Fewer workers? Retirement playlists? How local government workers are using AI
AI-generated summary reviewed by our newsroom.
- Cities report time savings and pilot deployments, but experts warn of bias, hallucination.
- Cities deploy AI tools to boost staff efficiency
- Officials say human review of AI work is important
Employees of Idaho’s third-largest city have access to “NampaGPT,” as they call it, a play on “ChatGPT,” the popular artificial intelligence chatbot.
In Garden City, employees have asked AI for everything from creating an anime image to designing parks.
Workers and managers have been grappling with how to harness generative AI since it burst into the spotlight in late 2022. Supporters believe the technology can increase productivity and boost the American economy. But detractors worry that AI is disrupting the job market, undermining the human experience and harming the environment. Others worry the economy is in an AI bubble.
How does AI work?
City governments in the Treasure Valley are increasingly embracing AI, such as chatbot services like ChatGPT, which are large language models. Those models work by predicting the next word in a sequence, sort of like autocorrect on steroids, according to Barrie Robison and Luke Sheneman. Robison is the director of the University of Idaho’s Institute for Interdisciplinary Data Sciences, and Sheneman is university’s director of research, computing and data services.
Large language models, or LLMs, offer a convincing mimicry of human thinking, but aren’t actually doing so, Robison and Sheneman said. An LLM’s knowledge is based on what it was trained on, such as news articles, books and websites.
There are inherent risks in using these tools. For example, these large language models can be trained on works with bias, which means they can repeat things that are biased or wrong.
Also, AI companies can collect personal or proprietary information that people provide in their prompts — questions or instructions — to the chatbots and use it to train future models.
These large language models also can hallucinate, or make things up. It’s not entirely clear what causes AI to spin fake facts out of nowhere. But a recent paper from OpenAI, the company behind ChatGPT, suggests it’s because these bots are designed to provide an answer instead of saying they don’t know. So the AI offers something, even if it’s wrong.
Some Treasure Valley governments are using the technology to try and get ahead and be more efficient. However, several have developed guidelines for employees. These include not putting private information in, reviewing what the AI puts out for accuracy, and in some cases telling the public when AI was used for a memo or press release.
How are cities using AI?
In Nampa, employees have asked NampaGPT for help with explaining a rescinded job offer, finding the title of a book about a hamster and a mole, and creating a retirement playlist. NampaGPT did tell one worker, incorrectly, that the year was 2023.
Nampa has spent about $115 a month on AI. Chief Information Officer Butch Schierman said AI is not a panacea. For security, Nampa’s IT department blocks access to external AI tools to city employees using city equipment, but city employees have access to an internal tool. That tool is similar to ChatGPT but it doesn’t have images, spokesperson Amy Bowman said in an email.
“We want to be smart with it,” Schierman said. “We have no shortage of work for people. There’s lots to do, and so if I can make their job more efficient, and they can get more done in a day’s time, we get more done as a city. Obviously, the taxpayers benefit.”
The city of Nampa is not planning to lay people off as AI becomes more integrated into work, he said. But Schierman said he can see a world where fewer staff per 1,000 Nampa residents are needed.
Schierman also touched on a new legal area for AI: whether AI prompts from city employees count as public records that cities must keep in case people want to see them.
Right now, most Nampa workers are deleting their prompts, and Schierman said that’s fully legal. It is unclear why employees are deleting their prompt histories, said Bowman, the Nampa spokesperson.
Not every city is using AI to the same extent. Caldwell employees are “not engaged in the robust use of” AI, according to spokesperson Char Jackson.
“However, AI tools may occasionally be used to assist with certain writing tasks, such as drafting or refining communications, to improve clarity and efficiency,” Jackson wrote in an email.
The city of Meridian, in response to a public records request seeking ChatGPT prompts, said the city itself did not have a ChatGPT account and didn’t know of any other city-sanctioned accounts.
The public records department did not ask each employee whether or not the employee had conducted city business using personal ChatGPT accounts, and indicated that asking such a question would cost the Idaho Statesman a lot of money.
“The City does not provide ChatGPT accounts to its employees, however use of AI tools like ChatGPT is allowed, provided they follow the City policy,” said Meridian spokesperson Trevor Smith said in an emailed statement.
One Garden City employee asked AI for funny things to put googly eyes on.
In Bonneville County and Idaho Falls, a new AI named Annie is answering non-emergency dispatch calls, according to East Idaho News.
Benefits and risks of AI
AI is best used as a tool and shouldn’t be relied on completely, said Margaret Sass, a lecturer at Boise State University who includes AI in her curricula. AI shouldn’t replace thinking, Sass said.
Charles Wadams, city attorney and director of human resources for Garden City, echoed that.
“The main point is that AI does not make decisions for the City of Garden City; people make decisions for Garden City,” Wadams wrote in an email. Garden City police do not use AI at all, Wadams said.
Garden City staff members are drafting an AI policy that would include allowing AI to draft policy documents, analyze data to inform decisions and help draft constituent emails. AI could also be used for simplifying “complex government language,” he wrote. But that’s only if humans review and verify what the AI has put out.
The City Council still has to approve the language.
A study found that the employees who get the most help from AI are more lower-skilled employees, said Sheneman, with the University of Idaho. It makes more sense to use AI for some task that someone is weak in and to take what the AI has done and build upon it, Robison and Sheneman said.
In Boise, employees used AI to help draft parts of the city’s AI policy, said Kyle Patterson, Boise’s director of organizational effectiveness.
Workers with ChatGPT licenses have self-reported saving four hours per week, he said.
“I’d say I really take seriously our role as stewards of taxpayer money,” said Patterson. Boise has spent $20,000 since June 2024 on ChatGPT licenses and has budgeted $60,000 next fiscal year. “If there are ways that there’s clarity and good research that suggests that using these tools help us do more with less resources, then I’m really excited about that as long as we can manage the risks.”
AI is still growing, so much of what it can do right now comes with an asterisk.
Ultimately, Patterson said, the organizational effectiveness director. One day homeowners could use AI to apply to build accessory dwelling units, or ADUs, on their properties.
“I don’t trust that we’re there yet,” he said.
This story was originally published October 13, 2025 at 4:00 AM.
