Glossary

AI evaluation language without hidden meanings.

Use these definitions when reading WAO pages, samples and controlled-beta materials.

AI Health

A decision-support view of evidence-backed dimension states, confidence, limitations and next actions. It is not a certification.

A specific provider, model, operation, account, region and evidence combination—not a broad provider logo.

How strongly the available representative evidence supports an interpretation. A score without adequate evidence can remain Needs Data.

Manually governed access with assigned capability, quota, support and evidence boundaries.

The period and exact test conditions to which an observation applies.

Accountable judgment using an explicit rubric, comments and adjudication where automated evidence is insufficient.

The form of input or output, such as text, image, document, audio, video or tool action.

A customer-controlled secret used only when an approved WAO workflow executes against that provider account.

A safe set of tasks and expectations that reflects the real decision better than a polished demo.

The consistency of successful operation under the tested conditions, including failures and recovery.

An accountable decision supported by defined evidence and open limitations—not a universal numeric pass mark.

Fictional data used to demonstrate a workflow or report structure. It is not a provider benchmark or customer result.