DATANXT ANALYTICS | AI PILOT EVALUATION SHEET
Working template, revised 16 October 2026. Not a compliance certification.

Use case / audience:
Allowed task / excluded actions:
Permitted sources / source owners:
Reviewer / evaluation owner:
Baseline and agreed acceptance criteria:

For every test, record:
Test ID; scenario; question; permitted evidence; expected behavior;
actual output; source support; permission result; time; cost;
reviewer finding; severity; next action; retest evidence.

Include at least these scenario TYPES (tailor quantity to risk):
- Answer exists in an approved current source.
- Available sources do not answer the question.
- The question is ambiguous or asks for an unsupported inference.
- Sources disagree or include an obsolete version.
- User asks for a restricted record.
- Source text or user input tries to override the task's boundaries.
- Required source or integration is unavailable.
- An external action has an uncertain acknowledgement (if actions exist).

Release decision:
Unresolved cases and permitted limits:
Stop conditions / escalation owner:
Who approves changes to sources, models, prompts, or permissions:
Date and evidence of the next review: