The city of San Jose is setting a precedent in the public sector’s adoption of artificial intelligence, with Mayor Matt Mahan openly integrating AI tools like ChatGPT into day-to-day governance and policy work.
From drafting speeches for ribbon-cutting ceremonies to helping shape the city’s $5.6 billion annual budget, Mahan and his team have embraced generative AI as a time-saving instrument that reduces administrative burden. “Elected officials do a tremendous amount of public speaking,” said Mahan. “You can knock out these tasks at a similar or better level of quality in a lot less time.”
The mayor’s office is not alone in this digital transformation. San Jose has already spent over $35,000 on 89 ChatGPT licenses, priced at $400 each, for city employees, with ambitions to train 1,000 staff (around 15 per cent of its workforce) in AI applications by next year. These tools are being used for everything from organising pothole complaints and rerouting buses to assisting in criminal investigations using surveillance data.
AI has not only been a strategic tool for internal bureaucracy but also a practical asset in securing funding. Andrea Arjona Amador, who leads electric mobility initiatives for San Jose’s Department of Transportation, used ChatGPT to help secure a $12 million grant for EV charging infrastructure. She created a customised AI agent to manage deadlines, organise correspondence, and even assist in writing the lengthy grant proposal. “We used to spend evenings and weekends pushing grants to completion,” she said. “This changed the game.”
Amador, fluent in Spanish and French before learning English, has also developed another AI agent to fine-tune the tone and clarity of her professional writing, streamlining communication across departments.
Despite the enthusiasm, Mahan remains cautious. “You still need a human being in the loop,” he said. “You can’t just press a couple of buttons and trust the output. Independent verification, logic, and common sense still matter.”
San Jose’s efforts are part of a broader trend across the Bay Area. San Francisco Mayor Daniel Lurie recently announced a plan to provide Microsoft’s AI assistant, Copilot, to nearly 30,000 city workers. The rollout includes clear ethical guidelines and privacy safeguards to prevent misuse.
While San Jose has not reported any serious issues with its pilot programmes, other jurisdictions have not been as fortunate. In Fresno, a school official resigned after relying on an AI-generated document that turned out to be inaccurate. Similarly, the US Health Secretary’s office faced scrutiny over an AI-assisted publication riddled with errors.
Elsewhere, cities like Stockton have experimented with more advanced AI agents capable of performing tasks such as calendar bookings and park reservations. However, Stockton ultimately shelved its plans, citing high costs. Gartner, a global market research firm, has predicted that over 40 per cent of agentic AI initiatives could be cancelled by 2027 due to financial and operational challenges.
Nonetheless, Mahan remains optimistic about AI’s capacity to modernise municipal functions. “There’s an amazing amount of bureaucracy that large organisations must manage,” he said. “Whether it is HR, accounting or grant writing, these are the roles where AI can make our workforce 20 to 50 per cent more productive, very quickly.”
With deep ties to major Silicon Valley firms such as OpenAI and Google, San Jose’s approach could become a model for local governments seeking to balance innovation with accountability in the digital age.