New York City’s recently unveiled artificial intelligence action plan aims to “strike a critical balance” between empowering agencies to deploy AI to improve lives, while protecting against harm.
Key to meeting that goal will be educating the city’s more than 300,000 workers about how they can build their AI knowledge and skills so they can “effectively and responsibly work with and on AI,” the action plan says. The city aims to develop centralized policy, guidance, support and resources for staff engaging with the technology and expects to partner with industry and academia on projects to address some of the city’s key challenges.
To build employees’ knowledge and skills, the city has various projects in mind. First up will be encouraging agencies to share information on their uses of and concerns with AI.
It also plans to identify the high-priority skills agency staff will need and update applicable city job descriptions and titles to reflect the range of AI talent required to support city efforts. To help bring the workforce up to speed on using the technology, the city will assess the training available both internally and externally and explore the feasibility partnerships that bring in outside people to work on short-term AI projects.
The best way to interest and educate staff on AI is to show them how it can be used in their jobs, experts say. That way, it goes from being abstract and scary to something that has practical, everyday uses.
“If I'm in government, I may not be that technical to begin with, and a lot of these terms may go over my head,” said Beth Noveck, a board member at nonprofit InnovateUS, which trains public sector employees on digital tools.
Staff may also be wary of “examples are not going to be relevant to [their] life, because they're often directed at computer programmers or technologists,” she said in an interview at the recent Google Public Sector Forum in Washington, D.C. Successful AI training must include “examples and references that will make sense” to staff in their daily work, she said.
At this stage, Noveck said the best way for employees to learn about AI is to experiment with it themselves, especially by using publicly available generative AI tools like ChatGPT at home. She said people must be willing to “stick a toe in the water in order to learn how to swim."
Training can also break down preconceptions that shows city staff that AI is “not the threat it’s made out to be,” said Dearborn, Michigan, Mayor Abdullah Hammoud during a prerecorded video at the Google Public Sector Forum.
New York City’s action plan also calls for supporting the implementation of AI by city agencies, including “comprehensive assistance” through the lifecycle of AI-related projects and initiatives. That effort involves identifying common challenges agencies face and empowering them to build AI tools in-house where possible and applicable.
In the near term, city leaders are looking to develop example lifecycles for AI projects that map out the roles agencies will play in implementation as well as the potential “bottlenecks” that may stymie implementation, including agency procedures, software or infrastructure availability, financial constraints and skills gap.
While building AI projects in-house may be attractive to many departments, experts warn it requires extensive collaboration across agencies, especially when it comes to data sharing and governance. A common data vocabulary can help make implementing AI projects easier, as will clean, accurate data.
“It's impossible to build and train algorithms that you can reliably trust without knowing you have quality underlying data,” Maryland Secretary of IT Katie Savage said in an interview at the Google forum.
To determine the viability of AI use cases, agencies should consider pilot projects, Jessica Gateff, Oklahoma’s deputy director of data services, said during a panel at the Google forum. The state has taken a “rigorous approach” to prototyping AI use cases with agencies, which helps determine their value and whether they can be scaled up.
But before AI solutions can make an impact in New York City, agencies must be able to buy and get support for the technology. That means producing contracting criteria that “support responsible use,” the action plan says, and finding ways to streamline AI contracting to eliminate redundancies and support cross-agency access to common tools.
New York City eventually wants to produce AI-specific procurement standards, terms and guidance for vendors to ensure that AI tools align with the city’s goals. Many procurement officials acknowledge the current process is outdated, but some governments have refined their procedures to keep pace with new technologies.
Wisconsin Secretary of Workforce Development Amy Pechacek said in an interview at the Google forum her agency has “flipped the script” on how they write and release requests for proposals and other procurement documents. Rather than treat an AI solution as the commodity to be bought, the agency instead prioritizes its desired outcomes.
“We said, ‘Here's our current state, here's our future state, this is what we want it to look like, this is what's going to be the outcome,’” Pechacek said. We tell the bidders where we are and where we want to be, then ask them to “propose how we get there. Whether that's AI, whether that's a different tool, whether it's the next thing after AI, bring it and that's what we're going to do.”