AI and Agent Observability: Tracing Failures in Production

What this session covers

• What organisations actually need to monitor across models, agents, tools, workflows and downstream actions.
• How teams identify whether a failure originated in the model, the underlying data, an integration or the wider workflow.
• Identity, permissions and controls for non-human systems acting without a human session.
• Building audit trails that support investigation, governance and regulatory requirements without creating unnecessary monitoring overhead.