Industry and health‑care witnesses describe productivity gains and narrow use cases for municipal AI

Industry and health‑care witnesses told the committee that AI can deliver administrative productivity gains and service improvements when deployed with human oversight, robust privacy protections and external evaluation.

Ravi Chawla, vice president and chief analytics officer at Independence Blue Cross, said IBX uses AI to predict health complications, reduce emergency room usage, personalize care management and automate administrative tasks. Chawla described internal governance (an AI center of excellence, an AI ethics committee) and regulatory compliance (HIPAA) and said IBX requires pre‑ and post‑implementation testing, member consent where appropriate and continuous monitoring for clinical uses.

Sorelle Friedler, a computer‑science professor and former White House official who helped draft the AI Bill of Rights, told the committee that AI’s statistical errors require active human review; a system with 98% accuracy still produces thousands of errors at city scale. She advised limiting generative AI in high‑stakes public decisions and favoring narrowly tailored, tested machine‑learning systems for administrative triage tasks. Friedler recommended a public inventory of AI systems, yearly updates at minimum and external audits for high‑risk deployments.

Andrew Jenison of CGI urged the city to adopt a responsible AI framework emphasizing human‑in‑the‑loop design, privacy and security, transparency of outputs and environmental sustainability. All industry witnesses supported stronger procurement clauses, ongoing monitoring and the use of external evaluators for high‑risk systems.

Panelists said generative AI can accelerate document drafting and developer productivity (GitHub Copilot was cited as producing 30–40% productivity gains for engineers in some contexts), but they warned that generative outputs require human verification to avoid ‘‘hallucinations’’ or factual errors. They recommended contract clauses that require vendors to disclose added AI features, to provide logs and to permit third‑party audits for systems that affect benefits, public‑safety outcomes or access to services.

The panelists’ testimony aligned with administration witnesses in urging human oversight, transparency and careful procurement language, while offering industry and health‑care examples of potential benefits when safeguards are in place.