Pong Xiong, Director of Minnesota’s Driver and Vehicle Services, knows the challenges of translating languages to improve citizen services firsthand.

Mar 19 2024

State Governments Deploy Contact Center AI to Bolster Customer Service

Automating tasks via artificial intelligence improves interaction with residents seeking services.

Minnesota’s Driver and Vehicle Services wants to improve customer service. The agency acknowledges that getting a driver’s license and other services there can be a drag for some, intimidating for others and seemingly impossible for non-English speakers.

“From our community listening sessions, we heard loud and clear that it was hard to get services from DVS, but it was significantly harder if you didn’t speak English,” says DVS Director Pong Xiong.

DVS has embraced artificial intelligence in its contact center to better serve its diverse population and improve customer experiences. Last March, the agency launched an AI-powered website chatbot that can answer citizens’ questions and provide services in the state’s four most widely spoken languages: Spanish, Hmong, Somali and English.

“We’re seeing great utilization,” Xiong says. “These experiences are in the languages that folks prefer and need, and it enables them to interact with DVS without needing help.”

Click the banner below to explore what transformation means for state agencies.


Local and state governments are increasingly using AI in their contact centers to augment their staff, improve citizen services and operate more efficiently.

Forrester analyst Max Ball says that organizations have begun to deploy AI in their contact centers in several ways: self-service functions, such as chatbots and call center interactive voice response systems; call routing; and agent assist, which provides live agents real-time information and recommended responses to answer questions or resolve issues.

Other contact center uses of AI are intertwined: Sentiment and quality management uses AI to measure and monitor call center agent performance, and reporting and analytics help organizations spot trends and patterns so they can make improvements, he says.

“Contact centers face more demands than they’re ever going to have the resources to provide, so anything that they can do to make things more efficient or improve self-service is great for the contact center. AI is a huge part of that,” Ball says.

Meet Minnesota’s Multilingual Virtual Assistant

DVS, a division of Minnesota’s Department of Public Safety, is deploying Google Cloud’s Contact Center AI cloud-based software, which uses AI to enable virtual agents to assist customers via web chat or over the phone.

The agency needed to improve customer service because its busy, 35-person call center staff answers only half of the 30,000 phone calls it receives each week. Many residents who do reach a live agent endure significant wait times, Xiong says.

The goal is to reduce call load and make it faster and easier for Minnesotans to get information and services in several languages. That way, “our call center team can focus on those difficult issues where you really need to talk to someone,” he says.

DVS and its consultants spent six months building the new virtual assistant website chatbot and rolled it out in March 2023. To implement it, they used Google Cloud’s Dialogflow, a tool with a natural language processing engine for developing conversational experiences with chatbots and virtual agents, and Google’s Cloud Translation API, which translates text into different languages.

The agency had a web chatbot before, but it was rudimentary and keyword-based. “The new, multilingual virtual assistant creates a more casual, conversational flow for our customers,” he says.

Source: Minnesota’s Driver and Vehicle Services

How AI Learns and Grows Dynamically

The DVS AI model continually learns and improves as more people use the virtual assistant. Over time, it will better understand the intent of people’s questions that were not initially programmed into the system.

“For example, as we have more Spanish-speaking users, the model learns more Spanish dialogue, and it makes suggestions for adjusting the way we present Spanish back to customers,” Xiong says. The large language model also adjusts to local Minnesota dialects.

DVS is working to integrate Google’s AI technology into its Cisco call center technology, which will allow virtual agents to converse with callers and answer common questions in the state’s four most popular languages. The agency wants to launch the service as soon as this year. “We’re hoping that resolves some calls before a caller gets to a live person,” he says.

EXPLORE: Data governance strategies for AI success.

The agency’s virtual assistant technology currently uses machine learning, but it plans to take advantage of Google’s generative AI capabilities in the future. Today, the agency still maps dialogue and conversations, but when it begins to use generative AI, it will be a more dynamic experience, Xiong says.

In the meantime, the new website chatbot has made a positive impact. Residents who speak Spanish, Somali and Hmong are thrilled that the agency is offering services in their languages, he says.

“I grew up translating for my parents,” says Xiong, who is Hmong American. “As an individual, you want to interact with the government on your own. Having to communicate through someone else makes you feel small, whereas now, you can independently interact with DVS. That is empowering for individuals.”

How Agencies Can Achieve Better Customer Interactions

New Jersey’s 211 system is also using AI to bolster its contact center operations by making web search results more precise and relevant and by updating and improving the accuracy of its resource database, says Rory Britt, IT manager for NJ 211.

The free, 24/7 information and referral service connects people to health, human and social services, using Five9’s Intelligent Cloud Contact Center software to manage and route phone calls to its multilingual team of 110 contact center agents.

The agents use different tools to text, email and hold live web chats with customers. They also use a resource database that houses information on 6,300 available government and community-based programs and services.

WATCH: Buffalo’s director of citizen services talks about taking the city’s 311 call center virtual.

Earlier this year, NJ 211 launched a new website that uses AI to improve its search capability. The website’s previous search capability was “too terse” and required people to use the exact taxonomy term to find services, Britt says.

To fix that, the organization built a new website on Google Cloud Platform that uses Drupal as its content management system and Elasticsearch, which contains a copy of the organization’s resource database, as its search engine.

Through Elasticsearch’s natural language processing and machine learning capabilities, the new search engine can better determine users’ intent and give them the search results that they are looking for, he says.

“When someone says, ‘I’m hungry,’ that could mean they need a food pantry, soup kitchen or food benefits,” Britt says. “Now, our search can figure out their intent and return suggested search results that include those options.”

State and Local Agencies Invest in Future Improvements

NJ 211 also uses data analytics to further improve search results. The agency has deployed Power BI on Microsoft Azure, which uses machine learning to analyze what people are searching for and what search results they are clicking on. Through this analysis, it makes suggested improvements that allow the NJ 211 staff to fine-tune search results further, Britt says.

The agency has partnered with a vendor that uses generative AI to improve NJ 211’s directory of services. It uses ChatGPT to compare database listings with the industry’s style guide to make sure each listing’s description, hours of operation, contact information and eligibility rules conform. Uniform listings make it easier for people to search the directory, he says.

The vendor is also using ChatGPT to check NJ 211’s resource database to make sure listings are accurate and up to date. The NJ 211 staff reviews the suggestions to make sure they are accurate and then updates the database.

“It can save us hours of updating records,” Britt says.

Photography by David Ellis

Learn from Your Peers

What can you glean about security from other IT pros? Check out new CDW research and insight from our experts.