How Government Call Centers Can Use Conversational AI

What Is Conversational Artificial Intelligence?

As IBM notes on its website, conversational AI includes technologies such as chatbots and other virtual agents that users can talk to.

“They use large volumes of data, machine learning, and natural language processing to help imitate human interactions, recognizing speech and text inputs and translating their meanings across various languages,” the company notes.

Conversational AI tools have two key elements: machine learning and natural language processing. “NLP processes flow into a constant feedback loop with machine learning processes to continuously improve the AI algorithms,” IBM notes. Conversational AI has principal components “that allow it to process, understand, and generate response in a natural way.”

EXPLORE: Virginia plans to expand its use of artificial intelligence in state agencies.

How Does Conversational AI Work? 

Natural language processing, as IBM notes, involves the combination of input generation, input analysis, output generation and reinforcement learning.

Unstructured data is “transformed into a format that can be read by a computer, which is then analyzed to generate an appropriate response. Underlying ML algorithms improve response quality over time as it learns,” IBM says.

First, with input generation: Users provide input through a website or an app via voice or text.

“If the input is text-based, the conversational AI solution app will use natural language understanding (NLU) to decipher the meaning of the input and derive its intention,” IBM says. “However, if the input is speech-based, it’ll leverage a combination of automatic speech recognition (ASR) and NLU to analyze the data.”

READ MORE: How is the use of AI accelerating in state government?

An NLU platform “evaluates a text string and attempts to decipher the author’s intent,” Nathan Cartwright, an intelligent customer experience architect at CDW, writes in a blog post. “An intent is the most basic task that is being requested by the customer. A customer may need to do something simple such as paying a bill or scheduling an appointment.”

Then the conversational AI tool forms a response, and machine learning algorithms refine responses over time to ensure accuracy.

“Since a chatbot is a form of AI that attempts to emulate human behavior, it must be able to decipher what a customer is requesting or validate data being given without having a list of keywords or phrases,” Cartwright says. NLU platforms are able “to set customer input into a pragmatic format to be passed to backend systems and then present data back to the customer in a human, readable format with context.”

RELATED: How does natural language processing help state governments?

How Are Government Call Centers Using Conversational AI?

Several government agencies have started using conversational AI technology in the past few years to improve their call centers.

One prominent example is the city of San Jose, Calif., which last year to deploy the tools for its call centers. The city had been missing its target thresholds for response times, according to Rob Lloyd, San Jose’s CIO and deputy city manager. The city wanted to help resolve nonemergency calls faster and take the load off its 911 contact center staff.

One of the key tasks San Jose focused on was deploying virtual agents to quickly resolve specific questions from residents.

As part of that endeavor, the city explored the evolution of translation technologies using neural networks. “In our research, we did find the language and literal translation as one of the human experience issues that people have when they’re dealing with their government,” Lloyd says.

That’s especially important in San Jose, which has sizable immigrant populations, including the largest Vietnamese population of any city outside of Vietnam. Spanish and Vietnamese are the two most prominent non-English languages spoken in the city. San Jose’s first pass at a constituent relationship management solution had good solutions for Spanish but not for Vietnamese, which is a complex language with influences that include Cantonese and French. For example, a test showed a notice about fireworks was translated as a bomb notice.

Still, the city kept at it. “We have a policy here to speak to our community where they are, and there are enough accessibility barriers. Language should not be one,” Lloyd says.

When San Jose worked with Google’s AI teams on the company’s Dialogflow CX tool, the quality of the translation was even better than expected, according to German Sedano, 311 products and projects manager for San Jose. “That’s the beauty of artificial intelligence, right? You can train a model, and the model gets better.”

Without the conversational AI tool, a resident would call the city’s 311 center, and the operator would need to bring in a translator if he or she did not speak Spanish or Vietnamese, for example. That cost both time and resources.

“Now, they don’t have to do it,” Sedano says. “They can enter their tickets in either Spanish or Vietnamese, and operators can focus on high-value work for residents. That certainly has a positive impact on the user experience.”

The most prominent way the city uses conversational AI is pushing out information to residents, Lloyd says. The city also uses it for asynchronous communication, where a resident submits a request in a foreign language, and the tools translate it. Conducting real-time translation in a call is the “most dynamic” use case, Lloyd says, but it’s still “an edge use case for us.”

Other cities have deployed conversational AI tools, including New Orleans, which in June launched an AI-powered chatbot called Jazz for its 311 call center. The city has faced a huge surge in call volume during the pandemic. “We quickly realized that we needed to diversify our communications platform and streamline the way we were delivering our services,” Tyrell Morris, executive director of the Orleans Parish Communications District, tells MeriTalk.

Staff at OPCD created Jazz’s knowledge base by connecting it to information from the 311 center’s customer relationship management platform and city websites, Morris tells Government Technology.

MORE FROM STATETECH: Learn why states are deploying computer vision technology.

The Benefits of Conversational AI to Government

Conversational AI can deliver many benefits for governments, as San Jose’s experience shows.

The first is simple efficiency. AI tools enable the city to handle more calls for service. In 2017, the city’s 311 center was handling about 165,000 tickets for service annually, Sedano says. Now, it is managing about 215,000. “I don’t think that would be possible without using this artificial intelligence technology,” he says.

The tool also improves customer satisfaction. In 2020, before the conversational AI tools were widely used, the city surveyed resident satisfaction with its 311 service and found that 28 percent of residents rated it as excellent or good and 69 percent rated it as poor. In 2021, those numbers were flipped, with 68 percent rating it as excellent or good and just 25 percent rating the service as poor.

The conversational AI tool was also better at translating human speech than auto-translate tools such as Google Translate — 22 percent better for Vietnamese to English and 51 percent better from English to Vietnamese, according to data provided by Sedano.

“That drive toward empathy and responsiveness and a faster, more efficient experience is what we’re focused on, and language is one ingredient in that equation,” Lloyd says. “If you’re highly responsive, you get back to them and have a nice service touch, they get their resolution quickly. And when we do the follow-up, then you have the highest ratio of satisfied customers.”