Policy & Practice | Fall 2025

state insights

By Heidi E. Mueller

Artificial Intelligence in Child Welfare

R ecently, I was asked how tech nology like AI and the data delivered by that technology are helping my team and me address chal lenges and work toward achieving our goals for the Illinois Department of Children and Family Services (DCFS). It’s an interesting question. Child welfare agencies are run on human power: with caseworkers and clinicians who provide support, and families and foster parents who provide homes and loving care. We face very human challenges: ensuring that our families have the resources they need to stay together safely and when that’s not possible, helping our kids find perma nent, loving homes; delivering the right services for the right amount of time for young people who have physical health and behavioral health challenges; and making sure that youth who are aging out of care are set up for success as young adults—just to name a few. Having the right data—information about a specific family or agency-wide trends—is critical in helping us under stand and address these challenges. In child welfare, accessing the right data can be difficult because most of the information about our children and families is captured in the reporting and observations by our staff contained in case notes. That’s why analysis of narrative data is so important. For more than two years, DCFS has been using AI from Augintel to access nar rative data to deliver insights that leadership can count on, better identify the needs of the families we serve, and empower staff. Here are just three examples of the impact of this work. Informed Data Delivery Our use of AI, and more specifically, natural language processing (NLP),

has greatly increased the accuracy of our data and expanded the amount of information we are able to gather. As a leader, it can be challenging to figure out how to access case note data to identify trends or evaluate processes. A recent DCFS audit pertaining to meeting inves tigation timelines is a great example. In Illinois, child protection inves tigators are required to make good faith efforts to visit, in person, alleged child victims within 24 hours and alleged perpetrators within seven days. Although, in most cases, inves tigators are able to conduct in-person visits within these timeframes, in some cases, despite good-faith efforts, a visit within these timeframes is not possible. These good-faith efforts, however, are often only recorded in narrative case notes, making them nearly impossible to quantify. Consequently, we have been unable to provide a comprehen sive account of our efforts, leading to underreporting and a lack of acknowl edgment of our caseworkers’ hard work and diligence. With the implementation of NLP to analyze narrative case notes,

the agency was able to verify, with data, thousands of investigations where good-faith efforts were made within the required timelines. These data resulted in much more accurate reporting and allowed us to evaluate and demonstrate significantly improved compliance. Data-Driven Approach to Identifying Unmet Needs Helping children and families thrive in our care requires a comprehensive understanding of their unique needs and delivery of the right services to address those needs. Caseworker notes are key to understanding those needs. One example of this is with our Home Visiting teams. They used NLP to identify cases of pregnant women in intact families across the state. Through this process, they discovered an addi tional 621 pregnant women who were not previously identified in the records. This new insight enhanced the team’s ability to identify unmet needs and provide available services to expectant mothers to address those needs.

Illustration by Chris Camnpbell

8

Policy & Practice Fall 2025