to have government pros on the show with us. Tell us a little bit about your role at the city of Amarillo, just to get us started.

>> Sure. So I'm the assistant city manager and CIO responsible for all technology in our city.

>> Casual, just all the technology within a municipality. Some pretty exciting stuff. So John, talk to me. I know you were really excited to bring Rich on. What's going on here? What's the partnership?

>> Look, about a year ago actually, we started working together, Dell and the city of Amarillo. And Rich will share, but he has got a transformative vision for how to take technology and apply it to good. And so what you're going to see here is ... and I'll let him talk more about it, but it's really about making life better for the citizens of his city. And not only that, but he's also advocating to do the same, and we're now working with him to do the same and make it more repeatable for other cities around the country.

>> Right.

>> So how did you end up meeting each other? Did you reach out to Dell? Yeah, what does this look like if I'm a city and I want to partner with a power player like Dell?

>> Sure, sure. It started actually after our broadband project. We had about half of our city that was missing-

>> This makes sense.

>> Right. So we closed that gap, and in doing that, what I discovered honestly about my own city is we have more refugees per capita than any city in Texas. 24% of our residents don't speak English at home. We have one middle school speaking 62 languages and dialects in the same school.

>> Oh, my gosh.

>> I'm a nerd. I started with Hugging Face.

>> You're safe here. You're amongst nerds.

>> I started playing with models, and I was really impressed with the abilities to translate and thought, "Hey, there's something here." Reached out to Dell and said, "Hey, I have this crazy idea. What if we use generative AI on our website and made that an interface into services?" So instead of having to know English or typing, you could just have a conversation and get access to information and municipal services.

>> That is so important. We talk a lot, and we joke about it a little bit. There's a lot of buzz around democratizing AI and what that really means, and I think there's a lot of lip service in that space a little bit sometimes. But the real practical application is how does it democratize access to even resources and tools and information, which is exactly what you're doing. You have a demo for us, don't you?

>> I do. I could show you.

>> Would you mind pulling that up? We'd love to see it.

>> Not at all. So this is actually our website in development right now. It'll go live in October. And it is a digital human that resides on our website that allows residents to connect to the website. She'll come up and ask, "What services do you need access to?" We do it in text for those that are hearing challenged. And then you just ask a question, so, "How do I pay my water bill?" And then she will tell you how to do that and then navigate you to the site, to the place to pay your water bill.

>> And this is Emma.

>> This is Emma.

>> And how'd you come up with a name? CrowdStrike has Charlotte. IBM in the sixties had a thing called ELIZA, which was the original AI chatbot. How'd you come up with Emma?

>> And then you've got Watson. Yeah.

>> To be transparent, it was the result of a dad joke. We were looking at the names and Emma was one of them. And I said, "Well, if she's Emma, her last name has to be Rillo." And everybody laughed. We were like, "Okay, that's a thing." It's Emma Rillo.

>> .

>> Wow. That is just-

>> Very good. I have two daughters.

>> That was great. How old are your daughters?

>> 30. I have three grandchildren now. So they're grown.

>> Awesome. What do your grandchildren or your children think of Emma?

>> They adore her. Yes.

>> What a lovely thing to say about an AI.

>> They actually do. And it was really funny, when we first started testing her, my head of innovation, Kristen showed her kids and her daughter said, "That's the sixth grade teacher everybody wants to have." And I knew we were into gold. That's the response that we wanted.

>> It's interesting, your reaction. I mentioned ELIZA.

>> Yeah.

>> ELIZA in the sixties ... I think it was late sixties, they would put ELIZA in front of folks that they felt had depression because ... Yeah, and so-

>> Companionship. Yeah.

>> Yes. And so you think about AI, it's much better than sitting on a phone and a call center, right?

>> Actually, that's something that we're having to deal with. So when we did our initial testing, what we discovered with our senior population is exactly that. Our initial testers, they won't stop talking to her. So I came back, and two and a half hours later, they're talking about their grandchildren. So instead of fighting that, we're trying to embrace that and work with some of the nonprofits, and actually work with our seniors community so that we could have a community companion, if you will, that addresses issues.

>> It's a weird dynamic. When you use ChatGPT, do you say please and thank you? I do.

>> I do sometimes.

>> Me too.

>> Just in case the machines come after us.

>> I feel like you need to. Emma Lu is one of the versions, right? Emma Lu is another version, if you will, iteration of Emma for the senior community.

>> Yeah. So we're working with non-profits to figure out, "Okay, how do we fund a companion for our seniors and let the seniors define what issues do they want to talk about? What personality will she have?" And I think in the long run-

>> Oh gosh, that's so precious....

>> we're going to start to see that. We're going to see it with our homeless community. We'll see it in education. So we're really making a study, we call it Conversational City. We're doing a study around what does generative AI for good look like in a city, in a community, and in society at large? How do we use that to make impact?

>> I'm real glad we're having this conversation. This is a great one. When you're communicating with your ... We all live in a tech bubble. We're talking to fellow nerds all the time. We all have our great opinions and hopes and dreams for this new technological revolution. How are the citizens of the city of Amarillo embracing this? What's the response been like as you do some of this user feedback?

>> It's been really interesting. In the beginning ... Anytime you innovate, whether that's in a corporate setting, you have to create a safe space to innovate in.

>> .

>> When you're in government, that safe space has to extend out to your residents. So we really didn't hide from that. We hit that on the front end. We passed what we call our Digital Dignity Rights and Privacy agenda. So it's a framework by which we can ingest new technology and protect the rights and privacy of our residents. We're very transparent. So we wanted to address the fear on the front end. That's been incredibly helpful. And now that there's a level of comfort that, "Okay, the city's doing the right things, and I can see that." Now they've embraced the technology, and it's really fun actually to introduce her to different groups. Because until you see her, it's just a concept. But I have employees that tell me that she's their favorite coworker.

>> So the vision here is to bridge that communications gap-

>> That's it....

>> amongst the citizens. But this obviously wasn't eight years in development. Just very rapid time to development, time to market. What's the tech underneath this whole thing?

>> So first of all, this is a collaboration between Dell and NVIDIA and some of our third party partners as well, and part of our open ecosystem. In fact, this is a great example of the Dell AI factory with NVIDIA in action, really more than anything else. We've been talking a lot about the factory. And underneath it, if you will, is Dell infrastructure. So you can imagine whether it's our servers and our storage, but also working with NVIDIA around their AI enterprise software and their infrastructure as well. And frankly, a lot of this is also our services. This was really critical here in the early going, and frankly, still ongoing basis is helping the community figure out what are the right data sources, how to integrate those data sources into the model, picking the model, picking the use cases, implementing the platform, integrating it with some other third party software that we're using in addition to NVIDIA and Dell to create that AI avatar interface that's so unique and so conversational, and really I think what's changing the game here. So underneath the covers, to your point, is Dell and NVIDIA infrastructure as well as some software. But what we're really doing from a Dell perspective is providing what the AI factory is, it's that easy button.

>> That's what everybody wants these days.

>> That's what they want. But yeah, under the covers, look, there's a lot of complexity for sure. But we're hoping that we're providing that great simplifier that allows them to turn all their great data, train it, optimize it, and create these outcomes that Richard's talking about.

>> It's a RAG. Are you vectorizing the data?

>> Yes, exactly.

>> Exactly.

>> And the vector database is a well-known vector database or an open source, or is it an integrated?

>> It is. It's actually Prion that's providing the vector database services for us. We made the decision early. You've got to think about where your data sources are. Actually, all of her data comes from our website, which has been really interesting to me because it's forced us to relook at the way we interface with our residents, period. So it didn't take very long for departments to realize, "Hey, we don't do this very well." It doesn't facilitate a conversation. And if it doesn't facilitate a conversation for Emma, it's clearly not answering the questions that our residents really have. So we've gone department by department and revamped our entire website. You could just bolt a digital human on your existing website, it won't work. You'll have garbage conversations. So department by department, we've engaged with our residents and asked, "What do you want to know about water? What do you want to know about finance?" And we've defined those conversations, that builds the website. And now as we learn more, we just update the website that gets ingested into the vector database and we use RAG and it keeps us updated. If you notice on the demo, there's a thumbs up, thumbs down. We actually track those. We track every conversation. Now we're anonymized. I don't know who's having it. But I can see if it was good or bad. I can tell if Emma could or couldn't answer it. And now we have SLAs for our directors. So they see, "Okay, here's questions we need to get answered in the next 48 hours." So the conversation improves every day as we go forward.

>> Yeah, I'm sure. We saw this with our CUBE AI. We had way more thumbs downs, and you have to respond to them and you have to continue to iterate. Are you using multiple LLMs? Are you playing around with them? What are you finding there?

>> Well, website's just our first use case. I have a two-year road map. And that is one LLM, but we're building a new city hall. When you walk in, Emma will greet you in the front door and tell you where to go. That can be a much smaller language model. So the persona will be the same, but the use cases will define what LLM that we use.

>> So she's obviously a humanoid looking robot or digital human.

>> Right.

>> How did you decide on her form factor?

>> That was like nine weeks worth of research. We worked with UneeQ who actually built her. We had our internal call center staff actually involved, and then we had residents involved, and we went through iteration after iteration. Emma is the brand for the city, but our long-term vision is now with speech to text, you can actually recognize languages and dialects. We want to get to the point where when you interface with us, Emma can recognize, "Okay, that may be a Somalian man." And then introduce a different avatar that is culturally appropriate to that person. So ideally, when we're all said and done, no matter who you are, when you interface with our city, you'll speak to someone that looks and feels like you. It seems like a little thing-

>> It's not....

>> but it's a really big thing. It makes people feel like, "Okay, they really care about us."

>> This is running in your data center?

>> It's not right now. So we took an old-school approach. We're developing in the cloud. I did not want to wait.

>> Yeah, great. Okay.

>> This gave me a year of development while Dell got their infrastructure pieces ready, and now we'll migrate it on-site.

>> Okay, so why don't you just keep it in the cloud? Why are you going to migrate it back on-site?

>> Because of the number of use cases that I've already formed, and it's cheaper to move the AI to my data than my data to the AI.

>> And why is it cheaper? It's just what it cost to just run tokens in the cloud?

>> Yeah. Well, and the datasets. So we have some use cases where we're working on something now ... You think about a conversation that's two-way. Right now, a resident asks a question and Emma gives an answer, but we're making some progress on follow-up questions from Emma. So now we can ask a question. Emma could ask a question and residents could give an answer. So a simple example is we're giving updates for council for the first time this week from Emma. So she'll tell residents, "Here's what happened in council."

>> That's so cool. Yeah.

>> I can ask, "What did you think about that?" And now my residents give answers. We're using insight software that typically you see in commercial operations to gather that, now I have a sentiment engine for my entire city that's updating on every topic. So you can see where that would get expensive as datasets start to grow.

>> How'd you pay for this? Was it incremental budget?

>> Good question.

>> Did you have to steal from other budgets?

>> No, actually I've been plotting this for a few years. So we had some funds for redeveloping our website, and we sat on those and added to them, and I kept telling our communications department, "We're not spending it yet because this generative AI thing is going to be something." And then when we ran the numbers for expanding our call center, in order to meet the need, we would have to spend about 1.8 million a year in additional salary to have all the languages and translators, and we're doing it for a fraction of that. So not only did we ... We had the money through other projects, but it actually saves us in -

>> Yeah, so the value add, which is exciting.

>> Yeah, exactly.

>> John, I'm curious, are other cities like this reaching out to you?

>> Absolutely, absolutely. So as I mentioned at the outset, Rich has been advocating as well, and he's actually been speaking to other customers with us on our behalf. There's a tremendous amount of interest. So really what we announced yesterday, it was the Dell gen AI solution for digital assistants, which is actually being demoed right behind us right now. And it really is ... We've taken all the learnings that Rich just shared with you and we've simplified it and made it a repeatable solution so that other cities can actually streamline the process. What we did with Rich, as I said, started a year ago, strong partnership. He's a much a partner as a customer, really at the end of the day. But we've taken those learnings and now I think we've really streamlined it and it's now the ability to ... We still need the services component to help customers figure out the data sources and everything else. But we've packaged the solution in such a way where now it's going to be easy to deploy, including on-prem or at the edge, because based on what Rich described, there's a lot of advantage to that, whether that's privacy or security or cost. So we're making that now simpler than ever to deploy on-prem.

>> You've obviously built up an application portfolio over the years. As you get experience with technologies like this, how are you thinking about your portfolio? In other words, what areas of the portfolio do you feel like, "What I was doing there, I can now do differently for one-tenth the cost."

>> All of them.

>> Everything?

>> Everything. So we actually restructured, even. We've reorganized around the change that's coming. We all see it, this wave of change coming in. And I had two concerns. I was really concerned with our change management process and how we innovate. So we restructured. I moved communication, innovation, and our PMO into the same office. So now we are actually going through our entire application set. I'll give you a simple example. Illegal dumping was an issue that we really struggled with. We put that in the innovation department. They did an engagement with our residents, and we found out that we didn't never solved the problem because it's actually six different problems. Engaging with residents, they told us exactly what the problems were. We're doing the same thing with applications. Well, the outcome of that is we use machine learning, we've got dump sites, we've dropped our fuel costs by $480,000 this year, and that's trash. That is applying just a little bit of AI just to solid waste. And we're seeing that across every department. So I think fundamentally, we're changing the way that we look at how we do business.

>> Oh, your budget director must be happy.

>> Oh, very.

>> Well, and you're getting teams out of silos. I think that's a big thing that we hear a lot. And even within the partner ecosystem that Dell has, and everyone else. It used to be this Silicon Valley secret, whisper, whisper, whisper, "We're doing these things or whatever's happening, or this team is fighting for this budget." And I really do feel ... We've talked about a lot on the show that this water level's rising together as we all collaborate to design the future of our lives and our family's lives and everything else. All right, gentlemen, last question for you. Since you're both CUBE Veterans and I'm sure we'll have you back, what do you hope to be able to say on this stage with Dave and I next time you're here that you can't say today?

>> That we're having two-way conversations, and that Emma has been extended beyond the walls of the city and is helping our little Guatemalan children in Pampa learn English. So I think we're going to see her applied more broadly across our community.

>> Love that. We can't wait to celebrate that with you. What about you, John?

>> I hope we have another 100 examples of what Rich just went through.

>> Yes, yes.

>> And look, not just in city government and in smart cities as he's really building here, but one of the things we didn't talk about is the same exact capability, this digital human capability and interface is useful in several industries. We're seeing interest not just in the cities, which I mentioned earlier because Rich has been helping, but in retail such as online as well as physical stores, actually even more so. We're getting interest in healthcare to really help bridge the care teams with the patients, whether it's when they first visit or when they leave, or actually during their stay, and financial service is another one. So I also hope that in addition to the 100 of those, we have 100 of other industries examples as well. So this has taken off. And again, we are so grateful to Rich and his collaboration here, and we hope to help to continue to grow his community as well.

>> Well, we look forward to hearing as the story evolves, and we'll have all 100 of those city municipalities up on this stage.

>> That's right.

>> One big, cozy, little happy family on our next podcast.

>> There you go. Deal.

>> Yeah, sounds great. Dell Tech World 2025. Can't wait to see it.

>> All right.

>> Dave, thanks so much for another fabulous interview. Thank both of you for coming to hang out. This is a real fun one that I think applies to a lot of people. We're making AI real here on theCUBE, folks. And thank you for tuning in from home. We're midway through day two of Dell Tech World here in beautiful Las Vegas, Nevada. My name's Savannah Peterson. You're watching theCUBE, the leading source for enterprise tech news.