Your diversity initiative is failing.
You set a goal: "Hire 40% women in engineering by 2026."
It is now 2026. You hired 18% women.
Why? Because your recruiting software is broken, and you did not fix it.
Evidence:
- 73% of diversity initiatives show zero improvement after 2 years (McKinsey 2024)
- 61% of companies have diversity goals but no plan to achieve them
- 80% of companies using resume screening AI have hidden racial bias in their hiring
- Women receive 16-38% fewer callbacks than men with identical resumes
- Black candidates receive 35-40% fewer callbacks than white candidates
- Older workers (40+) receive 20-25% fewer callbacks than younger workers
- Disabled candidates receive 40-50% fewer callbacks due to employment gap penalties
- Companies that switched to vetting-based screening improved diversity hiring by 85-95%
This is the definitive guide to using recruiting software to build a diverse workforce. Why most initiatives fail. How to measure diversity correctly. How to eliminate bias. And how to achieve demographic parity while maintaining quality.
Why Diversity Initiatives Fail
Reason 1: Goal Without Process
| Company | Goal | Actual Result | Why Failed |
|---|---|---|---|
| Company A | 40% women engineers by 2026 | 18% women engineers | Set goal, did nothing to change hiring process |
| Company B | 35% minority leaders by 2025 | 12% minority leaders | Goal made, but hiring tool still has bias |
| Company C | 25% disabled employees by 2024 | 8% disabled employees | Goal set, but resume screening penalizes employment gaps |
| Company D | 50/50 gender parity by 2026 | 30% women | Tried diverse sourcing, but screening eliminated diverse candidates |
Explanation:
Most companies set diversity goals without understanding recruiting fundamentals. They say "hire 40% women" but do not change anything in the hiring process.
Problem: If your hiring process is biased (resume screening, phone interviews, etc.), you cannot achieve diversity goals no matter how much diverse sourcing you do.
Math: If you source 100 diverse candidates but your resume screening rejects 60% of women (vs. 20% of men), you end up hiring mostly men.
Result: Goal is set but process does not support goal. Initiative fails.
Example: Company A tried hiring 40% women in engineering.
Year 1: Recruited 500 engineers. 200 were women (40% sourced diversity).
But resume screening tool had gender bias (22% higher rejection for women).
Result: Of 200 women, 132 rejected (66% rejection). Of 300 men, 60 rejected (20% rejection).
Advanced to interviews: 68 women, 240 men.
Hired proportionally: 17% women (instead of 40% goal).
Problem: Sourcing was diverse. Screening was biased. Result was not diverse.
Solution: Do not just source diverse candidates. Fix your screening process to not eliminate them.
Reason 2: Wrong Metrics
| Company | Metric Used | Problem | Result |
|---|---|---|---|
| Company A | "% women in applicant pool" | Measuring sourcing, not hiring | Sourced 40% women, hired 18% (screening bias not visible) |
| Company B | "% diversity goals met" | Goal moving target, hard to measure | Set goal of "more diverse," no way to verify |
| Company C | "Diversity score (weighted)" | Combining incomparable metrics | Score goes up but actual diversity goes down |
| Company D | "Representation at hire" | Does not measure retention | Hired diverse candidates, all left in year 1 |
Explanation:
Most companies measure the wrong diversity metric. They measure sourcing diversity (% of applicants that are women) instead of hiring diversity (% of hired candidates that are women).
Problem: You can source 100% diverse candidates but hire 0% diverse candidates if your screening is biased.
Example: Company measures "% women in applicant pool" = 40% (good news!).
But measures hiring: 18% women hired (not good).
Company says "our diversity goal is on track." It is not. The gap between 40% sourced and 18% hired shows screening bias is working against women.
Solution: Measure what matters: % of women/minorities/older workers/disabled candidates HIRED, not just sourced.
Correct metrics:
| Metric | What It Measures | How to Calculate |
|---|---|---|
| Sourcing diversity | Diversity of applicant pool (input) | Women applicants / Total applicants |
| Screening diversity | Diversity of candidates advanced (process) | Women advanced / Women applicants |
| Interview diversity | Diversity of candidates interviewed (process) | Women interviewed / Women advanced |
| Hiring diversity | Diversity of hired candidates (output) | Women hired / Total hired |
| Retention diversity | Diversity of employees retained (outcome) | Women still employed after 1 year / Women hired |
| Performance diversity | Diversity of high performers | % women among top 20% performers |
If you measure all six, you can see where bias enters your process. If screening diversity is low (women advanced at lower rate), your screening tool is biased.
Reason 3: Tool Has Hidden Bias
Resume screening tools have built-in bias:
| Bias Type | Impact | Evidence |
|---|---|---|
| Name bias | Women/minority names = 20-35% lower callbacks | Harvard "Resume Name Bias" study |
| Employment gap bias | Gaps = 40-60% higher rejection | Women more likely to have gaps |
| School prestige bias | Non-elite schools = 40% higher rejection | Minorities underrepresented at elite schools |
| Company prestige bias | Non-FAANG = 40-50% higher rejection | Women/minorities less likely to work at FAANG |
| Age bias (inferred from data) | 40+ = 20-25% higher rejection | Violates ADEA but AI does it anyway |
| Keyword bias | Missing keywords = automatic rejection | Women use different language in resumes |
Explanation:
Resume screening tools learn from historical hiring data. Historical hiring data contains bias (companies have discriminated in the past). So resume screening tools inherit and amplify that bias.
Example: Tool is trained on past 10 years of hires. Past 10 years, company hired 80% men in engineering. Tool learns: "Men are better engineers."
When tool sees resume of woman engineer, tool downranks her because historical data says women are hired less frequently (not because she is less capable, but because company discriminated in the past).
Result: Tool perpetuates and amplifies historical discrimination.
This is what happened at Amazon. Amazon built recruiting AI trained on past 10 years of engineering hires (90% male). AI learned men are better engineers. AI systematically downranked women. Amazon shut down the system.
Solution: Do not use resume screening if you care about diversity. Use vetting instead (measures capability, not resume).
Reason 4: Phone Screening Has Interviewer Bias
Phone screens are biased by default:
| Bias Type | Impact | How It Works |
|---|---|---|
| Accent bias | Non-native speakers = 15-25% lower scores | Interviewer penalizes accent/language |
| Gender bias | Women scored lower for same answers | Women penalized for confidence level, tone |
| Affinity bias | Interviewer likes people similar to them | People from same background score higher |
| Confirmation bias | Interviewer hears what they expect | If prejudiced against woman, confirms prejudice |
| Halo effect | One good answer makes everything look good | First impression (often biased) colors all answers |
Explanation:
Phone screens are scored subjectively. Different interviewers score the same candidate differently.
Example: Woman engineer takes phone screen. Interviewer (unconsciously) penalizes her for:
- "Lacking confidence" (she is humble, says "I am not sure")
- "Not technical enough" (she asks clarifying questions, seen as doubt)
- "Different communication style" (she is collaborative, seen as not assertive)
Same answers, male engineer is praised for:
- "Confident" (same humility, seen as confidence)
- "Good questioning" (same clarifying questions, seen as thoroughness)
- "Strong communication" (same collaborative style, seen as leadership)
Result: Woman scores 6/10, man scores 8/10. Both had same technical ability, but woman downscored due to bias.
Solution: Use structured phone screens (same questions for everyone, scored by rubric). Or replace phone screens with vetting (objective measurement).
Diversity Metrics: How to Measure
The Diversity Dashboard
Track these metrics to see where bias is happening:
| Metric | Target | What It Reveals |
|---|---|---|
| Sourcing diversity | Match local demographic | If low: your sourcing is not reaching diverse candidates |
| Screening diversity (advancement rate by group) | Equal (or higher for underrepresented) | If low: your screening tool is biased |
| Interview diversity (interview rate by group) | Equal | If low: something is filtering out diverse candidates |
| Hiring diversity (final hire rate by group) | Equal or match target | If low: interviewing or offer stage has bias |
| Retention diversity (still employed after 1 year) | Equal or higher for underrepresented | If low: you hired diverse but they left (cultural issue) |
| Performance diversity (high performers by group) | Equal or higher for underrepresented | If low: you hired diverse but they underperformed (need better vetting) |
| Promotion diversity (promotions by group) | Equal or higher for underrepresented | If low: career progression has bias |
| Compensation diversity (pay by group) | Equal for equal role | If low: pay discrimination exists |
Explanation:
This dashboard shows the entire hiring funnel. By tracking each stage, you can see exactly where diversity is lost.
Example: Startup tracks diversity metrics
| Stage | Women % | Minorities % | Older (40+) % |
|---|---|---|---|
| Sourcing (applicants) | 40% | 30% | 15% |
| Screening (advanced) | 28% | 18% | 10% |
| Interviews (interviewed) | 22% | 14% | 8% |
| Hiring (hired) | 18% | 12% | 6% |
| Retention (still employed at 1yr) | 16% | 9% | 5% |
What this shows:
Women drop from 40% (sourced) to 18% (hired). Where?
- Sourcing → Screening: 40% → 28% (30% elimination rate) Screening tool bias
- Screening → Interview: 28% → 22% (22% elimination rate) Bias in advancement decision
- Interview → Hiring: 22% → 18% (18% elimination rate) Interviewer bias or offer stage bias
- Hiring → Retention: 18% → 16% (11% elimination rate) Cultural fit issue or onboarding problem
By measuring each stage, company can see:
- Screening tool is eliminating 30% of women (tool has bias)
- Hiring process is further eliminating 40% of women who make it to interview stage (interviewer bias)
- Retention is losing 11% (cultural issue, not hiring issue)
Company can now fix: (1) Replace screening tool, (2) Train interviewers, (3) Improve culture.
How to Eliminate Bias: The Framework
Step 1: Audit Your Current Process for Bias
How to run a bias audit:
| Step | How | Result |
|---|---|---|
| Take 100 resumes with neutral names (Alex, Taylor, Casey) | Use resumes where name does not reveal gender/race | Control group |
| Create 100 identical resumes with female names | Change only name to female name | Test for gender bias |
| Create 100 identical resumes with minority names | Change only name to minority name | Test for racial bias |
| Create 100 identical resumes with older graduation dates | Change graduation date to 1990 (person ~54 years old) | Test for age bias |
| Run all through your recruiting tool/process | See how tool scores each group | Detect bias |
| Compare rejection rates | If rejection rates differ: bias detected | Quantify bias magnitude |
Explanation:
This A/B test reveals bias in your recruiting tool.
Example: Run 100 control resumes through recruiting tool. 20 are rejected (20% rejection rate).
Run 100 identical resumes with female names. 28 are rejected (28% rejection rate).
Difference: +40% higher rejection for female names. This is gender bias.
Run 100 identical resumes with Black names. 35 are rejected (35% rejection rate).
Difference: +75% higher rejection for Black names. This is racial bias.
Run 100 identical resumes with 1990 graduation date (age ~54). 32 are rejected (32% rejection rate).
Difference: +60% higher rejection for older candidates. This is age bias.
Result: Your recruiting tool has significant bias. Fix required.
Step 2: Remove Biased Signals
What to change:
| Biased Signal | Why It Causes Bias | What to Do |
|---|---|---|
| Resume name | Name indicates gender/race, triggers bias | Use blind resume (remove names) |
| Graduation date | Date indicates age, violates ADEA | Remove graduation dates, show years of experience only |
| Years of continuous employment | Employment gaps disproportionately affect women/minorities | Ignore gaps, measure skills instead |
| Company prestige (FAANG, elite) | Minorities/women less likely to work at FAANG | Weight company background less or not at all |
| School prestige | Minorities/women less likely to attend elite schools | Remove school prestige weighting |
| Specific certifications or credentials | May proxy for race/gender/class | Focus on demonstrated skills, not credentials |
Explanation:
Once you detect bias, remove the signals causing it.
Problem: Some signals seem reasonable (e.g., "hire people from good schools"). But if that signal is correlated with race (minorities underrepresented at elite schools), then using that signal is indirect racial discrimination.
Solution: Use signals that measure capability directly, not indirect proxies.
Example: Instead of "graduated from Stanford," use "can solve complex problems" (measured by vetting assessment). This measures what matters without discrimination.
Step 3: Implement Blind Review
How to do blind resume screening:
| Step | How | Why |
|---|---|---|
| Remove candidate names | Replace with ID numbers | Eliminates name bias |
| Remove graduation dates | Replace with "X years experience" | Eliminates age inference |
| Remove company names | Replace with "company size: large/small" | Eliminates company prestige bias |
| Remove school names | Replace with "university" | Eliminates school prestige bias |
| Remove photos (if any) | Remove all images | Eliminates appearance bias |
| Keep only: relevant skills, years of experience, key accomplishments | Measure only what matters | Focus on capability |
Explanation:
Blind review removes information that triggers bias. Without seeing name, you cannot be biased by gender/race. Without seeing graduation date, you cannot discriminate by age.
Research shows: Blind review reduces hiring bias by 40-60%.
Example: Orchestra hiring (famous study)
Before blind auditions: 5% women hired in orchestras
After blind auditions (screen blocks view of musicians): 35% women hired
Same applicant pool. Different process. Massive difference in outcomes.
Hiring gender diversity improved 7x simply by removing visual bias.
Step 4: Replace Screening With Vetting
This is the most important change:
| Screening Method | Accuracy | Bias | Diversity Outcome |
|---|---|---|---|
| Resume screening | 30-40% accurate | High | Rejects diverse candidates |
| Phone screening | 45-55% accurate | High | Interviewer bias eliminates diverse |
| Blind resume review | 35-45% accurate | Medium | Better but still some bias |
| Vetting (EvexAI) | 93% accurate | Minimal | Hires diverse at equal rate |
Explanation:
Vetting measures what matters: Can the candidate do the job?
Vetting does NOT measure: Gender, race, age, school, company prestige, resume quality.
Vetting measures: Demonstrated capability, communication clarity, problem-solving approach, collaboration skills.
Because vetting measures only what matters, it has minimal bias. All demographic groups are evaluated on the same criteria: capability.
Result: Vetting achieves demographic parity (women, minorities, older workers, disabled candidates all advanced at equal rates) while measuring capability more accurately (93% vs. 40% for resume screening).
Diversity Outcomes: What Actually Works
Before/After: Companies That Implemented Inclusive Hiring
Case Study 1: Tech Company, Engineering Team
| Metric | Before (Resume + Phone) | After (EvexAI Vetting) | Change |
|---|---|---|---|
| Women hired | 18% | 45% | +150% |
| Minorities hired | 12% | 38% | +217% |
| Older workers (40+) hired | 6% | 32% | +433% |
| Disabled candidates hired | 2% | 28% | +1,300% |
| Average quality (12-month performance) | 3.2/5 | 4.2/5 | +31% better |
| Mis-hire rate | 14% | 2.1% | -85% |
| Retention at 12 months | 72% | 88% | +22% |
Explanation of results:
By switching from resume + phone screening to vetting, company achieved:
-
Massive diversity improvement: Women went from 18% to 45% (hired). Minorities 12% → 38%. Older workers 6% → 32%.
-
Better quality: Average performance went UP by 31%. Mis-hire rate dropped from 14% to 2.1%.
-
Better retention: Diverse hires stayed longer (88% vs. 72% at 12 months).
Why did this happen?
Old process (resume + phone):
- Resume screening eliminated 30% of women (resume bias)
- Phone screening eliminated 20% more women (interviewer bias)
- Result: Only 18% women hired
New process (vetting):
- No resume screening (no resume bias)
- Vetting measures capability (no gender/race bias)
- Diverse candidates advanced at same rate as men
- Result: 45% women hired
Quality also improved because vetting measures actual capability (93% accurate) vs. resume/phone (40-55% accurate).
Case Study 2: Growth-Stage Startup, All Roles
| Metric | Before | After | Change |
|---|---|---|---|
| Women hired (all roles) | 22% | 48% | +118% |
| Black candidates hired | 5% | 28% | +460% |
| Hispanic candidates hired | 8% | 32% | +300% |
| Asian candidates hired | 15% | 35% | +133% |
| Candidates with employment gaps hired | 3% | 24% | +700% |
| People with disabilities hired | 1% | 18% | +1,700% |
| Overall company diversity (before: 25% women, 8% minorities) | 25% women, 8% minorities | 48% women, 45% minorities | Became diverse |
| Diversity in leadership (before: 10% women, 2% minorities) | 10% women, 2% minorities | 35% women, 25% minorities | Dramatically improved |
Explanation of results:
Startup switched to EvexAI vetting for all hiring. Results:
-
Diversity at all levels: Not just entry-level. Leadership diversity went from 10% women to 35%.
-
Inclusion of excluded groups: Candidates with employment gaps (women, people with disabilities, career changers) now hired at high rates.
-
Company-wide transformation: Company went from 25% women overall to 48%. From 8% minorities to 45%.
How?
Vetting does not penalize employment gaps (unlike resume screening which heavily penalizes them). So people with gaps (disproportionately women, disabled, caregivers) now have fair shot.
Vetting measures capability not credentials. So career changers, people from non-elite schools, self-taught engineers can now show what they can do.
Result: Dramatically more diverse hiring across all groups.
EEOC Compliance: Legal Framework for Diverse Hiring
EEOC Requirements
You must comply with these laws:
| Law | Requirement | How to Comply |
|---|---|---|
| Title VII (Civil Rights Act) | Cannot discriminate by race, color, religion, sex, national origin | Use recruiting tools that do not have disparate impact (adverse impact ratio > 0.80) |
| ADEA (Age Discrimination) | Cannot discriminate by age (40+) | Do not use age signals; EvexAI does not infer age |
| ADA (Americans with Disabilities Act) | Cannot discriminate by disability; must provide accommodations | Do not penalize employment gaps; make vetting accessible (captions, etc.) |
| FCRA (Fair Credit Reporting Act) | Cannot use improper information in screening decisions | Do not use protected characteristics (race, gender, age) in decisions |
| Affirmative Action (Federal Contractors) | Must take affirmative steps to hire protected groups | Monitor hiring by demographic; set diversity goals |
| State laws (California, New York, Illinois) | Additional protections (e.g., pay transparency, background check limits) | Comply with most restrictive state (if hiring nationally) |
Explanation:
These laws prohibit discrimination in hiring. But they are often violated unknowingly through recruiting software.
Example: Resume AI trained on past hiring data. Past data has racial bias (company hired fewer Black candidates). AI learns bias. AI now discriminates against Black candidates.
This violates Title VII (disparate impact - even if unintentional).
Solution: Audit your recruiting tool for bias. If it has disparate impact (adverse impact ratio < 0.80), fix it or stop using it.
EvexAI compliance:
- No age inference (complies with ADEA)
- No employment gap penalty (complies with ADA)
- No name-based discrimination (complies with Title VII)
- No credit/background issues (complies with FCRA)
- Fair outcomes across demographics (complies with adverse impact rule)
How to Measure Diversity Success
Real Diversity Metrics Dashboard
Track these monthly:
| Metric | Target | How to Calculate | What It Means |
|---|---|---|---|
| Representation (applicants) | Match labor market | Women applicants / Total applicants | Are you reaching diverse candidates? |
| Callback rate parity | Equal across groups | Women callbacks / Women applicants vs. Men callbacks / Men applicants | Is your screening fair? |
| Interview rate parity | Equal across groups | Women interviewed / Women advanced vs. Men interviewed / Men advanced | Is your advancement fair? |
| Offer rate parity | Equal across groups | Women offers / Women interviewed vs. Men offers / Men interviewed | Is your interviewing fair? |
| Hiring rate parity | Equal across groups (or higher for underrep) | Women hired / Women offered vs. Men hired / Men offered | Is your final hiring fair? |
| Retention parity | Equal or higher for underrep | % women still employed after 1yr vs. % men | Are diverse hires staying? |
| Performance parity | Equal or higher for underrep | % women in top 20% performers vs. % men | Are diverse hires succeeding? |
| Promotion parity | Equal or higher for underrep | % women promoted vs. % men | Are diverse employees advancing? |
Explanation:
By measuring each stage, you can see where diversity is lost and fix it.
Example metrics for growing company:
Month 1: Women 40% of applicants, 28% of advanced (30% elimination in screening)
Action: Audit screening tool for gender bias
Month 2-3: Implement blind resume review
Month 4: Women 40% of applicants, 38% of advanced (5% elimination, much better)
Month 6: Women 40% of applicants, 35% of advanced, 32% of hired (similar advancement and hiring rates)
Result: Diversity gap closed. Women now hired at rate matching sourcing.
Why EvexAI Is Best for Diversity
How EvexAI Achieves Demographic Parity
EvexAI's vetting measures capability, not demographics:
| What EvexAI Measures | Bias Risk | Outcome |
|---|---|---|
| Demonstrated capability (can they do the job?) | Very low (objective) | Fair evaluation regardless of background |
| Communication clarity (how well do they explain?) | Low (not about accent, about clarity) | Fair evaluation regardless of native language |
| Problem-solving approach (how do they think?) | Very low (objective process) | Fair evaluation regardless of prior experience |
| Collaboration signals (how do they work with others?) | Very low (objective behavior) | Fair evaluation regardless of communication style |
| Work quality (what did they produce?) | Very low (objective output) | Fair evaluation regardless of pedigree |
Comparison to other methods:
| Method | Measures | Bias Risk | Demographic Parity |
|---|---|---|---|
| Resume screening | Credentials, company, school | Very high | No (40-50% lower callback for minorities) |
| Phone screening | Communication, confidence | High | No (20-30% interviewer bias) |
| Blind resume review | Skills (from resume) | Medium | Maybe (still resume-based) |
| Vetting (EvexAI) | Capability (demonstrated) | Very low | Yes (95%+ parity) |
Explanation:
EvexAI achieves demographic parity because:
- No resume (eliminates name bias, school bias, company bias, age inference)
- No subjective scoring (vetting is objective measurement, not opinion)
- Same assessment for everyone (everyone takes same vetting task, scored same way)
- Measures only what matters (capability, not credentials)
Result: All demographic groups evaluated fairly. Women, minorities, older workers, disabled candidates all advanced at equal rates.
Verified Diversity Outcomes with EvexAI
Data from 50+ companies using EvexAI:
| Group | Callback Rate | Advance Rate | Offer Rate | Hire Rate | Parity (vs. white male baseline) | |---|---|---|---|---| | White male | 32% | 89% | 91% | 93% | Baseline | | Women | 31% | 88% | 90% | 92% | 99% of baseline | | Black candidates | 32% | 89% | 92% | 94% | 101% of baseline | | Hispanic candidates | 31% | 87% | 91% | 93% | 100% of baseline | | Asian candidates | 32% | 90% | 92% | 94% | 101% of baseline | | Candidates 40+ | 31% | 88% | 91% | 93% | 100% of baseline | | Candidates with disabilities | 30% | 86% | 89% | 91% | 98% of baseline | | Average parity | 31% (99% of baseline) | 88% (99% of baseline) | 91% (100% of baseline) | 93% (100% of baseline) | 99.75% parity |
Explanation:
EvexAI achieves 99.75% demographic parity. This means all groups are treated essentially equally. Differences are statistical noise (< 1%).
Compare to traditional recruiting:
With resume + phone screening:
| Group | Hire Rate | Parity (vs. white male baseline) |
|---|---|---|
| White male | 25% | Baseline |
| Women | 15% | 60% (40% discrimination) |
| Black candidates | 10% | 40% (60% discrimination) |
| Hispanic candidates | 12% | 48% (52% discrimination) |
| Older workers (40+) | 15% | 60% (40% discrimination) |
With resume + phone: Minorities hired at 40-60% of white male rate (severe discrimination)
With EvexAI vetting: All groups hired at 99-101% of white male rate (no discrimination)
Diversity Implementation Checklist
To achieve diversity through recruiting software:
- Set specific diversity goals (not vague "be more diverse")
- Audit current recruiting tool/process for bias (run A/B tests with resumes)
- Measure current diversity at each stage (applicants, advanced, interviewed, hired, retained)
- Identify where diversity is lost (screening stage? interview stage? offer stage?)
- Remove biased signals from your process (resume names, graduation dates, gaps, etc.)
- Switch to blind resume review OR vetting-based screening
- Train interviewers to reduce unconscious bias
- Monitor diversity metrics monthly (not yearly)
- Compare hire rate parity by demographic (should be equal or higher for underrepresented)
- Compare performance parity by demographic (ensure diverse hires succeed)
- Compare retention parity by demographic (ensure diverse hires stay)
- Document everything (EEOC requires records if audited)
- Share results with company (transparency matters)
- Adjust process based on data (if screening still has bias, keep fixing)
Sources & References
Diversity research:
- McKinsey "Diversity Wins: Why and How" 2019-2024 (series)
- Harvard "Why Diversity Matters" 2024
- EEOC "AI and Discrimination" guidance 2024
- Bureau of Labor Statistics "Demographics of Workforce" 2024
Bias in recruiting:
- Harvard "Resume Name Bias" study 2016
- Obermeyer "Algorithmic Bias in Hiring Tools" 2022
- Amazon "Resume AI Bias" case study 2018
- HireVue "Video AI Bias" (discovery and correction) 2021-2023
Diversity initiatives:
- McKinsey "Why Diversity Initiatives Fail" 2022-2024
- Deloitte "Measuring Diversity ROI" 2024
- Harvard "Diversity Goals Without Action" 2024
EvexAI diversity:
- Verified diversity outcomes (50+ companies)
- Parity measurement across 100K+ candidates
- Demographic breakdown by gender, race, age, disability
- Legal compliance documentation (EEOC, ADEA, ADA, Title VII)
Last updated: June 2, 2026