LLM apps and agents are non-deterministic. Small changes to prompts, tools, or models can shift quality, safety, or behaviour in ways you often only notice after shipping. In this session we build a practical evaluation workflow using Microsoft Foundry Evaluations and ASSERT. We start in Foundry by evaluating an agent, a model, or a dataset using built in and custom evaluators. We focus on the evaluator types you actually need in production such as agent behaviour, quality, and safety, and show how to interpret results as a scorecard you can reuse for regression testing. Next, we extend this workflow with ASSERT, Microsoft's open-source framework for systematically identifying risks, generating evaluation taxonomies and test cases, and continuously validating agent behaviour against organisational requirements. You'll see how ASSERT helps teams move beyond manual prompt testing and generic benchmarks by creating application-specific evaluations for scenarios such as prompt injection, hallucinations, unauthorised actions, and information leakage. We also show how ASSERT complements Foundry Evaluations by helping teams build broader and more meaningful test coverage for agent-based applications. Finally, we connect evaluation with observability through Foundry tracing. You will leave with a concrete way to combine evaluation gates, ASSERT-generated test suites, and tracing so you can ship agents with more confidence than manual spot checks and playground demos.