Leveraging Artificial Intelligence to Standardize Housing Interventions: A Participatory Model Integrating Practice Wisdom, Lived Experience, and Policy

Background and Purpose:
Despite decades of investment in housing and homelessness services, inconsistency across interventions continues to hinder outcomes, especially for people with serious mental illness (SMI). Arizona’s Medicaid agency partnered with academic and community stakeholders to address this gap through the development of statewide Standard Operating Procedures (SOPs) across six key housing interventions: Prevention, Outreach, Shelter, Recovery Housing, Rapid Rehousing, and Permanent Supportive Housing. This project leveraged artificial intelligence (AI) alongside participatory methods to integrate on-the-ground knowledge, existing policies, and service protocols into a cohesive SOP framework. The broader aim is to align fragmented housing services with public health goals, advance equity, and support scalable policy reform.

Methods:
The project involved three iterative phases: stakeholder engagement, AI-enhanced synthesis, and implementation planning. Over 200 stakeholders from across the state—including service providers, agency leaders, and individuals with lived experience—participated in workshops designed to surface core practices, barriers, and recommendations across six SOP focus areas: service requirements, program workflows, safety and crisis response, community integration, outcome measurement, and workforce development. Simultaneously, the research team used social work science and AI tools to systematically collect, analyze, and synthesize facts and truths about select housing interventions from documents, meeting notes, and open-text surveys. ChatGPT Enterprise was deployed to cluster themes, identify patterns, and draft SOP language efficiently while preserving stakeholder voice and intent.

Results:
The co-developed SOPs provide a structured, flexible framework tailored to Arizona’s diverse homeless service environments. Lived expertise directly shaped language, workflows, and expectations—particularly around equity in access, trauma-informed care, and cross-sector navigation. Participants emphasized the importance of culturally responsive, flexible guidelines that remain sensitive to rural constraints, staffing limitations, and funding variability. AI considerably enhanced project efficiency and outputs by reducing the time required to consolidate stakeholder input, providing rapid turnaround of draft materials for provider validation, and pinpointing thematic gaps that might have otherwise been overlooked. Resulting SOPs are aligned with the Arizona Health Improvement Plan (AHIP) and continuum-of-care strategic goals, positioning them for integration into statewide contracts and policy levers.

Conclusions and Implications:
This project presents a scalable model for policy-informed, community-centered SOP development, where AI and human expertise are complementary. The integration of lived experience alongside AI-enabled analysis yielded SOPs that are both technically robust and experientially grounded. The approach enhances transparency, accelerates documentation processes, and builds credibility among frontline practitioners. As public systems face increasing pressure to demonstrate value, equity, and efficiency, this method offers a replicable blueprint for transforming siloed service ecosystems into coordinated, outcomes-driven networks.