Data-First, Targeted AI Use: Key Takeaways From the 2026 APHSA NHS Summit

The 2026 APHSA NHS Summit in Arlington, VA, played host to a number of new conversations on Artificial Intelligence (AI) in health and human services. These conversations seemed to focus more on the practical than the hypothetical, an apparent shift from the futuristic scenario spinning regularly seen in the early days of AI experimentation and deployment. 

Conversations centered on forward-thinking agency uses, targeted AI deployment, mitigating caseworker burnout and offloading manual processes.

Practical Uses for AI in Health and Human Services

Among the case studies at this year’s event were those asserting that, more than it can as a generic chatbot, AI can be of immense operational value when narrowly targeted to fix distinct issues. A few examples of these dividend-paying uses include: 

• Note Entry: Implementing tailored tools like “Beam Notes” within Child Protective Services achieved an incredible 51% reduction in narrative entry time, turning hours of manual dictation into hyper-accurate case records.
   
• Records Processing: In Adult Protective Services, AI solutions scanned, uploaded, summarized, and cited dense, physical medical records in under an hour—a workflow that previously drained over 8 manual hours.
   
• User Guidance and Troubleshooting: Some state SNAP programs leveraged Large Language Models (LLMs) to scan highly irregular, blurry, or low-quality photos of pay stubs uploaded by users, providing immediate, real-time corrective feedback if more information was needed.
   
• Legacy System Translation: Agencies report utilizing AI as a modernization bridge, helping software developers translate legacy, 1980s “green-screen” computer code into modern infrastructure that can seamlessly interact with current APIs.
   
• Safer Service Training: Organizations are leveraging conversational, voice-enabled training simulators—essentially “flight simulators” for human services—to let new adjudicators safely practice trauma-informed communication skills with mock clients.

The Importance of a Data-First Approach

Success stories at this year’s event inspired discussions on the data precision and governance needed to achieve results. Without quality data, those at the summit held, AI tools might perform suboptimally or, worse, produce hallucinations. However, in an attempt to avoid the risks posed by poorly-sourced AI tools, agencies might consider any number of the following precautions:

• Ensuring Reliability Before Implementation: Centralize and verify workforce and constituent data before plugging in predictive or generative tools.
   
• Prioritizing Low-Risk, Internal Use First: Consider having initial deployments focus on internal operations (e.g., summarizing case notes, staff training, policy searching) rather than automated eligibility or client-facing applications.
   
• Strict Privacy Protocols: Safe AI deployment relies heavily on robust data governance, including data hashing to safely link records without storing Personally Identifiable Information (PII).

At Equifax, we understand the delicate balance between technological innovation and human responsibility. We are excited to partner with state agencies as they navigate this path, helping to build the secure, compliant, data-automated, and deeply human infrastructure that the future demands. Together, we can ensure technology serves as a powerful accelerator for your mission: taking care of the vulnerable community members who need it most.