On the first Wednesday of every month, Strategic Government Resources and the Alliance for Innovation present a highly useful class on using generative AI in different city management fields. The instructor for these sessions is Micah Gaudet, deputy city manager of the City of Maricopa, Arizona. In May, the topic was public safety. Back in February, we wrote about the class he held that month on AI and Human Resource management.

Gaudet has been learning about multiple practical uses for generative AI since late 2022. (We wrote a bit about his voyage of discovery in this April 2024 Government Finance Review article.) 

The purpose of these sessions: “To democratize AI and make it accessible for all,” he told us, pointing especially to its use in smaller cities and rural communities,” which are often understaffed and where “it’s like adding new staff members” to a team.

Of course, it’s important for communities to recognize risks. Based on his experience and conversations with government AI users all over the world, Gaudet counsels that users not put more faith in the information spewed out of a computer screen than they would from an employee who is assisting in research. His class demonstrates the advantages in AI’s quick ability to search for, and come up with ideas and approaches, but also the need to tailor general responses to individual city needs and to recognize limitations. “AI has never done a fire call or an EMS call,” he says.

The May generative AI for public safety session focuses on ChatGPT, both the unpaid and paid versions. Gaudet’s discussions of AI’s capacity are enlivened with examples of its capacity. Early in the class, for example, he enters a query about developing a testing process for sergeant positions, with an emphasis on the development of relational leadership skills and the soft skills needed on a police force. To the basic answers that AI delivers, he follows up with additional questions, such as “how to develop training programs,” and how to use a similar process in other areas such as fire service or code enforcement.

While many of the AI-generated answers are useful, in some cases they may seem obvious to experienced public safety officials. To dig farther, Gaudet shows how managers can spark other ideas and insights. During the webinar, he also demonstrates another AI platform, Claude, which makes it easier to do visualizations. He also uses Claude to show how generative AI can not only produce answers, but then provide the potential negative impacts of following its own advice.

We found the most intriguing part of the May session was Gaudet’s foray into using data sets that can be uploaded in ChatGPT-4. He shows, for example, that with the appropriate uploaded data, the $20-per-month version can match nuisance complaints with fire incident response data and the location of rental properties, yielding potential relationships between the different kinds of data a city has. 

The short time that generative AI takes to come up with answers allows Gaudet the chance to take a very small sip of coffee as the answers emerge. He also shows how persistence pays off when ChatGPT doesn’t deliver – for example in its unsuccessful first effort to generate an interactive map and how asking and re-asking questions in different ways can pay off.

#StateandLocalDataManagement #CityGenerativeAITraining #PublicSafetyGenerativeAITraining #GenerativeAI #GenerativeAIforPublicSafety #StragicGovernmentResources #AllianceForInnovation #SGR #SGRWednesdayAIWebinar #PublicSafetyandChatGPT #MicahGaudet #GenerativeAIandRuralGovernment #GeneratativeAIandSmallCities #GovernmentFinanceReview #CityofMaricopa