LESSON 1 OF 6 · ADVANCED
Choose a narrow, useful task
Begin with a specific task and define what a good output looks like. Drafting a generic interview agenda is different from ranking real candidates. Higher-impact uses require more scrutiny, appropriate expertise and organizational approval. NIST’s framework is voluntary guidance for managing risk; it is not an employment-law certification or a guarantee of safe use.
Classify three possible uses by impact, data needed and human decision owner.
About 2 minutes reading + 12 minutes practice. Take longer when needed.
Apply the method
Define the task, intended user, input data, decision impact and acceptable output before selecting a tool. Distinguish drafting or organizing information from evaluating a real person’s eligibility. Higher-impact uses need stronger expertise, oversight and organizational authorization. Begin a pilot with fictional, low-risk information and a review standard. NIST’s voluntary framework supports risk-management thinking; it is not proof that a particular product is safe, lawful or suitable for employment decisions. Document the reason to use a tool and the conditions under which the team would not use it.
A worked application
Fictional example: A recruiter tests AI assistance to draft a generic intake agenda for a fictional service role. A reviewer checks relevance and factuality. That trial does not authorize feeding real applicant résumés into the same service or automatically rejecting people. Those uses involve different data and consequences and require separate assessment.
Common mistake to avoid
Do not let a successful low-risk drafting demonstration become blanket approval for high-impact candidate decisions.
Work through the case
FICTIONAL SCENARIO
A tool can help brainstorm role-intake questions from fictional information. Automatically rejecting applicants based on an opaque score has different consequences.
Make it practical
Classify three possible uses by impact, data needed and human decision owner.
Compare your reasoning with a worked response
Keep candidate decisions with authorized people and obtain appropriate review before deploying higher-impact systems. A useful pilot begins with low-risk, non-sensitive inputs.
Takeaway
Use the method to explain your decision, identify what you still need to verify and choose an accountable next step.
Treat the response as a model for reasoning, not the only possible wording. Adapt it to the role, evidence and appropriate organizational guidance.