For employers
Choose the right starting point for workforce or AI-data work, then prepare a brief that makes the decision, evidence, and review path clear.
Begin with the system or workflow, its intended use, and the decision the work should support. Then describe the source material, expected output, and evidence a reviewer can use. This keeps the conversation anchored to an actual operating question instead of a broad job title or an undefined request for “AI talent.”
People and review capacity
Use this path when the central question is who should do the work. Define candidate qualification, language and domain context, calibration material, reviewer roles, and how edge cases will move through the team.
Material and quality evidence
Use this path when the central question is what material will be produced or reviewed. Define intended use, task units, rubric versions, validation samples, and acceptance questions.
The system, workflow, relevant users, intended use, and decision the work should inform.
Representative inputs, expected outputs, references, constraints, and difficult examples.
The judgment reviewers must make, the evidence available to them, and when they should escalate.
What your team will accept, revise, reject, or investigate after reviewing the results.
The AI evaluation task-scoping checklist translates the public NIST AI Risk Management Framework into questions about context, task evidence, rubric criteria, reviewer context, disagreement handling, records, and acceptance.
Share the task context, required judgment, source material, reviewer needs, and the decision your team is preparing to make.