Session (CTI)

Issue: Agentic AI governance in evidence synthesis usually focuses on what could go wrong if teams adopt agentic AI too quickly. That caution is necessary, but incomplete. HEOR, medical information, and market-access teams also face risks from delay: systematic reviews become stale, comparator landscapes shift, guidance and reimbursement dossiers age, and teams may duplicate manual work while informal agentic AI use spreads outside controlled workflows. At the same time, formal agentic AI guidance often lags the systems it is meant to govern. Many recommendations still fit prompt-based assistance or static tool validation, while current workflows increasingly use retrieval, agents, audit trails, local validation sets, and structured human adjudication. EU JCA and IRA drug-pricing timelines make timely, defensible comparative evidence more urgent. The debate is not acceleration versus caution. It is how to govern agentic AI assistance against the real baseline of slower reviews, outdated evidence, and delayed decisions. Overview: This panel will examine the governance models organizations can use while formal agentic AI guidance catches up. The discussion will focus on practical decision-making: which evidence tasks can move now, which require stronger controls, and what documentation makes agentic AI-supported work defensible to medical, HTA, regulatory, and market-access reviewers. The panel combines perspectives from a technology builder, pharma medical information, evidence-production operations, and an HTA-methods moderator. The aim is to help attendees distinguish unmanaged adoption from controlled, risk-tiered use of agentic AI in evidence workflows.