Can Chat-GPT Style AI Help Us Understand Emergency Shelter Data?

Canadian Alliance to End Homelessness Conference, October, 2025.

For program design and evaluation, it’s important to understand how large groups of people interact with the housing and homelessness system. Much of this interaction is recorded in case notes form. Since a major shelter can record thousands of case notes a year, it’s not practical to read all these notes to understand the needs and experiences of all the people accessing care.

This talk explores using natural language AI to read a series of case notes and to determine if each case note contains a particular theme (ie. employment, mental health challenges, drug poisoning, etc.). Human decision makers can use these theme labels to extract notes related to a particular topic or simply see the prevalence of each theme to better understand housing and homelessnes system use at the population level.