AI systems will fail. The question is whether they fail safely.  

This session focuses on failure as a first-class design concern. We examine common failure modes in LLM and agentic systems and how teams can design for resilience, graceful degradation, and recovery when things go wrong. 

Rather than aiming for perfection, we focus on building systems that remain trustworthy under stress.  

What you will learn  
• Common failure patterns in AI and agentic systems  
• Designing fallbacks and safe default behaviors  
• Human override and escalation strategies  
• Incident response and recovery for AI in production  

This session is essential for teams running AI systems in real-world, high-stakes environments.