Hiring decisions deserve proof, not guesswork.
This page explains how our tests are built and checked, where AI is and isn’t used, and how we protect the people who take them. We’ll keep it updated as we grow.
How every test is built
- A psychologist defines exactly what the test measures and how it relates to the job.
- Questions are written, including AI-drafted ones, and every one is reviewed by a psychologist.
- Real people try the questions while we watch for confusion.
- Questions are piloted unscored. Weak or unfair ones are dropped.
- We check reliability and set norms against real applicant groups.
- We publish a technical manual for each test.
Fairness
We check every question for differences in how groups perform, and we drop questions that behave unfairly. Every customer gets pass-rate monitoring by group, with alerts if a cut score starts to screen out a group unfairly. Once we’re in market, an independent firm will audit us for bias every year, and we’ll publish the summary here.
AI, with limits we publish
- Suggests tests, with reasons you can check
- Writes candidate feedback reports
- Flags activity for human review
- Never rejects or ranks people automatically
- Never analyzes faces, voices or emotions
- Never sees candidate names or emails
Every AI output records which model and prompt version produced it, so any result can be reproduced and reviewed.
Security and privacy
- Data is encrypted in transit and at rest.
- Answer keys never reach a candidate’s browser.
- Candidate data is kept for 24 months by default and deleted on request.
- We don’t sell data, and our AI providers don’t train on it.
- We’re preparing for a SOC 2 audit and will publish the report date here once it’s booked.
Rules we design for
The EU AI Act treats AI used in hiring as high-risk, with requirements that apply from December 2, 2027. In the US, we design for the EEOC’s guidelines on employee selection, New York City’s bias-audit law and state AI-in-hiring notice rules. We build to these now rather than retrofitting later.