How San Diego built a foundation for AI at scale

10 August 2026

by Jonathan Andrews

San Diego is combining centralised governance, workforce training and practical deployments as it expands the use of artificial intelligence across its 13,000-strong workforce.

The city completed its AI policy and governance framework in 2024, bringing together the Department of IT, Human Resources, Office of Race and Equity, City Attorney, City Clerk, public safety, communications and performance and analytics teams.

All city employees then completed mandatory training on the policy before San Diego began deploying tools including Microsoft Copilot 365, Google Gemini and Copilot Chat in 2025.

Jonathan Behnke, CIO, City of San Diego, said rapid adoption and growing demand for more sophisticated applications led the city to establish a cross-functional AI centre of excellence.

Jonathan Behnke, CIO, City of San Diego

“We wanted to support that large volume of AI requests and minimise the risk of rogue AI across our city departments,” he told Cities Today.

The centre manages the city’s overall AI strategy, governance, standards, data enrichment, risk management, security, infrastructure, change management and financial oversight. Contracts with managed service providers have also been amended to align their work with the new structure.

An AI launchpad on the city’s intranet gives employees one place to access policies, governance information, approved solutions and training. It also includes an intake process through which departments can propose applications.

An AI governance subcommittee reviews those proposals, with 33 use cases submitted so far. The process gives the city a central route for reviewing departmental proposals while reducing the risk of uncoordinated AI use.

San Diego’s initial deployment of Microsoft Copilot 365 also underlined the importance of investing heavily in communication and training. Employees can access weekly live sessions, lunch-and-learns, office hours and guidance tailored to particular departments and functions.

“What we thought was maybe going over the top was exactly what the employees needed,” Behnke said. “Some people learn better with videos, some with hands-on workshops, some with live training and some with cheat sheets. We wanted to make sure we were providing all of those channels because everybody learns differently.”

A survey following the deployment estimated annual savings of 47,500 employee hours, equivalent to US$3.8 million in value. Most respondents said the technology improved the accuracy of their work, while most said withdrawing it would cause significant disruption after it had become part of their daily operations.

The city is now working with Google on an AI-supported procurement workflow being developed in response to California’s AB 339. Behnke said the law, which took effect on 1 January, requires government agencies to give labour organisations 45 days’ notice before certain procurements involving work performed by represented employees.

With Human Resources anticipating more than 12,000 submissions this year, the workflow is designed to help departments prepare statements of work, check submissions for compliance and allow corrections before human review. A conversational AI agent will be able to analyse submissions and identify potentially affected job classifications. The system will then support the final notification and start the 45-day review period.

Behnke sees scope to adapt the model to other processes involving submissions, departmental reviews and statutory deadlines.

The city will share its experience when it hosts the City Innovation Network (CIN) Leadership Forum North America on 9-10 September. The San Diego event will be the 61st city leadership meeting convened by the Cities Today Institute and the third regional meeting of CIN.

Behnke said discussions will cover internal efficiencies, public-facing applications, governance, executive sponsorship and securing budget support. One emerging concern is vendors’ shift from per-user licences towards consumption-based pricing, which makes future AI costs harder to forecast.

“I think it’s so beneficial for us to collaborate and exchange ideas, and learn from each other’s experiences,” he said. “I’m very excited about the September meeting.”

Main image: Sean Pavone | Dreamstime.com

https://cities-today.com/wp-content/uploads/2025/01/dreamstime_m_158418624-image.jpg

How collaboration can improve micromobility within cities