Before the mayor of San Jose, California arrives at a ribbon-cutting ceremony for a new business, his aides ask Chatgupto to help draft some story points.
“Elected officials do an incredible amount of public speaking,” he said that his recent itinerary has taken him from openings in new restaurants and semiconductor startups to a low-ride automotive culture festival.
Other politicians may be Skittish that the chatbot helped co-write speeches and drafted a new fiscal year's budget of $5.6 billion, but Mahan is looking to embrace artificial intelligence technology by increasing the number of nearly 7,000 government workers who run Silicon Valley's biggest cities by citing the example.
Mahan said adopting AI tools would eliminate Drudge's job and help the city provide services suitable for around 1 million residents.
He is not the only public or private sector executive to direct AI-or-bust strategies, but in some cases workers have discovered that expensive techniques can add hassle and mistakes.
“The idea is to try things out, get really transparent, look for issues, flag them, share them with different government agencies, and work with vendors and internal teams to solve problems,” Mahan said in an interview. “It's always bumpy with new technology.”
By next year, the city intends to have 1,000 or about 15% of workers trained to use AI tools for a variety of tasks, including pothole complaints response, bus routes, and using vehicle tracking surveillance cameras to solve crimes.
One of the early adopters in San Jose was Andrea Arjona Amador, who heads the Electric Mobility Program in the city's transport department. She has already secured a $12 million grant for electric vehicle chargers using CHATGPT.
Arjona Amador has set up a customized “AI Agent” to review the communications received for various grant proposals, and asked them to help organize incoming information, including due dates. She then helped draft the 20-page document.
So far, San Jose has spent over $35,000 to purchase 89 ChatGPT licenses ($400 per account) to use city workers.
“We used to work, but before we started using this, we spent a lot of nights and weekends trying to get a grant to the finish line,” she said. The Trump administration later retracted the funds, so she pitched a similar proposal to local funders who weren't tied to the federal government.
Arjona Amador, who studied Spanish and French before learning English, also created another customized chatbot for editing the tone and language of professional writing.
With close ties with some of the biggest players in the tech industry, including San Francisco-based Openai and Google based on Mountain View, mayors of the largest cities of the Bay Area help to promote the types of AI adoptions the tech industry is striving for, as well as guidelines and standards to avoid the harmfulness of technology.
San Francisco Mayor Daniel Lurie has announced plans to provide nearly 30,000 city workers, including nurses and social workers, with access to Microsoft's Copilot Chatbot. According to the San Francisco plan, it comes with “robust privacy and bias protection measures, and clear guidelines to ensure that technology is strengthened rather than achieving human judgment.”
San Jose has similar guidelines and has yet to report any major accidents in the pilot project. Such issues have attracted attention elsewhere due to the trend in misinformation-spewing techniques known as hallucinations.
The ChatGpt digital fingerprint was found in an error-filled document issued in May by US Health Secretary Robert Kennedy Jr.'s “Make America Healthy” committee.
In Fresno, California, school officials were forced to resign after saying they trusted the AI chatbot that created information in the document too much.
While some government agencies are secret about when they turn to chatbots for help, Mahan is open about the background notes written by his chat grit, which he relies on when giving speeches.
“Historically, it would have been hours of phone calls and reading, and you wouldn't have gotten those insights,” he said. “You can knock out these tasks at similar or better levels of quality in a much shorter time.”
But he said, “You still need humans in the loop. You can't trust the output by pressing some buttons. You still have to do independent verification. You need logic and common sense and you need to ask questions.”
Earlier this year, when Openai introduced a new pilot product called Operator, it promised a new kind of tool that goes beyond the capabilities of chatbots. Instead of analyzing documents and writing text, you can also access a computer system to schedule calendars or perform tasks on behalf of someone. The development and sales of such “AI agents” is currently an important focus for the high-tech industry.
You drive over an hour east of Silicon Valley. The Bay Area joined Central Valley Farm Country and Stockton City Director of Information Technology Jamilniazi had a big vision of what such agents could do.
Perhaps the Parks and Recreation Department can help AI agents book a park or swimming pool for their birthday parties. Or, residents could learn how busy the pool was before packing their swimming clothes.
However, six months later, after completing the proof of concept phase, the city did not purchase a full license of the technology due to costs.
Gartner, a market research group, recently predicted that over 40% of its “Agent AI” projects will be cancelled by the end of 2027.
The mayor of San Jose remains bullish about the possibility that these AI tools will help workers speed up digital paperwork “in the gut of bureaucratics.”
“There's an incredible amount of bureaucracy that a large organisation has to have,” Mahan said. “Whether it's writing finance, accounting, HR, or grants, these are the types of roles that employees think can be 20% more productive.”
