How decisions are made
Yareta is decision support, not an automated decision-maker. Every hire, investment or selection decision is made by a person.
Human review
- A named human reviewer in your organisation, or a Yareta psychologist you’ve engaged, makes the hire / invest / select decision.
- Review must be meaningful: the reviewer should have the authority and the context to disagree with or override what Yareta shows. Yareta outputs shouldn’t be used as the sole basis for a decision about a person.
- Every report states this near the top and in the footer.
- This applies to every output, including Predictive Numbers and team breakdown risk signals. A signal is a prompt to investigate, not a verdict.
If you request a psychologist in the loop, Yareta assigns one on top of the automated report. Your organisation remains accountable for the final decision unless your contract explicitly states otherwise.
What the numbers are
- Every score comes from the deterministic scoring engine, using fixed rubrics and your framework configuration.
- Language models never create or change a score. They help interpret text and write the narrative.
- Scorecards record the engine version and framework version, so any score can be traced to exactly how it was produced, including after a framework is recalibrated.
- A Success Framework Score measures fit to a framework. It becomes a Predictive Number only once that framework’s predictive validity has been proven against real outcomes.
Provenance
Every finding traces back to its source: documents, transcripts or assessments. When a report is regenerated, findings are carried, enriched, added or retired, never silently dropped, and the audit trail is kept. Human edits are attributed to the person who made them.
Confidence is not fit
Low confidence means “not enough evidence yet,” not “poor fit.” Use probing questions and the optional Yareta Assessment to raise it. See How scoring works: Confidence.
Transparency to the people being assessed
- Reports disclose that content is AI-assisted.
- Where AI interview tools (such as live avatar) are used, candidates are told they’re interacting with AI and that it informs their assessment.
- People in Yareta’s assessed network are only discoverable to customers if they’ve consented.
- Individuals can request an explanation of how they were assessed and ask for a human review. To do this, email connect@yareta.co.
Regulation
Hiring and people-assessment uses of AI are treated as high-risk under the EU AI Act and are subject to UK GDPR rules on automated decision-making. Yareta is designed to support these obligations through human oversight, logging, versioning, provenance and transparency. Your organisation, as the user of the system, keeps its own obligations as the decision-maker. This page describes how the product works. It isn’t legal advice for your organisation.
Access and security
The customer interface never exposes source code or system secrets. You access Yareta through the app, managed integrations and the keys you create (developer, Zapier, OAuth), which you must store securely. See Authentication.