Washington, D.C.: helping social workers save time using AI-assisted contact notes

Working as a child protective investigator in D.C. is a high-pressure job in a high poverty and high-needs city—and getting workers up to speed is a difficult task. The team working on D.C.’s applications of AI were fully aware of how much time social workers spent “bogged down,” looking through case files, “trying to find information in years-long cases.”

Washington, D.C. is deploying a new, agency-wide child welfare information system. The first step in this process was digitization: turning paper into screens. Once that foundation was in place, the team began to explore where AI could make the biggest impact. Their first use case mirrored Arizona’s—policy search and guidance. With a new system and hundreds of new workers coming on board, it was critical to make policy access frictionless—as well as helping workers learn a new child welfare information system. The developers of the first use cases simply wanted to give workers a “leg up” in addition to the training they received. In theory, the AI application would support the training, and serve as an expert to advise workers. Initial adoption of the agency’s AI-powered ‘Case Agent’ was gradual, as staff adjusted to a new system and associated changes in workflow. The lesson was clear: timing matters. AI cannot succeed if underlying workflows are not yet stable.

The agency narrowed down from a list of 30 to just 4 use cases, with others in the pipeline. Some are more advanced than others.

One key use case for the agency is Case Agent, an AI-generated agent that quickly gives workers answers on a particular case. This is critical given the volume of information that can be collected for a single child in care. Alleviating this administrative burden placed on workers gives them time back to think about how to meet the needs of children in care. While not live yet, this use case could make a caseworker’s job more manageable.

Contact notes reflected another critical use case for the D.C. agency. Some case workers, while they are in and out of their cars for meetings, prefer to write their contact notes on paper. AI-enabled contact notes allow busy workers to take a quick snapshot of their written notes, and have AI generate typed summaries required for the child welfare information system—cutting tremendous amounts of time and energy typing these notes. Double data entry is eliminated. The agency is now testing this with a group of social workers, and it has generated a great deal of interest.

Workers can also dictate notes into their agency-issued device, record them, and AI will generate typed and cleaned up, professionally formatted contact notes—ready to be validated and edited by the worker, before being ingested into the child welfare information system rapidly and seamlessly.

Again, this approach can save time. The agency realized that contact notes represent the lion’s share of the documentation needed in the high turnover job of child protective investigators facing “huge caseloads” in Washington, D.C. Adding an investigation summary capability to rapidly document and transcribe contact notes could be a game changer, and that could be replicated across the United States.

The team also developed an AI-driven service request assistant that scans court orders and automatically drafts court-ordered service requests, helping workers keep up with the myriad of required documentation and tasks needed. Given the complex needs of families in the city, time is of the essence and AI can help.

Other pilots focused on matching children and families to services—from therapy to housing—using data-driven recommendations, while keeping human decision-making intact.

Marina Havan, from the Washington, D.C. team, pointed out that the D.C. child welfare system is excited about implementing multiple use cases developed by social workers, and focused on a return on investment. Havan advised that this needs to be a cost vs. benefits analysis, and sometimes it is difficult to ask for money for AI projects based solely on the metric of saving time.

It was also highlighted that implementing AI use cases is not simply about identifying the best use cases. This also involves incorporating change management as part of a holistic approach to AI adoption.

Attribution: Case study provided to the AI Navigator courtesy of the IBM Centre for The Business of Government, Apolitical's content partner. “AI in State Government: Balancing Innovation, Efficiency, and Risk”, https://www.businessofgovernment.org/reports/ai-in-state-government, IBM Centre for The Business of Government.