JAI Bot fairness report
Every application on JAI Bot is read on the experience it shows, and a person always makes the hiring decision. This page shows the results across all employers on JAI Bot, anonymized, so anyone can check them.
Covers the 12 months to 2026-10-26. Updated weekly. Last updated 2026-10-26.
Selection rates by group
From 332 applicants who chose to answer the voluntary self-identification form. Their answers are never used in grading. The ratio compares each group with the group with the highest rate; below 0.80 (the “four-fifths rule”) means “look closer”, not proof of unfairness. Each employer makes its own hiring decisions.
Gender
| Group | People | Moved forward | Interviewed | Hired | Evidence score above median |
|---|---|---|---|---|---|
| Man | 200 | 1% ratio 1.00 | 1% ratio 1.00 | 1% ratio 1.00 | 47% ratio 0.88 |
| Woman | 129 | 0% ratio 0.00 | 0% ratio 0.00 | 0% ratio 0.00 | 53% ratio 1.00 |
1 smaller group is not shown to protect privacy.
Race / ethnicity
| Group | People | Moved forward | Interviewed | Hired | Evidence score above median |
|---|---|---|---|---|---|
| Asian | 262 | 0% ratio 0.00 | 0% ratio 0.00 | 0% ratio 0.00 | 56% ratio 1.00 |
| White | 43 | 5% ratio 1.00 | 2% ratio 1.00 | 2% ratio 1.00 | 19% ratio 0.33 |
3 smaller groups are not shown to protect privacy.
How JAI Bot keeps people in charge
- Applicants who are not selected can be given the reason, approved by a person at the employer.
- Any applicant can ask for a person to review their application, and is told when it happens.
- Self-identification answers are never used in grading, and JAI Bot’s grader learns with names and other identifying details left out.
Read more: How JAI Bot uses AI. This report supports, but does not replace, any independent bias audit an employer may need.