Generative AI and Local Governments.
When he first heard news reports about ChatGPT and its powers to search the internet and create readable text, rhymes, and fanciful tales, Micah Gaudet immediately saw the use for entertaining his eight-year-old son with "corny dad jokes and little bedtime stories."But within weeks, as deputy manager of the fast-growing City of Maricopa, Arizona, he started imagining ways to use this newly available form of artificial intelligence (AI) in his work as deputy city manager and chief public safety officer. Initially, he tapped its power for dealing with email responses, short memos, and editing, and then began talking with his staff, and city officials both in his own city and elsewhere, about its potential.
He began holding webinars to teach and share information and by November 2023, he had started a network of AI and generative AI enthusiasts to share the ways they're using these powerful technological tools in their own local governments. "I wanted to know how people are using this because it is the future," he says.
The speed of change
Across the United States and the world, the previously unfamiliar term "generative AI" has opened up a world of opportunities for government. "It's disruptive to local government in ways that we haven't seen before because of the pace at which change is coming," said Ron Holifield, chief executive officer of Strategic Government Resources (SGR) and interim executive director of the Alliance for Innovation.
When Gaudet's jump into generative AI caught Holifield's attention, Strategic Government Resources ran a webinar featuring an interview with him--and 700 local government officials signed up.
A tool to combat the workforce shortage
One of the key factors generating interest in generative AI is the ongoing public-sector workforce shortage, which is beginning to look less like a temporary phenomenon and more like an ongoing issue, based on demographics, continued competition with the private sector, and changing attitudes toward long-term employee-employer relationships.
According to Ryan Oakes, global health and public services industry practice chair at Accenture, the issue with recruiting and retaining a full staff has created a need to "generate the kind of productivity increases that will allow us to compensate for workforce challenges across the country."
The potential is particularly profound when governments need to collect and impart information. Combined with other aspects of artificial intelligence, fertile spots for generative AI lie in routine, repetitive, and rule-based functions such as claims processing, eligibility automation, and administrative authorizations. But humans are still needed for overseeing the way the rules are developed and implemented.
Exploring potential uses
In late 2023 and early 2024, a great deal of exploration was taking place on ways to use generative AI. And at the same time, states and local governments were busy developing policies and guidelines to ensure that productivity gains would not compromise accuracy or introduce bias through AI algorithms.
As those policies are being developed, Gaudet emphasizes the importance of having a "human in the loop" perspective. "Don't just take what ChatGPT writes, copy it, paste it, and call it good. You have to review this in the same way you would normally review anything. Treat it like an intern who is helping to augment your work," he says. With a background in military intelligence, Gaudet was brought up in a world that emphasized the importance of verifying information.
In one-on-one meetings with department heads, Gaudet has communicated the potential of AI without pushing its use on those who are not ready. The city is exploring uses in a variety of ways. For example, the police department is exploring how generative AI can analyze and generate reports on body-worn camera footage and provide material for training officers on language use.
Speed is a critical benefit. Last summer, the City of Maricopa, Arizona, celebrated the 20th anniversary of its incorporation with a big anniversary party at a city park. Gaudet relied on police and fire officials to develop an incident and traffic management plan, which they worked on over several months.
As a test of Maricopa's premium version of ChatGPT, Gaudet uploaded a screenshot of a Google map, explained the event and its planned 10,000 attendees, and asked for its incident and traffic management analysis. "The response that Chat GPT gave was almost the exact same one that our police and fire management team put together," Gaudet says. "I think it took ChatGPT all of two minutes."
Controlled searches
One way in which a small number of governments are using generative AI is to provide a new kind of controlled search that is limited to their own websites, rather than the broader internet, where the volume of information and misinformation can provide confusing and potentially inaccurate results.
One of the first to try this was the City of Grand Rapids, Michigan. As of early February, its website still offered two search options. The traditional option is flagged by the familiar tiny magnifying glass. The other option gives visitors the chance to try out a new AI search tool, "powered by Polimorphic," a software company that specializes in customer service.
We asked both the traditional search tool and the generative AI one the same question: "Do I need a permit for opening a new coffee shop in Grand Rapids?"
The traditional search brought up 2,376 results. The first reference led to an article about a venture fund aimed at boosting businesses owned by people of color in West Michigan.
The AI search response was simple and to the point. While it said that Grand Rapids website sources could not find specific information on coffee shop permitting, it suggested contacting the Grand Rapids Business Licensing Office for more information on permits and licenses and provided both a phone number and website link.
"The key part is that you're basically limiting the part of the internet that you're looking at," said Parth Shah, chief executive officer at Polimorphic, which started providing "Ask AI" technology to cities and counties in October 2023.
Andy Pederson is village manager in the Village of Bayside, Wisconsin, a small community just a few miles from Milwaukee. He began using "Ask AI" in early November 2023. "It's more user friendly for everybody. It not only searches our website and summarizes information, but it then provides links to the resources that people need to use," he said, adding that it has also been useful for training new staff.
One challenge--and benefit--of "Ask AI" is that it reveals outdated or missing material. For example, in the early days of its use in Bayside, officials noticed that questions about community policing provided the name and email of a former police chief, necessitating an update.
"When you start to use AI, you start to see where it's pulling from content that might be out of date," said Shah, who is also working on a voice version so callers can also have access to a generative AI search. This developing technology also dramatically affects the quality of responses to citizen service requests, reducing the need for employees to direct telephone, email, text, or chatbot traffic to the right place.
The need for caution
Pawel Maslag, a member of the city council in the City of Garfield, New Jersey, became aware of the potential for generative AI while working on a master's program at Yale University, following the release and major publicity surrounding the release of ChatGPT in late 2022. But Maslag is also highly aware of the need for caution. "I think there's a lot of places rushing into services without fully understanding the implications," he said.
In an opinion piece for North Jersey. com, a digital news service, in summer 2023, Maslag singled out three policy steps that he considers especially critical:
1. Mandatory training for staff.
2. Transparency ordinances so its use is "transparent and understandable."
3. Impact assessments, so that when AI or generative AI are used, there is an analysis of how it affects different populations, whether there are security vulnerabilities, and whether potential biases might affect results.
"As technology becomes more intertwined with our daily routines, our dependency on it increases. With dependency comes vulnerability," he said.
A couple of the basic ideas: 1) Don't put personally identifiable information in any generative AI tool that you're using, and 2) Make sure that you're vetting the results for accuracy.
Reynolds also emphasized the importance of a broad approach to staff training to understand the rules and the risks that emerge as government processes change and more data and information is more widely available. As she points out, generative AI brings with it a drastic change in who can have access to data.
"Artificial intelligence has brought data, and the ability to analyze data to the masses--not just to the individuals with data analytics expertise," she noted.
Katherine Barrett and Richard Greene are principals of Barrett and Greene, Inc.
 Printer friendly
 Cite/link
 Email
 Feedback
| |
| Author: | Barrett, Katherine; Greene, Richard |
|---|---|
| Publication: | Government Finance Review |
| Article Type: | Viewpoint essay |
| Geographic Code: | 1USA |
| Date: | Apr 1, 2024 |
| Words: | 1471 |
| Previous Article: | GASB 102 and More Disclosure for You. |
| Next Article: | Will the Real Municipal ESG Please Stand Up? |
| Topics: | |