There are a number of different tools that can be used for this purpose, and the emergence of such tools is not slowing down. The technology is expected to change work, helping make certain processes simpler or more efficient, including data management.
One example is Augintel, software that leverages natural language processing (NLP), a subdiscipline of AI, to help process extensive amounts of case note data for those in behavioral health and child welfare industries.
This tool has been adopted in government by entities including the Illinois Department of Children and Family Services (DCFS) and the Allegheny County Department of Human Services (DHS) in Pennsylvania.
As Katy Collins, chief analytics officer at DHS, told Government Technology in 2022, case notes hold large quantities of textual data, and this technology enables workers to combine the information and get broader insights about the people they serve.
The software could, she said then, help flag early warning signs for opioid use to help caseworkers understand and address these risks earlier.
She underlined the value of this software in the instance of a case transfer, offering a broad overview for caseworkers about what has happened within a case before they dive into multiple years of case notes.
For Illinois DCFS, the tech was adopted to offer child welfare caseworkers, supervisors and private agency provider staff a way to better manage narrative data about cases — and it has resulted in tangible time savings for caseworkers.
"Reading all of the notes word for word is very time consuming, especially when the reader is trying to find a specific piece of information," Julie Barbosa, DCFS chief deputy director of strategy and performance execution previously told Government Technology in a written statement.
Augintel CEO Marty Elisco defined NLP as the ability of computers to understand the written word.
He said that generative AI, which has become more mainstream recently, has extended the work the company is doing by taking critical data and reconstructing it into what he called “abstractive summarization.”
He noted that for public-sector entities, adopting AI technology is not as simple as integrating off-the-shelf tools like ChatGPT into workflows. Because of the unique needs of public-sector agencies, he said there is work to be done to ensure a language model understands the nuances of the organization’s language needs; this includes regionally specific terms, agency-specific acronyms and more. This language model was designed to understand child welfare, behavioral health and social services.
What the software does is identify critical information buried in notes, he said; it does not create or generate anything new. It is not intended to replace the work of a human, but rather to complement it. Elisco said this tool can save caseworkers five hours a week.
In addition to the time savings for caseworkers, there is also the impact from an organizational perspective. Being able to look at narrative data across an organization can provide insight into how well that agency is performing and justify the tax dollars being spent. Elisco said Augintel customers have increased their compliance by up to 30 percent, and realized a 20 percent time savings on administrative tasks.
The software supplements an agency’s existing system with the ability to integrate the data from Augintel.
“And this is possible because, in a sense, what we’re doing is structuring the unstructured data,” Elisco said. “Our technology is making that qualitative data quantitative, so that it can be used for trending analysis.”
While primarily used in behavioral health and child welfare, Elisco said the tool can be useful for any organization providing value-based care. Its language model has not yet been trained in these areas, but Elisco sees potential expansion opportunities for the tool into public safety and other government services where officials use narrative data to provide care.
“What our tool does is enable these organizations to identify the individuals that require social care,” he said.
Augintel is not the only tool helping government agencies manage data with AI. Another example is mySidewalk, which recently launched an AI-powered data assistant, Sidekick. According to CEO Stephen Hardy, this new tool offers a way to put actionable data into people’s hands. Signups for Sidekick began in January.
MySidewalk pulls data from 65 sources including federal government repositories like the Census Bureau and the Environmental Protection Agency, and propriety data sets. Its large language model provides actionable insights informed by a decade of collecting comprehensive community data. The company also does its own projections to fill gaps. Hardy said roughly one-third of the data the company uses is publicly available in various formats.
MySidewalk CTO Matt Barr said plans include leveraging more qualitative data from areas like the language guidance of a public health department’s brand book, to ensure an inclusive data story is told. Sometimes, Barr said, data analysis projects are not undertaken because of uncertainty about the length of time required or the benefit or cost. A big goal for Sidekick is to make the process simple and affordable enough for organizations to do those projects.
“We know who we work for, and it's people for whom trust — [and] accuracy — is really important,” said MySidewalk Vice President of Marketing and Business Development Adam Runner. This new tool makes the expertise and data built into the company’s products more accessible through technology-enabled interfaces. “It’s all about trust.”
The AI-powered data assistant helps lower the barrier to entry, Barr said, for people looking to interface with data to better serve their communities, which really illustrates the potential of AI in this space: “Our goal is to democratize data.”
