Template preview (excerpt)
4) Performance metrics & thresholds
- Validity metrics (fit-for-purpose)
- Fairness checks (where applicable)
- Calibration and error analysis
Evidence pointers
- Oversight records (approvals, overrides, escalation)
- Candidate disclosure artifacts + UI proof (if applicable)
View the worked example
Boundary:
- The system provides recommendations only.
- Humans make final hiring decisions.
- Always-review: any rejection recommendation with confidence < 0.7.
Before you begin
For compliance, risk, product, and ML ops teams shipping agentic Processes into regulated environments.
A system-type Annex IV template for hiring/HR screening: rankings, scoring, recommendations, and the evidence you need for oversight and fairness reviews.
It keeps the focus on controls and proof: disclosures, reviewer Processes, and traceability across versions.
When to use this resource
- Your system screens, ranks, or scores candidates or influences hiring decisions.
- You need to show oversight, explainability, and fairness-oriented evaluation evidence.
- You are preparing internal governance reviews or external audits.
Information to gather
- A clear statement of intended purpose + boundaries (especially for hiring decisions).
- Your data sources and retention windows.
- Evaluation approach and threshold governance.
- Oversight SOP and monitoring plan references.
Review checklist
Use these checks in the review with your system owner. Confirm the applicable requirements and attach evidence for the decisions your team makes.
- Outputs are explicit (rankings/scores) and boundaries are explicit (what is not automated).
- Data minimization and sensitive data handling are defined.
- Fairness-related checks and thresholds are documented (where applicable).
- Oversight rules define always-review vs sampled review and recorded rationale.
- Logs and exports can prove which model/policy version produced a recommendation.
Operating controls and evidence
- Govern
Policy-as-code checkpoints that block or require review for high-risk actions.
Versioned change control for model/prompt/policy/Process updates.
- Assure
Risk-tiered sampling reviews (baseline + burst during incidents or after changes).
Near-miss tracking (blocked / nearly blocked steps) as a measurable control signal.
- Prove
Configurable retention schedules, integrity verification, and an append-only Audit Trail.
Evidence Room export bundles (manifest + checksums) so auditors can verify independently.
Questions about this resource
Is hiring/HR screening commonly considered high-risk?
Many employment-related decision-support use cases are commonly treated as high-risk categories, but classification depends on intended purpose and context. Confirm with counsel.
What evidence matters most for HR screening?
Oversight records, evaluation results (including relevant subgroup/segment analysis where applicable), and traceability from recommendations to versions and policies.
Do we need candidate disclosures?
Disclosure obligations depend on the system and jurisdiction. Many teams include clear candidate-facing transparency and retain evidence it was presented.
How do we avoid sensitive data leaking into logs?
Use redaction rules, minimize logged data, and enforce access controls; record all export actions.
How often should we refresh the documentation?
On material changes, incidents, and monitoring findings; many teams align updates to change control cadence.
What do auditors reject?
Documentation that does not tie claims to evidence: missing review records, missing evaluation results, or missing version traceability.
Download file language: English