Screening job applicants

Task boundary

Screening job applicants is assessed as a specific task and workflow, not as a general endorsement of AI for the surrounding job. The v1 product boundary excludes AI influence over employment screening.

Assumptions

• The proposed AI output would rank, filter, summarize, score, or recommend applicants.
• That output would influence who advances in an employment selection process.
• The v1 house boundary applies regardless of whether a person makes the final click.

Decision

• Disposition: Do not use AI
• Pattern: Human-only
• Inherent risk: critical

Rules that drive the decision

• Employment screening is an explicit all-AI-influence exclusion in the v1 methodology.
• A disclaimer, human approval, or lower autonomy cannot override that exclusion.
• This is a conservative house-policy boundary, not a claim that every jurisdiction universally prohibits all AI use; the U.S. Department of Justice source describes a relevant U.S. employment context.

Why alternatives were rejected

• Ranking or summarizing candidates still shapes the prohibited decision.
• Controls do not convert an all-AI-influence exclusion into a proceed result.

Next safe step

• Keep applicant comparison, ranking, summarization, and advancement decisions human-only.
• If exploring AI for unrelated administrative work, define a separate task that cannot influence selection and run a new assessment.
• Have qualified employment and legal owners confirm the process and jurisdiction-specific obligations.

Reassess the changed task

• Only a genuinely separate administrative task may be reassessed; renaming or adding approval to screening does not change the exclusion.
• Revisit this published boundary only through a reviewed methodology update, not an operator exception.