Reliability in agentic systems is not just about the model. It is about what happens when a tool call times out, when an agent loops, when a retrieval step returns garbage, and when a compliance checkpoint needs a human in the loop. In this session, Rajat Mehta, Morgan Stanley, will cover the orchestration patterns, error handling strategies, and fallback logic that keep agentic pipelines running in environments where downtime and wrong outputs both carry a cost. What this session will cover: - Orchestration patterns for multi-step agentic pipelines in financial workflows - How to handle tool-call failures, agent loops, and bad retrieval gracefully - Where to place human-in-the-loop checkpoints without killing iteration speed - Fallback strategies that maintain compliance requirements when agents go off-track