AI-powered features are becoming ubiquitous, but how do you test them when responses are non-deterministic? Traditional assertions fall short with LLM integrations. This talk explores practical strategies for automated testing of AI features.

We'll tackle unique AI testing challenges: managing non-deterministic, validating safety guardrails against harmful outputs and PII leaks, and observability practices for debugging AI test failures.

Walk away with concrete patterns for building confidence in your AI features without breaking the bank or sacrificing reliability.