Responsible AI and Fairness
How Coco HR approaches AI in hiring — what it does, what it deliberately does not do, and how to use it fairly and lawfully.
Hiring decisions affect people's livelihoods. Coco HR uses AI to make screening faster and more consistent, but it is designed so that a human, not a model, decides who gets hired. This page sets out our approach and your responsibilities as a customer.
AI assists — humans decide
Coco HR's AI generates questions, transcribes spoken answers, and scores responses. It produces decision support: structured scores, per-question feedback, and evidence.
What it does not do:
- It does not output a hire / no-hire recommendation.
- It does not automatically reject candidates based on scores or proctoring signals.
- It does not set a candidate's verdict.
The recruiter reviews the report — scores, answers, transcripts, recordings, and anti-cheat signals — and sets the verdict (Pending / Selected / Rejected). AI narrows and organises the review; the judgement stays with a person. See How AI scoring works.
Use assessments as one input
An assessment score is one signal among many. To use it responsibly:
- Combine it with other evidence — interviews, work samples, references, and role-specific judgement.
- Review the underlying answers and recordings, not just the headline percentage. Transcription errors, phrasing, or nerves can move a score without reflecting ability.
- Be consistent. Apply the same assessment, criteria, and review process across candidates for a role.
Accessibility and accommodations
Candidates may need adjustments to demonstrate their ability fairly. The assessment includes an accessibility accommodation option, and you should:
- Make candidates aware they can request accommodations.
- Offer alternatives where an assessment format disadvantages someone for reasons unrelated to the job.
Transparency and consent
- Candidates are recorded during assessments (webcam, screen, snapshots) and consent is collected before the assessment begins.
- Be clear with candidates about how their data and recordings are used, and for how long. See Security and privacy.
Bring-your-own model changes the provider
A company can configure its own AI model — OpenAI, Anthropic, OpenRouter, Groq, Mistral, or a custom endpoint — instead of the platform default. If you do:
- Scoring quality and behaviour depend on the model you choose. Different providers reason differently; results are not guaranteed to match the platform default.
- You control where assessment content is sent for evaluation, which has data-handling implications.
- Re-baseline what a "good" score looks like for your roles after switching, and validate the model behaves acceptably for your use case.
Comply with local hiring law
You are responsible for using Coco HR in line with the laws that apply to you. Common examples:
- United States — anti-discrimination and fair-hiring rules (e.g. EEOC guidance), plus state and local rules on automated employment decision tools.
- European Union — GDPR for candidate personal data, including recordings and automated-processing transparency.
Coco HR gives you the tools (human-in-the-loop review, accommodations, recordings, consent capture); applying them lawfully in your jurisdiction is your responsibility. When in doubt, consult your legal or compliance team.
Summary
| Principle | In practice |
|---|---|
| Human decides | AI scores; a recruiter sets the verdict. |
| One input | Assessments complement interviews and other evidence. |
| Review the evidence | Read answers, transcripts, and recordings — not just the score. |
| Accessible | Offer accommodations. |
| Transparent | Candidates consent to recording up front. |
| Lawful | Comply with EEOC, GDPR, and local rules. |