Evals for Beginners: How to Test Your AI Agents скачать в хорошем качестве

Evals for Beginners: How to Test Your AI Agents

Your AI agent writes LinkedIn posts. Classifies support emails. Answers customer questions. But how do you know it's doing a good job? Most people rely on spot-checks and gut feeling. That's fine for small projects and POCs. Not fine for production systems handling real business work. Evals (evaluations) are systematic tests that measure AI quality, reliability, and performance against specific criteria. They let you catch mistakes before customers do, test prompt changes without breaking existing workflows, and quantify improvement over time. This workshop shows you how to build them. Jan will be joined by [Marcel Claus-Ahrens](  / marcelclausahrens  , n8n Ambassador (a.k.a Dr Pure Eval), who'll walk through what evals are, when to use them, and how to implement them in n8n. Marcel will demo a production eval system, then guide you through building your own from scratch. *You'll learn:* • What evaluations are and why they're becoming critical for AI systems • The main types of evals • When evals are worth the setup time (and when they're overkill) • How to create test cases when you don't have historical data • How to build and run evaluations in n8n *Who this is for:* Anyone building AI agents or workflows for real business use. If you're wondering whether your AI is reliable enough to run unsupervised, this session gives you the tools to find out. This 90-minute session includes a full demo, hands-on build, and time for Q&A.