AIM & IBM CXODiscussion 2026

The third table session establishes that agentic AI demands a fundamental transition from model-level governance to comprehensive system governance. Because agents execute actions rather than merely generating text, organizations must enforce pre-deployment governance across centralized platforms, manage explicit agent identities and permissions, and establish continuous monitoring with automated circuit breakers. To scale without operational bottlenecks, leaders propose risk-based human oversight, reserving manual intervention for high-threshold or transactional tasks. The framework prioritizes full auditability, decision replayability, foundational master data quality, and model-agnostic architectures. Integrating governance across the AI SDLC ensures that permissions, approval checkpoints, and audit mechanisms are designed, tested, and continuously validated as agentic systems evolve. Crucially, accountability remains with business process owners, supported by standardized AI "fact sheets" and centralized agent catalogs to maintain visibility and enterprise control.