Field notes
AI evaluation, without the demo-day optimism.
Short, practical guides for teams shipping agents and RAG into revenue workflows. Every guide ends with the same question: what evidence would survive an incident, a board review, or an auditor?
Agent reliability
AI agent evaluation: the production gap
Why an agent can pass your eval suite and still fail a customer—and what to measure instead.
Read the guide → RAG reliabilityRAG evaluation: the 12-check pre-production review
A practical checklist for retrieval quality, groundedness, refusal, cost, and regression testing.
Read the guide → AssuranceWhat AI assurance actually means
A plain definition, the evidence a board or auditor expects, and where most teams stop short.
Read the guide →