AI can help health and human services in several ways, from surfacing information insights at scale to providing information on public health issues proactively to improving data management. Some governments are exploring AI to address workforce shortages, which could support the human services sector.
“We know that across every state, there’s a crisis in staffing in human services,” said Gregory Kwaczala, vice president of strategy, equity and advancement for Youth Research Inc. (YRI), the nonprofit affiliate to the New York Office of Children and Family Services (OCFS).
AI can help address this staffing issue by increasing caseworkers’ efficiency and providing them coaching to improve their job performance, Kwaczala said. He noted that state and local governments are limited in their capacity to provide this individualized support, which is where an AI tool — Lyssn, in the case of OCFS — can fill the gap.
YRI is charged with implementing evidence-based practices across the state, providing staff support for OCFS and supporting the fidelity monitoring aspect of MI, which can be a massive lift based on the number of caseworkers across the state, YRI Director of Research and Policy Erica Thompson said. Implementing an AI tool into the process filled the gap in a scalable way. In addition to creating training opportunities for employees, the tool provides useful quality assurance metrics.
There are risks regarding AI, however, and in the health and human services area, there are unique data privacy concerns; for YRI, it was important to consider what data was being collected.
Notably, Lyssn allows users to decide what happens to their data, its Chief Technology Officer Mike Tanana explained. This can include how long the data is stored and whether it is able to be used to train models. Agencies using Lyssn to train staff can have it analyze real client conversations — but YRI opts to use simulated scenarios for training, to preserve client confidentiality and security.
Stearns County, Minn., has also been exploring how AI can support its Family and Children Services Division’s work; the county is already using AI in some ways and is exploring adopting it for other purposes, the division’s Human Services Director Nicholas Henderson said.
Minnesota has a state-run, county-administered human services model. This means the state provides guidance and funding, and the county has specific performance measures it is required to meet for quality assurance — which Henderson described as a huge focus, as the state works to implement MI within the child welfare system.
The state has a partnership with Lyssn, which will impact all child welfare agencies including those at the county level; Henderson said the process of implementing this technology is underway.
In the area of child welfare, Henderson said the county is already starting to explore how generative AI technology — such as ChatGPT — can assist in the creation of safety plans. To protect data privacy, the county currently only uses these tools to create an outline, without inputting any sensitive data, which can be added afterward. What used to take staff two hours is now taking less than 30 minutes, Henderson noted.
A jurisdiction’s AI policy must be considered, as well, to ensure AI adoption or new use cases comply with its rules; the county’s policy allows some products and use cases, but not others, Henderson said, to mitigate risk around unassessed use cases or products.
A fourth of the county’s residents receive public assistance, so the county is implementing AI to meet call demand, support staff and improve customer service in its Gateway Services Division — adjusting its AI policy to strike a balance between compliance and use.
AI is already being used by other county platforms such as Microsoft Teams, which has AI technology built into its meetings for transcription purposes.
The county will protect clients’ data while increasing efficiencies with AI, Henderson said: “And hopefully for us, those efficiencies will mean that we’re doing greater work in the community versus spending time on our community computers doing data entry.”
In Wyoming, its Department of Family Services uses Lyssn’s AI for fidelity monitoring, and officials have recently decided to expand its use to enable staff to segment large data sets by district. This replaces a manual process, said Maria Vasquez-James, a department social services policy analyst.
Confidentiality laws regarding the child welfare system were a primary consideration before any AI was adopted, Vasquez-James said. Families had to sign a release form before their recordings could be uploaded into Lyssn.
She likened adopting AI in spite of concerns about the technology to exposure therapy; the idea can be scary, but familiarizing people with the technology has significantly reduced concerns.
Now, the data the platform provides informs state MI training. And because conversations with families often occur in noisy environments, this technology can be very impactful in their understanding, especially with something “as complex as human language,” she said.
Emily Smith Goering, Lyssn’s senior implementation success manager, is a former social worker who directly interacted with child welfare agencies. Health and human services’ environments are very demanding and often offer limited support in terms of professional training, she explained. AI, however, can provide feedback and coaching that may not otherwise be available.
A big part of working with government agencies, she said, is educating them on the security of the tool, AI’s benefits and the company’s approach to AI bias — on which it provides an annual report.
And while the ultimate goal is to provide agencies with the workforce development information that can help them better support their clients, Lyssn data also helps states report to the federal government on their use of evidence-based treatments, Tanana said.
But humans power the technology and the services it benefits, Smith Goering emphasized. AI, she said, will not remove the human element from the human services field: “Not only are we not trying to replace humans, we’re trying to augment what they do and support what they do.”
