Use Case 1
Quarterly OFCCP Compliance Audit Before a Federal Contract Renewal
Trigger: Legal counsel notifies the DEI Lead that an OFCCP compliance review is scheduled in 45 days.
1.DEI Lead provides the 12-month pipeline demographics dataset and JD corpus folder path.
2.Agent ingests all data, validates demographic self-ID completeness, and flags job families where self-ID rates fall below 70%.
3.Agent runs adverse impact calculations across all five funnel stages for each protected class, applying Fisher's exact test and Bonferroni correction.
4.Upon detecting High severity findings (AI ratio below 0.70), agent sends a Slack alert and pauses external report distribution.
5.Agent scans all 47 JDs in the corpus against the organization's approved bias lexicon.
6.Agent generates a complete OFCCP-formatted PDF audit report with methodology appendix, findings tables, and funnel charts, then creates a Gmail draft for legal counsel and Head of Talent.
✓
The DEI Lead receives a fully documented, OFCCP-aligned audit package within hours rather than weeks, with High severity findings escalated immediately for remediation planning.
Use Case 2
Proactive JD Bias Sweep Before a High-Volume Engineering Hiring Push
Trigger: The Head of Talent wants to ensure all 60 draft engineering JDs are inclusive before posting.
1.Head of Talent uploads all 60 draft JDs to a Google Drive folder.
2.Agent reads each JD, scans for terms on both the built-in bias lexicon and the organization's custom approved lexicon.
3.Agent produces a prioritized Google Sheet ranking all 60 JDs by flag count and severity.
4.For the 15 High-priority JDs, agent generates a DOCX summary with side-by-side flagged excerpts and recommended rewrites.
5.Agent sends a Slack message summarizing the sweep results: total JDs reviewed, flag rate, most common problematic terms, and a link to the full Google Sheet.
✓
The Head of Talent receives a ranked, actionable bias audit of all 60 JDs before a single role is posted — with specific term-level findings and rewrite suggestions.
Integrations
RequestApproval (formal approval gate before report distribution)
GeneratePPTX (board-ready executive summary deck)
ATS / HRIS Direct API Integration (Greenhouse, Workday, Lever)
Industries
technologyretailhospitalityhealthcare
See It In Action
A sample conversation with this agent
What does this agent actually do and how do I kick off my first audit?
The agent explains its five audit phases (data ingestion, adverse impact analysis, JD language scan, scorecard pattern review, and report generation), lists the Google Drive folders and Sheets it needs, and walks the user through the two inputs required: a pipeline demographics export from the ATS and the JD corpus folder path.
Run a full adverse impact audit on our Q2 hiring pipeline. The ATS export is in the 'Hiring Data / Q2 2025' folder on Google Drive.
The agent ingests the pipeline demographics sheet, computes selection rates for each demographic group at every funnel stage, applies the EEOC 4/5ths rule and Fisher's exact test (with Bonferroni correction), flags any stage-group combination with an adverse impact ratio below 0.80, suppresses cells with n<5, and produces a formatted PDF audit report with executive summary, methodology appendix, and funnel visualization chart. Slack alert is sent for any High severity findings (ratio below 0.70).
We only have demographic self-ID data for about 55% of candidates. Can you still run the analysis?
The agent proceeds but automatically inserts a prominent data quality disclaimer in every affected section, noting that self-identification rates fall below the 70% confidence threshold. The methodology appendix explains the statistical power implications and recommends improving self-ID collection before the next audit cycle. Findings are labeled with reduced-confidence indicators throughout.
Compare our interview scorecard ratings for the last three quarters broken down by interviewer panel composition and candidate demographic group.
The agent reads scorecard data from Google Sheets, computes mean scores and score distributions per demographic group for each interviewer and panel, runs statistical tests with Bonferroni correction, and produces a heatmap chart and detailed findings table identifying specific interviewers or panels with the largest disparities. Output includes recommended calibration interventions.