Can AI be helpful for your city?
Most of us with a smartphone are using artificial intelligence every day whether or not we think we are. Are you asking Siri or Google to dial a phone number for you or give you directions? It’s AI. AI is becoming more prevalent and, frankly, more sophisticated than ever before, and more municipalities are using AI and generative AI in multiple ways.
Everything from adding a chatbot onto a city’s website — like Colorado Springs, Colo.’s, ASKCOS, which provides information on city services — to sorting trash and recyclables using AI technology like the city of Portsmouth, Va. But more commonly, municipalities are using generative AI — or artificial intelligence that is capable of creating text, images or videos often in response to prompts.
According to a June 4, 2024, article in The World Economic Forum — “Cities are Sizing up the Generative AI Skyline but First They Have to Establish the Ground Rules” — 96 out of 100 mayors and city staff have expressed interest in exploring the possible applications of GenAI rather than actively implementing it. While 96% of mayors are interested in using GenAI, many have not developed comprehensive policies and only 2% of cities are actually adopting the technology.
That article also claimed that the GenAI market is projected to expand to $255.8 billion by 2033, compared to $13.5 billion in 2023.
The Municipalspoke to representatives in two cities that are using AI, including Murray City, Utah.
Doug Hill, chief administrative officer for the city of Murray, a suburb of Salt Lake City, said, “Our city recorder, Brooke Smith, has been our most interested person in learning about AI.”
He said she’s attended a lot of conferences and brought information back to the city. “She’s our most proficient user of AI,” he said, adding that she generally uses ChatGPT for RFPs (requests for proposals) for procurement, and she also does meeting minutes using voice recognition software that’s then fed into an AI program and writes the minutes. She then reviews them and makes sure they are accurate.
Hill said Smith has done a lot of research on different programs.
“We created an AI committee with representatives from most city departments. We meet monthly,” he said. Since the committee was just formed a few months ago, it has only met a few times so far to discuss how members are using AI.
“We decided to develop a policy on how to use it and talk about what type of software to use if not on an approved list.”
There is software out there that Murray is not using, and Hill gave a couple of examples, which include software for police departments that write reports using body cam footage and software for the water department that monitors water conservation efforts. He said, unlike some cities he’s aware of, the city of Murray is currently only using AI that has been approved by Homeland Security.
As for AI’s biggest benefits, Hill said, “It’s a tremendous timesaver. We’re generally using it for minutes, procurement policies, articles, speeches, simple things like emails, press releases — things that would take several hours to produce before, Grammarly can do quickly.”
Hill gave an example of someone in the parks department coming to him because he needed specs for landscaping, and he wondered if there was a template document for that. Murray might have one, but Hill suggested he go to ChatGPT and type in landscaping specs and tell it specifically whether he wanted landscaping for a pool, park, tennis court, etc.
“He came back and showed me what it produced, and what would have taken hours, it took 15 minutes by AI — so the greatest benefit so far is time savings,” Hill said.
But, Hill admitted, “Right now, we have more questions than answers” with some AI, but he hopes to use it to make their jobs easier.
Concerns about AI include privacy, accuracy and giving proper credit. Hill learned about the term “hallucinations” regarding AI — also known by some more colorful terms — which is when AI presents inaccurate information as fact. One study analyst said chatbots hallucinate as much as 27% of the time with factual errors present in 46% of generative AI.
So, Hill said staff members have to do a lot of fact checking. There are also concerns about information becoming public, since right now AI is an open infrastructure. One error they keep finding is “right now, they can’t distinguish between Murray City, Utah, and Murray, Ky.”
Even when Smith uses the voice recognition to take minutes, Hill said, “It’s not entirely accurate.” He added, “Most people don’t want to read word-for-word minutes, so she asks for a summary — but we have to constantly fact check.”
He shared he was going to use generative AI for the mayor’s state of the city address, but “at the end of the day, it didn’t sound like the mayor, so we didn’t use it.”
Murray City also wants to be transparent about the use of generative AI. Hill said, “If someone is using ChatGPT, it says so somewhere as well as that it’s been edited and fact-checked by staff.”
San Jose, Calif.
Albert Gehemi, chief privacy officer for the city of San Jose, shared, “The way the city of San Jose thinks about AI is how can it help provide services more efficiently and help our services be more accessible.”
As for efficiency, staff are using GenAI to draft content “by editing and reviewing rather than writing from scratch.”
Gehemi said staff uses it for anything from purchasing documents, memos, emails and more. “They figured they’re saving on average an hour per task and for heavy users up to 8 hours drafting larger documents — policies or solicitations. It helps us provide services we need to do faster and more efficiently.”
As for accessibility, San Jose has a “fantastic tool for language access for residents who can’t come to city hall or who don’t know how to communicate with us. We rolled out AI that basically translates city council minutes into 50 different languages,” he said. “Our city clerk sings its praises — it allows us to give access to so many residents in a way that wasn’t feasible or affordable before.”
The city is piloting a “computer vision” technology to detect road conditions. Gehemi said, “It’s still in testing, but we’re exploring how capable AI systems are able to detect issues in the street like potholes, etc. instead of relying completely on residents reporting.”
San Jose is “very excited” about an upskilling program and it’s working with six to seven departments to demonstrate what AI is and how to use it in their specific departments. “We’re going back and forth on how to better integrate it with our systems,” he said.
The main two concerns about AI are the same as Murray, Utah: data privacy and misinformation. Gehemi said, especially with “new types of AI like chatbots,” city leadership wants staff to know that information being put into the system “can be exposed in a new way that was never an issue before.”
He said AI is really useful in finding things the city thought were hidden, but it wasn’t. So, staff is keeping in mind what’s going on public websites and are asking questions like who has access. The city is setting good guidance and keeping a tight grip on how staff and residents are navigating privacy concerns and it has developed AI guidelines.
Misinformation and deep fakes are a concern, so city leadership is educating staff on how to verify before acting. A part of that education is how to spot a deep fake and check AI for accuracy. He said the guidelines and policies the city has come from the GovAI coalition.
Because the AI space is moving so quickly, San Jose has more detail in the guidelines versus policy because it is always evolving. Gehemi said the city has already updated the guidelines once.
For the public, he said San Jose is focusing on education and working, especially with small businesses, on how to use AI tools safely and responsibly. There is also a focus on understanding and spotting deep fakes.
He said the city works with neighborhood associations on how to use AI, especially if they don’t speak the same language. The city helps small businesses use AI tools to streamline their business, especially if English is not their native language. This help highlights how AI can help them fill out paperwork, draft emails and run their businesses better.
The biggest advice Gehemi offered was to “join the GovAI coalition — it’s incredibly valuable. Any agency getting started with AI should check out the GovAI coalition — there are over 1,000 practitioners, 350 state, local and federal agencies working on how to use AI effectively and responsibly.”
He said the city of San Jose’s policies have come from there and that’s where he heard about the road detection software.
“We learn from each other and pool resources. It makes adopting AI easier and more responsible — and (offers) more (varied learning) than we could learn by doing it on our own, because there are so many different perspectives in the room.”
Membership in GovAI Coalition is free, and interested agencies can just fill out the GovAI membership form. There’s also going to be a GovAI Coalition summit held in San Jose on Dec. 4-5 in partnership with The Center for Public Sector AI and Government Technology.
The city follows eight guiding principles in its approach to AI.
- Effectiveness — Ensure that systems are reliable, meet their objectives and deliver precise and dependable outcomes for the utility and contexts in which they are deployed.
- Transparency — The purpose and use of systems is proactively communicated and disclosed to the public.
- Equity — The systems deliberately support equitable outcomes for everyone; effectively managing any bias in systems.
- Accountability — Roles and responsibilities govern the deployment and maintenance of systems and human oversight ensures adherence to relevant laws and regulations.
- Human Centered Design — Systems are developed and deployed with a human-centered approach that evaluates AI-powered services for their impact on the public.
- Privacy — Privacy is preserved in all AI systems by safeguarding personally identifiable information and sensitive data from unauthorized access and manipulation, according to the city council’s Digital Privacy Policy.
- Security and Safety — Systems maintain confidentiality, integrity and availability through safeguards in accordance with the city’s Information and Systems Security Policy.
- Workforce Empowerment — Staff are empowered to use AI in their roles through education, training and collaborations that promote participation and opportunity.
When a city department wishes to procure an AI tool, it follows a review process to assess the benefits and risks of the AI system.
In its use of AI, Murray City aims to be transparent, particularly if using GenAI. Pictured is Murray City Hall. (Photo provided by Murray City, Utah) From left, Albert Gehami, city of San Jose Privacy Officer, and Leila Doty, privacy and AI analyst, helm a booth during National AI Literacy Day. (Photo provided by San Jose, Calif.) Murray City, Utah, has used AI to transcribe meeting minutes, which has saved time. However, the AI transcription is not always accurate and requires human proofing. Pictured is Murray City’s council chamber. (Photo provided by Murray City, Utah) Albert Gehami, city of San Jose privacy officer, drives the city’s road safety condition pilot test vehicle. (Photo provided by San Jose, Calif.) 
Artificial intelligence AI and machine learning concept. Computer processor chip with microchip isolated and electronic circuit pattern on white background. (Shuttershock)
