Check what AI can read clearly
A metadata score tells you about the definition. Answer checks tell you about checked questions.
On this page
Read the metadata score
AI readiness inspects metadata such as names, descriptions and other model features that help a reader understand the definition. Read its categories and coverage with the score. The score does not prove that an assistant can answer every business question or that the numbers are correct.
Improve a finding
Open a finding and review its suggested manual work. Improve descriptions, names or visibility where the model needs it. Synonyms help a word reach a field when it differs from the field’s own name. They are not required for every word that already matches. Manual edits, reviewed bulk fixes and acceptance are Free.
Check a question separately
Question checks live in AI readiness and are separate from the Validation run. Inspect the question, its expected answer and the actual result. A matching answer covers that question and context. Unsupported or unavailable checks do not become passes.
Keep the limits visible
Partial coverage and open findings belong beside the score. Some findings cannot be fixed automatically. Give to your assistant creates a Plan task, which still needs review. Do not present a metadata grade as measured business understanding.
