Eight competencies.
Eight AI agents.
Hired on merit.
The MERIT-8™ methodology — developed by FirstPanel Research — deploys eight AI agents, one per competency, on a single adaptive interview. A consensus engine turns their evidence-linked scores into one Role-Fit Index your legal team can defend. Nothing about the person is measured except what they can do.
5× throughput · 80% less screening time · 100% of applicants interviewed · 0 biometric inference
Every candidate. Every competency. Same rubric.
Each AI agent owns one competency, reads the same transcript, and must attach verbatim evidence to every score — or abstain. No evidence, no score, no guessing.
Role Expertise
Job knowledge and technical depth, tested through role-calibrated scenarios and verifiable claims. The single strongest predictor of performance.
Problem-Solving
Reasoning on novel, role-relevant problems the candidate hasn't rehearsed — the interview-native measure of cognitive ability.
Communication
Clarity, structure, and listening — did they answer the question that was actually asked?
Interpersonal Effectiveness
How they work with others: teamwork, cross-functional collaboration, conflict and feedback — one competency, because it's one factor.
Work Discipline
Conscientiousness in evidence: planning, follow-through, verification habits, ownership of quality.
Values & Drive
Motivation for this role and alignment with your documented values — never similarity to current employees.
Adaptability
Learning agility: response to change, feedback absorbed, skills acquired, plans revised.
Integrity & Judgment
Situational judgment on realistic dilemmas, consistency of account, ethical reasoning.
AI Fluency
Not a ninth competency — a weighted module you switch on per role. It injects AI-competency criteria into Role Expertise (tool use & delegation), Work Discipline (verifying AI output), Integrity & Judgment (responsible use), and Adaptability (adopting new tools). And because the interview itself is a conversation with an AI, we don’t ask candidates whether they can work with AI — we watch them do it, live.
Proof-carrying scores
The Adjudicator — a consensus engine from FirstPanel Research’s multi-agent research — aggregates the eight agents into a weighted Role-Fit Index with honest uncertainty. Agents that disagree are resolved by re-reading the evidence, never by averaging the disagreement away.
Click any rating, see the exact moment that earned it.
A CSM and a senior engineer need different profiles. Strength elsewhere can't buy back a failed floor.
Configure your hiring bar freely — the system rejects anything that touches or proxies a protected attribute.
Bias controls your lawyers will actually believe
Fairness in MERIT-8 isn’t a policy statement — it’s a set of engineering constraints tested on every model release.
Input blinding
Agents never see names, gender signals, age indicators, photos, or postal codes. They score what was said, not who said it.
Content-only signals
Zero facial-expression, voice-tone, or biometric analysis — the capability doesn't exist in the architecture. EU AI Act Article 5 compliant by construction.
Counterfactual testing
Matched transcripts with varied demographic signals must score identically, or the model version doesn't ship.
Four-fifths monitoring
Live adverse-impact dashboards per requisition, with automatic alerts and optional shortlist holds.
Accommodations, invisibly
Voice, video, or text; extended time; pause and resume — none of it visible to scoring agents.
Independent audits
Annual third-party bias audits to the NYC Local Law 144 standard, summaries published.
Test MERIT-8 on your own requisition.
Free pilot: one live role, your rubric, real applicants. First shortlist inside two weeks — judge the method by its evidence.