LensReading which lens this session carries.

Fairness readings computed on modeled data demonstrate the instrument, not a real audit result. The grid below is produced by the same code that would run against a real population; the population underneath is modeled.

Govern · people-affecting decision · HR2A

Candidate screening and shortlist

Scores applications against stated requirements and produces the shortlist a recruiter reviews. The agent does not reject; it ranks and recommends.

All decision points
People affected
8,420
Each month
Decision kind
Selection
A lower rate is the harm
Readings
30
0 below the reporting floor
Outside threshold
3
across 6 periods
Open alarms
4
4 raised in total
Appeals
5
2 changed the original outcome

Who is accountable, and what the oversight actually is

Named, with the measure written out rather than implied by a role title.

Agent
Screening Agent
Autonomy: A3
Accountable human
Recruiting pod lead, per requisition
Recruiter
Oversight measure
A named recruiter reviews every shortlist before it reaches a hiring manager and can add any applicant the agent ranked below the line. The recruiter sees the ranking rationale and the fields the agent used.
Review cadence
Monthly cohort review, quarterly threshold review
Reviewer: Nadia Kovac, Data Protection Officer, with Priya Raman
On schedule
Monitoring status
Monitored
Readings computed for every reporting period and every declared cohort.

Regulatory position

How this decision is classified and what the affected person can see.

Risk class
High risk — EU AI Act Annex III(4)(a)
Employment: a system used to filter applications and evaluate candidates in the course of recruitment. High risk on the face of Annex III.
Explanation surface
Published
A standing explanation is available to the affected person without asking.
How the outcome is contested
Talent Acquisition Lead for the region
Service level: 120 hours to a human response
Full risk record for this agent

Selection rate by cohort, by period

Impact ratio against the reference cohort. A cell with no reading means the sample was too small to compute one — it is not a pass.

CohortReferenceMar 2026Apr 2026May 2026Jun 2026Jul 2026Aug 2026
Women
Gender
Men
0.93
16.5% · n=3,705
0.92
16.1% · n=3,705
0.90
16.3% · n=3,705
0.89
15.9% · n=3,705
0.88
16.2% · n=3,705
0.87
16.0% · n=3,705
40 to 49
age-band
30 to 39
0.91
16.8% · n=1,937
0.90
15.9% · n=1,937
0.89
16.2% · n=1,937
0.87
16.3% · n=1,937
0.86
15.3% · n=1,937
0.85
15.5% · n=1,937
50 and over
age-band
30 to 39
0.84
15.0% · n=1,179
0.82
14.5% · n=1,179
0.81
14.8% · n=1,179
0.79
14.3% · n=1,179
0.78
14.5% · n=1,179
0.76
14.0% · n=1,179
Under 30
age-band
30 to 39
1.04
18.9% · n=2,273
1.05
18.5% · n=2,273
1.06
19.6% · n=2,273
1.07
19.2% · n=2,273
1.08
18.8% · n=2,273
1.09
20.1% · n=2,273
Minority ethnic groups, aggregated
Ethnicity
Majority group (per local reporting standard)
0.90
16.0% · n=2,610
0.91
17.0% · n=2,610
0.89
16.1% · n=2,610
0.90
15.8% · n=2,610
0.88
16.3% · n=2,610
0.89
15.9% · n=2,610
Within thresholdMoving but not yet outsideOutside the disparity thresholdSample below the reporting floor· means no reading exists for that period
Cohorts declared for monitoring that never produced a single reading
Gender · Non-binary and self-described — No reading has ever been produced for Non-binary and self-described on this decision. The cohort is declared in scope and the population exists, but it has never once been large enough to report against Men at the confidence this program requires.

Alarms on this decision point

Each with the action recorded against it and the person who owns it.

FAIR-0001Sustained driftMedium
170 days openAcknowledged
The ratio for Women has moved from 0.93 to 0.87 across 6 periods without ever breaching. This is inside tolerance and heading out of it, which is the reading a threshold alone would miss.
Action: Extrapolate the trend to the threshold, put a date on it, and bring that date to the next review rather than waiting for the breach.
Owner: Nadia Kovac · Data Protection Officer
FAIR-0002Never reportableLow
170 days openMonitoring
No reading has ever been produced for Non-binary and self-described on this decision. The cohort is declared in scope and the population exists, but it has never once been large enough to report against Men at the confidence this program requires.
Action: Either aggregate this cohort across decisions and periods so a reading becomes possible, or state on the record that this cohort is not monitored here and why. Leaving the cell empty is the one option that is not defensible.
Owner: Nadia Kovac · Data Protection Officer
FAIR-0003Sustained driftMedium
170 days openAcknowledged
The ratio for 40 to 49 has moved from 0.91 to 0.85 across 6 periods without ever breaching. This is inside tolerance and heading out of it, which is the reading a threshold alone would miss.
Action: Extrapolate the trend to the threshold, put a date on it, and bring that date to the next review rather than waiting for the breach.
Owner: Nadia Kovac · Data Protection Officer
FAIR-0004Threshold breachHigh
78 days openOpen
50 and over is being selected at 76 percent of the rate for 30 to 39, below the four-fifths threshold. The breach has held for 3 of the six periods.
Action: Hold the decision at the current autonomy level, re-run the cohort with the ranking features ablated one at a time to find what is carrying the disparity, and record the finding against this decision point before the next cycle opens.
Owner: Priya Raman · Chief People Officer

Appeals against this decision

Every contest recorded, with who reviewed it, how long they spent and what changed.

RefSubjectGroundsRouted toService levelHuman timeOutcomeEvidenceStatus
APL-0011
CAND-49512
candidate · US
Applicant requests human review of an automated ranking under state law.
Ben Ortiz
Talent Acquisition Lead, Americas
168 / 120 hr—PendingHR2A-1021#3Past service level
APL-0009
CAND-49330
candidate · AU
Applicant states an accessibility adjustment was not accounted for in the assessment stage.
Lin Wei
Talent Acquisition Lead, APAC
62 / 120 hr44 minOverturned
The adjustment request had not reached the assessment stage. Reassessed with the adjustment in place and shortlisted.
HR2A-1008#2Closed
APL-0006
CAND-49014
candidate · SG
Applicant requests the reason for rejection and the data used.
Lin Wei
Talent Acquisition Lead, APAC
43 / 120 hr19 minOriginal upheld
Rationale and field list supplied. No change to the outcome.
HR2A-1002#2Closed
APL-0002
CAND-48377
candidate · US
Applicant asks on what basis the application was ranked and requests human review.
Ben Ortiz
Talent Acquisition Lead, Americas
96 / 120 hr22 minOriginal upheld
Ranking rationale supplied in full. Human reviewer reached the same conclusion on the stated requirements.
HR2A-1021#2Closed
APL-0001
CAND-48120
candidate · UK
Applicant states that relevant experience held under a former job title was not recognized by the ranking.
Aisha Bello
Talent Acquisition Lead, EMEA
71 / 120 hr38 minOverturned
The former job title was not in the skills taxonomy. The applicant was added to the shortlist and the taxonomy was extended.
HR3B-1010#1Closed
How to read this page

The impact ratio is the cohort’s selection rate divided by the reference cohort’s. For a decision where being selected is the benefit, the threshold is four-fifths and a ratio below it is the alarm. For a decision where being selected is the harm, the test is inverted and a ratio above 1.25 is the alarm. Both are stated on the reading rather than left to the reader.

A reading is only computed when the cohort and the reference cohort each carry at least thirty people and at least five events. Below that, no ratio is produced and the cell stays empty. This is the single most common way a fairness dashboard misleads — a small cohort producing a comfortable number — and it is refused here by construction.

This decision runs in US, UK, SG, AU. Local law shapes what may be measured: where a works council agreement or a local statute restricts cohort analysis, the exclusion is recorded in the monitoring status rather than silently applied.