SANTA CRUZ COUNTY WILL FURTHER integrate artificial intelligence into its government workforce in an expansion of a policy that tested how the new tools could be used in 2025.
The county Board of Supervisors heard a presentation on Tuesday about the move to a new phase of AI use that will standardize workforce training, uses and oversight.
The new phase follows testing of AI that started after the county crafted a policy governing its use in 2023. The county encouraged individual employees to start using large language models — known as LLMs — and generative AI for certain tasks, as long as they did not remove human decision making.
The goals of creating the policy were to explore what tasks AI could help complete and how a chatbot could improve searches on the county’s website, according to Tammie Weigl, director of the county’s Information Services department.
Supervisor Justin Cummings said he would support the policy but said it was important for people to still connect with county staff when they want to speak to a person. He also said it was critical that the tools did not take the jobs of staff in the future.
A chatbot is an LLM coded and trained specifically on a limited dataset or website. In this case, the chatbot would help users navigate the county’s website, find answers to common questions, and connect them with county staff who can best help them.
Weigl said developing the chatbot was one of the main reasons for creating the county’s AI use policy because the search function on the county’s website often did not return usable results. A specialized LLM chatbot could help residents find answers to a range of questions about county operations, services, and general information, she told the Board of Supervisors.
“I’ve gotten enough feedback over the years that sometimes on our website it’s a little difficult to find quick answers,” she told the Board.
The AI use policy created in 2023 focused on protecting privacy, ensuring human oversight of its use, and disclosing and getting user consent when interacting with AI.
New phase, tighter oversight
The informational presentation detailed a new phase of the county’s embrace of AI, which will help department heads track usage and ensure it meets the county’s policy and come up with new ways to use the tools.
Previously, employees had been experimenting with using the tools on their own accounts. Now, the county will manage staff accounts. Staff will use AI for various needs, including automating tasks and generating work products.
The county has narrowed the choices of providers to either ChatGPT, owned by OpenAI, and Claude, the AI tool from Anthropic, both of which are based in San Francisco. The selection will be made in May.
The county’s use is being expanded under a plan called the Elevation and Standardization Workplan, which aims to also create more comprehensive oversight of employee use to ensure it conforms with the county’s policy on AI. Weigl said the policy was designed to “elevate” the work of staff, not replace human judgement.
She said her team used Claude Sonnet 4.6 to make Tuesday’s presentation slides before they reviewed and approved them, which Weigl said was an example of using AI in a way that was in line with the use policy of using it for efficiency without removing human oversight.
The new phase will standardize training and uses that have otherwise been largely left to employees to do on their own. It will also involve prompt training to help staff get the most out of the tools, Weigl said.
Data on the county’s website will need to be edited and standardized in ways that help the chatbot be more accurate — a process known as “cleaning” the data — before the chatbot can debut.
Weigl, who has worked for the county for decades, said that when email was new, only some employees were assigned addresses and used it. She said it gradually expanded to more employees over time before becoming a routine use for nearly all of them, and she predicted the growth of AI use would be similar.
About 1,800 county employees already use AI in some form. She said the new phase will make it a routine, daily part of their lives.
Supervisors voted unanimously to authorize the new phase and update the AI use policy to reflect its increased usage. They also asked for a progress update in September, which would be shortly after two automation tools debut, according to the current timeline set forth by the Information Services Department.