The AI craze sweeping the globe has already led many companies and industries to adopt artificial intelligence tools.

Now, the city of Oakland is dipping its toes into what some are predicting will be the biggest technological shift in decades.

For over a year, Oakland officials have been quietly exploring how to use AI to improve city services, and city staff have been working with a local university to come up with ideas, they told us.

Meanwhile, some city workers are already using AI.

Deputy City Administrator Joe DeVries told The Oaklandside that he and the city’s IT director formed a working group to explore AI roughly a year and a half ago. Mindful of the risks, DeVries said the group, which included staff representatives from every department, began by developing interim security guidelines for using AI in city work.

“Clearly, there’s a lot of fear and concern as a public agency,” said DeVries. “We wanted to make sure we did it right.”

In the near future, officials plan to ask companies to pilot AI tools across city departments. A draft document that the city shared with The Oaklandside lists 30 potential “use cases” for AI. 

The city wants to partner with companies to test some of these ideas. In exchange, Oakland would provide feedback on products before firms try to sell them to the broader market. 

The list of use cases serves as a useful window into some of the problems bedeviling Oakland’s government. Just a few examples:

Chatbots that can talk with residents about anything from code enforcement and permits to fielding complaints about potholes

A tool Oakland’s police watchdog can use to scan OPD body camera footage to speed up investigations

Software for the Finance Department to stop “revenue leakage” 

An AI agent to help the Oakland Animal Services Department match people with pets up for adoption

AI-enabled drones so the fire department can monitor fires. 

Technology that can comb through surveillance footage to catch people illegally dumping trash 

A platform that prevents corruption by reviewing the campaign finance reports of politicians to identify red flags 

A tool to review development proposals so the permitting process speeds up

In some of those cases, DeVries said, “obviously a human has to do the final look.”

Oakland established rules for using AI, but are they being followed?

The city working group issued guidance to staff in December 2024 on how to ethically use AI. For one, they have to get permission before putting city data into an AI tool.  

Another rule: Staff have to include citations when AI is used in any reports, memos, or other public records. 

City staff are also prohibited from using AI tools to gain access to unauthorized data or feeding AI systems city data that includes personally identifiable information, such as social security numbers or credit cards. The city’s guidelines also bar staff from using emails, Microsoft Teams messages, or any data from city applications or databases that contains information that could identify residents. And there is a blanket ban on “any non-public data.”

It’s clear that these guidelines are not uniformly applied, though. In one case where we’re aware AI was used — to generate an image of City Hall for a mayor’s office press release this month — no citation was included. And in records of chat logs we’ve received, where city staff ask ChatGPT for assistance writing emails, the city redacted some names before sharing them with us, indicating personally identifiable information was entered against the rules.

Along with these guidelines, the city adopted an “equity statement” applying to AI use. That document says Oakland will develop and use technology that “reduces racial biases” — a concern with many AI tools — and ensure privacy, transparency, and community engagement. 

Oakland is taking baby steps compared to San Francisco, which is home to major AI companies. In July, Mayor Daniel Lurie, who had OpenAI CEO Sam Altman on his transition team, announced that nearly 30,000 city employees would have access to Microsoft 365 Copilot Chat, an AI assistant that, according to Microsoft, can draft reports, analyze data, and summarize documents. 

Oakland contemplates dozens of uses for AI

According to DeVries, staff from about 25 different city departments attended a workshop in April at Northeastern University’s Mills College campus. During the meeting, staff and a couple of Northeastern professors developed a list of roughly 60 potential uses for AI in city services.

DeVries said the city plans to solicit proposals from AI companies that might be interested in working on the ideas. The original list of 60 ideas was whittled down to 30 in a draft “request for information,” a precursor to any contracts Oakland might sign with companies. The city is hoping firms will pilot projects over 16 weeks. 

”If it works well, we can consider actually developing and using a product more completely,” DeVries explained.

Northeastern is also lending graduate students to the city to help departments develop tools already in use — primarily Microsoft 365 Copilot. DeVries said about 24 members of the city’s AI working group will be trained at the end of the month on how to use Copilot when it’s attached to the internet. DeVries added that the version of Copilot he currently uses doesn’t have access to the internet, but he still finds it useful for recording and summarizing meetings.  

Nikki Lowy,  Northeastern’s director of community and city outreach, said students will study how Oakland staffers use Copilot and what challenges they encounter, with the goal “that this can form a larger rollout throughout the city at some point in the future.” 

Mayor Barbara Lee has a representative on the city working group. Through $2.14 million from the Kapor Foundation, her office is pursuing a number of initiatives, including a “tech week” for Oakland and launching an “innovation lab” at Northeastern that would include AI training and entrepreneurial support for the public.

“AI has the potential to dramatically improve city services for Oakland’s residents,” said Lee spokesperson Justin Phillips. “From helping to guide people through the permitting workflow, as San Jose is doing, to integrating AI into 311 to help residents access city services, as Denver is doing, the opportunity to use AI to benefit Oaklanders is there for the taking, and we are taking steps to ensure we do that in a responsible, ethical, and equitable way.”

How will AI impact the city’s workforce?

A prominent concern with AI is that it could displace human workers. DeVries said that Oakland will support workers by paying for most to further their education, and that city leadership wants workers to get trained in AI.

DeVries said this will hopefully “allow them to do more with less, which is a reality we’re facing in these tough budget times.”  

To gauge what the city’s workforce thinks of AI, DeVries said several departments were surveyed last year. About 85 employees responded, 55 of whom expressed concerns about using AI as part of their jobs. Respondents raised fears about data leaks and human workers losing their jobs to machines. Across the U.S., one in five workers reports using AI in their job. 

Others noted that investing in AI would put the city out of alignment with its own environmental policies. The AI industry is currently engaged in a data center building boom that is consuming vast quantities of materials. When complete, the data centers require enormous amounts of electricity, some of which is generated by greenhouse gas-emitting power plants.

One respondent pointed out that AI relies on human content, which means the city could end up replicating problematic trends. 

“If — to use my field as an example — the existing trends in city planning are problematic, content generated by AI will be trained on those problems and reproduce them,” the staffer wrote, citing problems like car-centric planning and disparities in access to city services. 

A minority of respondents, 30 city workers, said they didn’t have concerns with AI. Several pointed out potential benefits, like customer service chatbots, the ability to quickly translate city reports into languages other than English, smoothing out Oakland’s permitting process, and making complex policy papers more accessible for a general audience. But even the people who saw upsides to AI urged Oakland leaders to be cautious about how they implement the technology. 

“If the city is to use AI, I think it needs to be for simple, clerical processes,” wrote one respondent. “I don’t believe we should be using it for content generation or policy purposes at this phase.” 

“*” indicates required fields