The state has been an early adopter of AI technology, from the February 2024 AI policy requiring risk assessments, to the use of AI in supporting job seekers, to the September 2024 launch of a beta version of an AI chatbot.
Now, SOS has partnered with Amazon Web Services (AWS) to migrate two divisions’ legacy Microsoft systems to the AWS cloud. According to SOS CIO Robert Fulk, public-private partnerships support innovation for the state by enabling it to leverage industry expertise: “The industry partners bring the latest technologies to bear for us.”
To enable this migration to the cloud, the state has adopted a “lift and shift” migration strategy. His nearly three decades of experience in the private sector, Fulk said, equip him with that sector’s mentality and speed with which to complete projects. The strategy used here entails taking existing legacy stack applications, moving them into a modern infrastructure, containerizing them, and then building around them. According to Fulk, this approach is faster, more effective and cheaper, and it will allow the state to take advantage of the latest tools and technologies — including AI.
As Fulk explained, the state needed an infrastructure with which it could take advantage of AI and data analytic capabilities. By partnering with AWS on cloud infrastructure to support enterprise systems, the state can build and implement AI tools to support its customer experience.
One such example is chatbots. For SOS, two new chatbots will support the user experience. To support the state’s INBiz platform and auto dealer licensing system, the state partnered with Google to launch “Liz” and “Otto,” respectively. Both tools currently support English and Spanish interactions. Prior to these tools‘ launches, chatbots for the platforms were not AI-powered tools, but rather, used decision tree answers that could support specific paths but had a limited scope.
In the coming months, the capabilities for these two new tools will expand with multilingual, multimodal and multi-platform capabilities, Fulk said.
The state is working to improve both products, using analytics and user feedback to make improvements as needed before adding other languages and capabilities. The tools support access to key metrics for how they are being used, even providing sentiment analysis of interactions. The state will then further train the tools based on initial interactions, to ensure answers align with what SOS wants to convey. As Fulk put it, training a system like this is akin to teaching a child how to address specific situations.
A notable piece of the state’s chatbot creation process is limiting the data that is used to inform the responses it generates, because when governments use the entire website as the data input, they risk including information that is potentially out of date, inaccurate or in a format that makes it difficult to interpret.
“If you try to boil the ocean and plug in everything, you can’t control everything; you don’t have the data integrity,” Fulk said.
To solve this problem, the state used internal scripts for call center environment employees with specific questions and answers to train the virtual agents. The data simply had to be restructured for this purpose. These tools will not use data from outside sources like the Internet at large to answer questions customers ask, which is an important guardrail for data integrity and security. This approach allows the state to generate success stories in a controlled way at an efficient pace.
Another AI-powered project in the works for SOS involves using AI technology to search, index, read and interpret the data within more than 100 years of records across four divisions — tens of millions of records. That project is slated to launch in the coming weeks.
“But the only way we can do that is a modern cloud environment,” the CIO said, noting the limitations of a legacy data center. “We can’t instantaneously scale the hundreds of servers or take advantage of the power of AI without being on the cloud.”
Another way the state is exploring the use of AI is by embedding AI into digital service processes themselves as a way to go beyond supplementing those processes with a chatbot to provide user support. As Indiana is designing more effective systems and processes, exploring how to automate those processes where possible can help the state scale its team and reduce costs.
However, Fulk did emphasize that while the technology can make processes more efficient for state workers, “AI will never be able to replace the humans.” He describes it as an enabler to support humans’ ability to work efficiently. SOS intends to use AI to better organize and disseminate the information that businesses and customers in the state need.
Looking forward, the state plans to use its efforts in mapping user journeys and migrating to the cloud to support other modernization efforts to meet users’ needs, which Fulk underlined are not monolithic. Users in the state, he explained, have differing technology needs and desires, so public-sector solutions tend to not be one-size-fits-all. For example, some customers may prefer to access services from a mobile device while others may prefer to visit a government building in person. As such, SOS is trying to provide alternative experiences for different users and ensure they are all effective and available to constituents.
“We’re here to serve the constituents and public,” Fulk said, “so I want to make sure we’re doing it in a modern way.”
