Policy & Practice | Fall 2024

technology speaks

By Steve Roth

Artificial Intelligence in Action: Key Lessons Learned Over Eight Years in Production

T rust but verify. These three simple words are critical when it comes to using artificial intelligence (AI) in human services. They have also provided the foundation upon which Northwoods built our flagship software, Traverse, since its inception in 2018. Our journey developing Traverse has resulted in many lessons that agency leaders, fellow software providers, and other key stakeholderås can learn from to move forward responsibly while preserving the core values of human services. HowTraverse Leverages AI The case discovery module of Traverse leverages several different AI tools, including a large language model (LLM), natural language processing (NLP), automatic speech recognition (ASR), deep learning, and machine learning. Through this combination of tools, Traverse can read, understand, and turn all the content it is provided into searchable, discoverable text. This includes text from case notes and nar ratives, text from casefile documents and photos, and text transcribed from audio and video files. Case discovery provides agencies with a variety of best-of-breed features to help workers, supervisors, and administrators discover insights about the case file, including: n People mentioned in case documentation and the relationship between them n Events, including placement changes, medical events, legal events, and others found when dis playing the case in a timeline view

n A single content item summary to get a quick snapshot of large items (like long medical and psychological records) n A full case summary that provides an overall snapshot of what has happened in the case

n Drugs mentioned in the case, including medical and nonmedical

Traverse also supports workers through a chatbot, Ada, our Agency Digital Assistant. Administrators use Ada to help workers get quick, accurate answers about agency policies, laws, provider information, or anything else that they want

n Objects found in photos in the case (injuries, weapons, drug paraphernalia)

Illustration by Chris Campbell

Policy & Practice Fall 2024 34