Operating ProcedureExecution layer

AI Screening Calibration (QA)

A checklist with 6 steps: Define the pass-through rate for AI pre-screening before launch.

By InnovaAI ResearchPublished

What are the steps?

checklist

AI Screening Calibration (QA)

  1. 01

    Define the pass-through rate for AI pre-screening before launch

    Set a target percentage of candidates that should advance to human review, based on historical quality data from the client's past hires.

  2. 02

    Run a shadow test on the last 50 closed roles

    Feed past candidate resumes through the AI screener and compare its shortlist against the candidates who were actually hired, noting any mismatches.

  3. 03

    Review AI screening decisions for demographic bias signals

    Check whether the screener disproportionately filters out candidates from certain backgrounds, using the client's own diversity data as a baseline.

  4. 04

    Interview the hiring manager on shortlist satisfaction

    Ask whether the AI-selected candidates were a better fit than the previous manual shortlist, and document specific examples of misses.

  5. 05

    Adjust screening thresholds based on the shadow test results

    Lower or raise the AI's scoring cutoff until the pass-through rate aligns with the target, and re-run the shadow test to confirm the change.

  6. 06

    Document the calibration settings and rationale for the client

    Provide a one-page summary that lists the threshold, the bias check results, and the shadow test outcomes, so the client can audit the process.