How generative AI is being used to put the humanity back into government

In the U.S., residents who speak limited English face challenges native speakers can only imagine—from navigating day-to-day challenges like reading signs in English and conversing with members of the community to accessing basic services such as education, health care, housing and employment. 

Not being able to communicate with state and local officials to, say,  get a drivers’ license or register a vehicle makes it difficult for these residents to find and keep a job in order to support their families. 

In Minnesota, where 21% of the population speaks a language other than English at home, the Department of Public Safety’s Driver and Vehicle Services division, or DVS, is making it easier for non-English speakers to access its services,  using a generative AI-based virtual assistant. 

Between its implementation in March and August, the virtual assistant has held about 150,000 conversations, the bulk of which are not in English, said Pong Xiong, director of DVS. The virtual assistant uses Google’s Translation AI technology to automatically translate English into Hmong, Spanish and Somali—the three most spoken non-English languages in the state. The new service replaces DVS’ previous chatbot, TAMI, which did not use generative AI or offer translation.

Generative AI, which is the same technology behind ChatGPT, allows DVS website visitors to ask the virtual assistant questions about how to apply for a driver’s license or register a car. The more the service engages with users, the more it learns and improves its responses. To ensure the translations’ accuracy, DVS also works with state residents who speak those languages. “We know we’re translating difficult terminology because we’re talking about DMV terminology,” Xiong said. 

DVS has also added self-service options through the virtual assistant. Specifically, users can ask the virtual assistant about the status of their orders, such as license plates or car titles. In its first two months of use, it helped with more than 6,100 self-service transactions.

Additionally, the virtual assistant has a built-in feedback tool. So far, DVS has received almost 1,650 comments through that mechanism. The feedback has mostly focused on any pain points in the process. So far, for instance, the department learned of and resolved problems users had logging into the self-service portal, as well as about inconsistent information on its website on how to request a driving record and pay online.

The department began researching ways to better reach non-English speakers after feedback from a 2020 town hall revealed that those residents found navigating DVS services challenging. “We know how critical it is to work with DVS to get your driver’s license, to get your car registered. These are all tasks that enable Minnesotans to do all the things that they need to do to be successful and access all the resources that the state has to offer,” Xiong said. 

Before using the translation tool, DVS sent material to a vendor to translate. When the translated text came back, the department disseminated it. “That became a really challenging process—one that’s not very agile, one that doesn’t react and change with what the customers need,” Xiong said.

Another benefit the virtual assistant provides is increased employee bandwidth. “This tool hasn’t taken any jobs away,” he said. “It’s allowed us to focus on those sticky situations where customers really need someone to look at their unique situation and help them understand and navigate that.”

One benefit to a more seamless translation is harder to quantify. “Now we’re offering services and communication to a population of Minnesotans that before this would have no way to interact with us independently,” Xiong said. “I grew up translating things for my parents, and with this available to them, it just makes them feel just a little bit less small, when they’re working with DVS, or more human.”

The rollout of the virtual assistant is a five-year, multiphase effort that includes adding more languages—Arabic and Russian are likely next—and integrating with the phone system and allowing for speech to text.

Beyond the translation to other languages, DVS is working to make the English it uses simpler and more accessible.  “We’ve worked really hard to remove a lot of the jargon,” Xiong said, “and just speak in a plain manner to our customers.”

Xiong said he believes DVS is the first motor vehicles department to introduce a generative AI assistant, but the technology is gaining traction and adoption in a variety of ways nationwide

At a conference in May, Amy Glasscock, program director for innovation and emerging issues at the National Association of State Chief Information Officers, or NASCIO, said that every conversation she had ended up talking about generative AI. “It’s just been increasingly at the top of mind for state CIOs.”

The organization has launched a generative AI working group for members. It meets virtually every month and invites states to share how they use or plan to use the technology, as well as the policies and legislation they’re working on.

“States are excited about what it can do to improve the citizen digital experience, but what’s really key … [is making] sure that citizens are aware of what’s going on and that their data is protected,” said Alex Whitaker, director of government affairs at NASCIO.

Local governments are also embracing generative AI. Boston is looking at use cases such as automating data analysis and coding, and Seattle has published guidelines for generative AI use (so has Washington state)., Experimentation is not limited to big cities: Wentzville, Missouri, a city of 45,000 west of St. Louis, is testing generative AI tools to automate some city communication.

Much like the Minnesota DVS, cities are using generative AI to improve the quality of engagement, according to Joshua Pine, urban innovation program manager at the National League of Cities. “What we’re seeing with cities is they’re using it to make the human connection that much more meaningful— whether it’s through providing translation services … or being able to automate … some of those tasks that really enhance the human experience rather than replace the human experience,” he said.