Evaluating AI agents in production
What breaks when agents meet real tools, and how to test for it before launch.
Insights
Written from engagements, not press releases
Three series are in preparation. Articles appear here as they are published.
A buyer's guide to model evaluation
How to read an eval report, and the questions that expose a weak one.
Test automation that survives handover
Why most vendor-built frameworks die within a year, and how to build one that does not.
AI red teaming: a practical checklist
The campaign structure, harm taxonomy and retest loop we run before any launch.
RAG evaluation metrics that matter
Retrieval precision, answer faithfulness, and the gates that keep a pipeline honest.
Performance testing for seasonal peaks
Building load models from production traffic, and fixing what they expose in time.