Building an AI Committee at Your Hospital

AI governance committees are becoming mandatory at academic medical centers and large health systems. This episode covers how to build one that makes real decisions instead of producing reports nobody reads.

Committee Composition

• Clinical champion (physician who uses AI daily, not just an executive sponsor)
• CMIO or clinical informatics lead
• Compliance officer with HIPAA and regulatory expertise
• IT security representative
• Legal counsel familiar with AI liability and vendor contracts
• Quality and patient safety officer
• Nursing informatics representative
• Patient representative (for transparency and trust)

Decision Framework

• Tier 1 (low risk): AI tools that do not touch patient data and do not influence clinical decisions — can be approved by department heads
• Tier 2 (moderate risk): AI tools that process patient data or assist clinical workflows — require committee review, BAA verification, and a 90-day pilot
• Tier 3 (high risk): AI tools that generate diagnoses, treatment plans, or patient-facing content — require full committee vote, legal review, and ongoing monitoring

Common Mistakes

• Committees that only review tools after they are already deployed
• No sunset or re-review process — tools approved once stay forever without reassessment
• Over-representation of IT and under-representation of frontline clinicians
• Confusing vendor security certifications with clinical safety validation