Executive summary

A method for finding out whether an AI answer service can be trusted to cite the right evidence, before anyone relies on it.

What is inside

What the baseline diagnostic covers

The asset walks a diagnostic lead through the question inventory, source inventory, question set, the three measures, access tests, failures, claim grading, and sign-off. It ends with a diagnostic questionnaire and an evidence inventory.

  • Inventorying questions and where they arrive
  • Inventorying sources and their approval status
  • Building a representative question set
  • Measuring citation accuracy, retrieval relevance, and acceptance

Create a free account to view the full page