Session

Artificial intelligence is advancing rapidly across healthcare, but broad strategies and principles alone do not resolve the practical questions organizations face when implementing AI. Health plans and health systems must determine whether AI is appropriate in a specific use case, how benefits and risks should be weighed, what data and infrastructure are required, how workflows and people will be affected, , and which outcomes demonstrate value. Responsible AI implementation depends on context. The intended purpose, users, data, workflow, level of human oversight, and potential consequences differ by use case.   

This session will use NCQA's broader AI strategy and learning collaborative model to frame a cross-sector discussion about implementation. NCQA will explain why it is bringing stakeholders together in structured, peer-to-peer, environments around defined use cases and share how risk-benefit criteria can support more disciplined evaluation. AHIP will address the health plan perspective, including the importance of data readiness, foundational capabilities, and implementation considerations. The Duke Institute for Health Innovation will share real-world health system experience, including how implementation choices, workflow integration, and outcomes measurement affect adoption.  

Together, the panel will demonstrate why cross-collaboration among quality organizations, health plans, and health systems is essential to building a more practical and evidence-informed approach to responsible AI adoption.  

Learning Objectives:

• Explain why AI implementation should be evaluated at the use-case level, using intended purpose, data requirements, workflow integration, and potential benefits and risks. 
• Demonstrate how cross-stakeholder collaboration enables organizations to build on shared AI implementation lessons, avoid duplicating efforts, and how these insights are applied through NCQA’s inaugural learning collaboratives. 
• Identify foundational data and operational prerequisites, including data quality, infrastructure, governance, workflow ownership, and measurement readiness. 
• Introduce the idea of applying a risk-benefit lens to an AI use case, considering expected value, affected stakeholders, potential harms, safeguards, human oversight, and evidence. 
• Apply use-case evaluation, collaborative learning, implementation readiness, and risk-benefit considerations to inform responsible AI implementation decisions.