County considers how to deal with AI

“A lot of risk, a lot of problems, a lot of solutions”

by Katherine Nettles

Gunnison County has started the process of creating policy and a handbook for using artificial intelligence (AI) systems either directly by county representatives or on their behalf. During a work session with county information technology director Chris Lindner on June 23, county commissioners and staff reviewed a draft policy and handbook Lindner had prepared and agreed that approving an overall process to review new technology is the way to go versus trying to keep up with each new technological shift.

The goals defined by Lindner in the draft AI policy were to provide clear guidance for anyone involved with the county making decisions around or with AI, ensuring that everyone adheres to values around AI systems, define roles and responsibilities around the county’s AI usage, establish and maintain processes to assess and manage risks presented by AI, align AI system governance with existing data governance, define prohibited uses of AI systems, establish “sunset” procedures to safely retire AI systems as needed and define how AI systems may be used for legitimate Gunnison County purposes in accordance with applicable laws and policies.

Lindner said the issue had been increasingly pressing as his department finalized other aspects of technology improvements, and as AI technology improves and becomes more effective and widespread among government agencies.

“I think the last few years we’ve focused on operational stuff, replacing equipment and getting things in good shape for everybody,” he said. “But we didn’t really have a lot of policy, at least in the last decade, so I took a step back and focused on that.”

He said he pulled a lot of resources from established sources considered by the industry as best practices.

“Most of these policies were templates that were derived from an organization called GovAI,” said assistant county manager for operations and sustainability John Cattles. He explained that it is a coalition based in Silicon Valley and San Jose focused on empowering governments to leverage AI for the public good.

“We don’t know where AI is going to go in government. So, you start with broad [concepts], and what I think Chris did a really great job of is create processes by which you’re going to analyze future opportunities or tools to decide if they’re appropriate for the county,” said Cattles.

The county’s policy defines AI as “a machine-based system that can, for a given set of human-defined objectives, make predictions, recommendations or decisions influencing real or virtual environments.

The goal overall was to provide some structure to assess new technology in a systematic and deliberate way without being too prescriptive, said Cattles. The policy identifies different levels of risk which warrant more oversight from commissioners or the county manager.

“Lack of understanding leads to high-risk behavior,” said Cattles of AI. “And we want to make sure that our staff understand what could be high-risk.”

Lindner said many county departments could find AI useful and even important. “There’s a lot of data entry; there are a lot of tasks…we’re not trying to replace anybody’s job or reduce the workforce. We’re trying to free up time for high-value projects. I think we’re a way out from even that,” he said, but acknowledged that increasing efficiency might eventually lead to fewer new hires. County manager Matthew Birnie and Cattles agreed that it was difficult to predict what the distant future might hold.

Cattles stated, “It’s going to be easier for a tool to give us an analysis and have a human double-check that. I think that’s going to be a big part of our deployment. But we are not going to automate any work and leave humans out of it.”

An example was the county’s use of AI to evaluate county road conditions across the entire road system, and then use human expertise to further identify their relative importance and usage.   

There was some discussion around the policy’s recommendation for public engagement where relevant, as an additional safeguard and accountability factor. Cattles said that would be particularly important if there were a new technology such as a software being rolled out from which members of the public would not be able to opt out. The process of oversight by the commissioners, county manager and public input would need to be fleshed out more by the board and staff before a final draft, he noted.

Commissioners had some comments and questions but expressed general comfort with the draft. Commissioner Laura Puckett Daniels said transparency was important, while balancing efficiency. “To your point, we’re okay to move slowly. We want to do things right,” she said. 

Lindner said the county currently uses Copilot for free, and there are some upgrades available for various levels of cost that interact with organizational data and then require more deliberation and security. “There’s a lot of risk, a lot of problems, a lot of solutions,” he summarized. He asked commissioners how aggressive they want to be with all that as they consider the budget for 2027 in a few months, or in the more distant future. He also acknowledged the environmental concerns associated with “overusing AI.” He posed the question, “How much do you want to invest in the next level?”

Commissioner Liz Smith said a threshold for her was a demonstrative benefit to the public, and she would need more information to work that out.

Commissioner Puckett Daniels said she wanted to evaluate the options but felt that something should be approved quickly or else staff would proceed with unauthorized use anyway. Everyone else at the meeting concurred.

“We don’t get to be stagnant,” said commissioner Jonathan Houck. He said new technologies can be difficult at first, but learning to navigate and use them is the only way to adapt. Commissioners gave Lindner approval to continue researching an enterprise solution.

Lindner said at about $30,000 per year, Claude is a more cost-effective option than ChatGPT, for example, but he will come back to the board with more specifics for approval.

“This conversation is a great start,” said Cattles. He noted that the policy could start going out for all onboarding of new staff, all staff training and to managers. “We know we’re behind and we need to go quickly to integrate this. So we’ll start working with HR right away.”