“You Need to be Super Careful...”: Service Providers’ Perspectives on AI Use In Case Documentation

Background and Purpose: The rapid advancement of artificial intelligence (AI) captures considerable attention to its applicability in social and health services. Recent studies have examined the use of AI in social work case management (Li et al., 2026) and clinical documentation in healthcare (Bracken et al., 2025). Case documentation is closely tied to clients’ service receipt and contains identifying information and clinical narratives often used to justify professional decisions, such as service eligibility and treatment planning. As AI is increasingly adopted in social and healthcare documentation, understanding the uptake and perceptions among service professionals becomes imperative. This presentation explores social service professionals’ perspectives on and experiences of using AI in case documentation.

Methods: Data were drawn from a larger ongoing Community-Based Participatory Research project exploring how trafficking survivor narratives are created and used in clinical decision-making by anti-trafficking social service providers in the U.S. Semi-structured interviews were conducted with 18 service workers (e.g., counselor, case manager, and clinical supervisor) providing direct/clinical services to trafficking survivors across eight states in the U.S. The interview guide included a question exploring participants’ perspectives and experiences regarding AI use in clinical case documentation. Their responses to this particular question were analyzed by two team members using reflexive thematic analysis (Braun & Clarke, 2022); preliminary codes and finalized themes were refined through discussions among two primary coders and consultation with the research team.

Results: Many participants reported that the use of AI (e.g., Large Language Models (LLM) like ChatGPT and Gemini) for case documentation was prohibited within their organizations, primarily due to privacy and confidentiality concerns. Most agreed with these restrictions, noting that AI tools lack clear legal protections and pose risks of exposing sensitive client information. Participants who did report using AI for documentation generally preferred Gemini over ChatGPT for its enhanced privacy protection features. Organizations that use Electronic Medical Records (EMR) systems may opt for built-in AI features; however, the significant cost presented a financial barrier. Despite these concerns, most participants acknowledged AI as a useful tool for improving efficiency and enhancing the professional writing quality of case documentation. Some, however, emphasized that AI should be used sparingly and only when necessary. Finally, participants, especially those in supervisory roles, expressed a pragmatic view that a complete restriction of AI use might not be possible.

Conclusion and Implications: Findings reveal reservations among service workers about AI use in case documentation primarily due to privacy and confidentiality concerns. However, service providers are aware of and receptive to the potential benefits of AI in case documentation. Given the rapid advancement and growing prevalence of AI across social service sectors particularly in case documentation, social work researchers and service providers should proactively engage diverse stakeholders and service user groups in constructive dialogue around ethical AI use in clinical practice and documentation, develop training that promotes ethical and client-protective AI use, and advocate for the development of AI tools that incorporate robust privacy protections and adhere to established ethical guidelines.