An HR adviser at a local council is drawing up a shortlist of ways the HR team could use Claude. The council's AI-use policy allows AI to draft and summarise material, but states that every decision about an individual member of staff or applicant is made by a person. Which use cases belong on the shortlist? Select TWO.
- ARanking job applicants against the person specification and rejecting those below a set score
- BDrafting job adverts from approved role profiles for a recruiter to review and edit before posting Correct
- CDeciding the outcome of disciplinary cases by comparing investigation notes with past outcomes
- DSummarising anonymised staff survey comments into themes for managers to discuss and act on Correct
- EMarking probation reviews as pass or fail from managers' notes so HR can close them faster
Why A is wrong: Screening large applicant pools is a tempting time saving, but rejecting candidates on Claude's score is a decision about individuals, which the policy reserves for people and which carries bias and fairness risk.
Why B is correct: Correct. This is drafting from approved source material with a person reviewing before anything is published, which the policy permits and which saves real time on repetitive writing.
Why C is wrong: Consistency with past cases sounds fair, but the outcome of a disciplinary case is a decision about a person. The policy requires a person to make it, and past outcomes may carry their own bias.
Why D is correct: Correct. Summarising a large volume of free-text comments is a strong fit for Claude, the comments are anonymised, and managers keep the decisions about what to do with the themes.
Why E is wrong: Speed is appealing, but a probation pass or fail decides whether someone keeps their job. Letting Claude make that call breaks the policy even if a manager wrote the underlying notes.