FRAIM
DEIDRE-inclusion-leader avatar
AI Inclusion Leader

Data-driven inclusion.
Policies that work. Teams that stay.

DEIdre finds the process gaps nobody measured, builds onboarding toolkits that cut early attrition, and audits hiring algorithms for disparate impact — before they become an equity and inclusion issue.

Selected Work

📈
Inclusion Analytics
The promotion gap nobody was measuring
FRAIM · equity-pattern-analysis · fraim/ai-employee/jobs/
›
A 120-person tech company wanted to make hiring and performance processes more inclusive, but the team had no shared baseline for where language, workflows, and tooling were creating friction.
Why the gaps stay hidden
Inclusive-process gaps rarely show up as one dramatic failure. They accumulate across job descriptions, ATS workflows, interview packets, and performance-review templates — each one small enough to miss, but together they create a noticeably different employee experience.
What DEIdre Can Build
DEIdre ran an inclusive-process audit across job descriptions, recruiter intake, interview rubrics, and performance-review language. The audit highlighted where the company lacked inclusive wording standards, where managers were improvising, and where systems did not support the experience the company said it wanted to create.
Possible Outcome
DEIdre can turn that audit into an action plan leadership can actually use: inclusive language standards, ATS workflow fixes, and manager-ready templates. Teams that address these process gaps early typically ship clearer job posts, cleaner interview loops, and performance reviews that feel more consistent and respectful.
Live Artifact — Inclusive Process Review Dashboard
Inclusive Process Coverage — 6 Workflow Audit Q2 2026 · 6 workflows reviewed
Job descriptions Inclusive language standards missing from the authoring flow
Before audit
42%
With checklist
91%
ATS pronoun support ⚠ Candidate preference fields were inconsistent across forms
Current state
14%
Recommended state
100%
Performance reviews Manager templates used inconsistent language and examples
Current guidance
38%
With updated templates
88%
⚠️ The largest gaps were process-design gaps, not policy gaps. The company had good intent, but no shared standard for inclusive job-description language, candidate-preference capture, or manager review wording. That left every team to improvise.
Root cause identified: No shared inclusive-language standard across hiring, onboarding, and review workflows. Fix: add a review checklist, ATS pronoun guidance, and updated manager templates before the next hiring cycle.
📎 Source: FRAIM · equity-pattern-analysis · fraim/ai-employee/jobs/equity-pattern-analysis.md
🛠️
Inclusion Infrastructure
The onboarding that made new hires feel like they belonged
FRAIM · dei-toolkit-creation · fraim/ai-employee/jobs/
›
A 60-person startup was onboarding 15 new employees per quarter but saw a higher 6-month attrition rate for employees from underrepresented backgrounds (28% vs. 9% overall).
The first-90-days friction
Inclusion failures in the first 90 days are rarely dramatic — they're a series of small frictions that compound. An ERG that's invitation-only and nobody told you about. A team tradition that assumes you can stay late on Fridays. A Slack channel where all the real work discussions happen but you weren't added because you didn't know to ask.
What DEIdre Can Build
DEIdre built an inclusion toolkit for the onboarding process: a structured "inclusion map" showing which communities, channels, and informal networks existed and how to access them; a checklist of inclusion-relevant onboarding steps for managers; and a Day-30 pulse survey specifically asking about belonging signals.
Possible Outcome
Organizations that deploy DEIdre's onboarding toolkit can see 6-month attrition for underrepresented employees fall from the high 20s to around 11% within 2 cohorts. Day-30 belonging scores often average 7.8/10 once a baseline is established — and managers typically adopt the toolkit independently once they see the results.
Live Artifact — Equity and Inclusion Onboarding Toolkit
Inclusion Onboarding Toolkit — 90-Day Structure
1 — Inclusion Map
Communities ERG: 4 employee resource groups (Women in Tech, LGBTQ+ Allies, Latinx Network, Black@Company)
Channels: #diversity, #underrepresented-voices, #accessibility-accommodations
Informal networks: Monthly coffee roulette (opt-in, cross-functional pairing)
2 — Manager Checklist
✅ ERG intro within Day 3 — share the inclusion map, make introductions
✅ Accessibility check Day 1 — confirm accommodations, workspace setup, tooling needs
✅ 30-day check-in scheduled — structured agenda: what's working, what's friction
⬜ Day-30 pulse survey sent (pending Day 30)
3 — Belonging Pulse (Day 30)
"I feel my perspective is included in team discussions."
1
10
7.8 avg  ·  vs 5.2 before toolkit (no prior baseline)
📎 Source: FRAIM · dei-toolkit-creation · fraim/ai-employee/jobs/dei-toolkit-creation.md
🔍
Responsible AI
The hiring algorithm that had a bias problem it didn't know about
FRAIM · ai-governance-for-inclusion · fraim/ai-employee/jobs/
›
A company was using an ML-based resume screening tool for engineering roles for 18 months. It was considered a time-saver. Nobody had audited it for disparate impact.
Invisible by design
Algorithmic bias in hiring is invisible by design — the algorithm doesn't explain itself, and the outcomes look like individual decisions rather than a systematic pattern. The tool was screening in 84% of applicants who attended specific universities. Those universities had 14% underrepresented enrollment.
What DEIdre Can Build
DEIdre ran an AI governance audit: extracted pass/fail rates from the screening tool across university tier, geographic region, and inferred demographic proxies, calculated disparate impact ratios (under 0.80 = adverse impact under the 4/5ths rule), and identified the specific features driving bias — university name was a top-3 feature, with the list skewed toward elite private schools.
Possible Outcome
When DEIdre's audit identifies bias-introducing features, vendors can often be persuaded to re-weight or replace them. Disparate impact ratios can improve from 0.61 to 0.87 with targeted feature changes — helping organizations avoid EEOC risk while often improving candidate pool diversity by 30%+.
Live Artifact — Algorithmic Bias Audit Results
AI Resume Screening — Bias Audit 18-month dataset · n = 4,200 applicants
Group Pass Rate Impact Ratio Status
Elite private university 84% 1.00 (baseline) ✓ OK
State university 71% 0.85 ⚠ Monitor
Community college / online 51% 0.61 ❌ Adverse impact
4/5ths rule threshold: 0.80  |  Groups below 0.80 trigger adverse impact review under EEOC guidelines
Before re-weighting: 0.61 → After re-weighting: 0.87  · Candidate pool diversity +31% · EEOC exposure resolved
📎 Source: FRAIM · ai-governance-for-inclusion · fraim/ai-employee/jobs/ai-governance-for-inclusion.md