What SF can learn from San Jose's AI deployment

Driving the news: San Jose began piloting AI software this month that serves as a "pre-check" for permitting ADU applications to ensure answers are tailored to what's required.

• Over 90% of ADU applications in the city are returned to the customer because of missing or incomplete information, Mayor Matt Mahan told Axios.
• "There's no reason for two human beings to go back and forth in a game of ping pong over email for weeks, if not months, when you could have an AI co-pilot … giving you that feedback in real time, as though you were speaking to a city planner," Mahan said. "We can serve many more people with the same amount of tax dollars."
• The AI assist involves CivCheck software; staff are still manually reviewing applications as of now. Mahan says they plan to roll out the AI co-pilot to help ADU applicants as early as next year.

What they're saying: San Jose's pilot is part of a broader trend of local governments testing "really, really narrow use cases where tech can actually free up time," Kimberly Lucas, public policy professor at Northeastern University, told Axios.

• "Starting small" in a community-specific context is crucial to ensure AI isn't used inefficiently or improperly, Lucas noted.

What we're watching: San Francisco could benefit from a similar tool in its permitting process, which is notoriously slow.

• The city has "reviewed several AI tools" for permitting and is evaluating "how to incorporate them," Michelle Lynch Reynolds, programs communications manager of the San Francisco Office of Small Business, told Axios via email.
• Lynch added that they are also testing AI to help optimize workflows in 311 customer service, traffic safety and street response.

The intrigue: San Jose also recently released the results of another AI pilot program that involves using cameras on city vehicles to spot potholes, parking violations and illegal dumping in real time.

• The system — which Mahan says automatically blurs personally identifiable information like faces — identified potholes with 97% accuracy and trash or debris with 88% accuracy.

San Jose is using AI to optimize transit routes as well. The software taps into sensors on traffic signals and buses across the city to shorten and lengthen traffic lights' duration, predictively timing routes so buses are less likely to encounter a red light.

Reality check: The Bay Area is already seeing the effects of gen AI's dependence on large data centers, which require a high amount of electricity and water.

• The rapid expansion of AI infrastructure has also increased pollution and health risks, especially among marginalized communities, per a 2024 study from UC Riverside and Caltech.

Between the lines: Researchers have increasingly urged the public sector to adopt best practices, such as mandating investments in on-site renewable energy systems, regulating utility rate hikes to prevent cost-shifting to residents and requiring energy audits and disclosures.

What's next: San Jose is building an open-source AI training database for the public sector through the GovAI Coalition, which the city launched in 2023. San Francisco is one of over 800 members across the nation.

• The coalition's second annual conference will take place Nov. 5-6 at the San Jose McEnery Convention Center.