AI Interviewer Buyer's Guide

1. Can recruiters edit questions and scoring logic per role?

A strong answer shows that every job role supports an independent rubric and that each question's weight can be configured separately. Platforms that apply a single template to all roles fail to enforce the primary finding of structured interview validity research — same questions, same criterion weights.

Intrvio ships a rubric builder with role-based competency packs. Recruiters can create, edit, or delete any question; assign per-question weights; and add behaviorally anchored rating descriptions for specific roles. Changes are versioned: old rubric versions are archived and remain in the audit trail matched to previous interview results.

2. Does the AI adapt follow-ups or follow a fixed script?

A strong answer demonstrates that the system can generate live follow-ups that directly respond to what the candidate said. Pre-built question trees produce questions that sound realistic but are unconnected, which leads to 'interview-able behaviors' rather than authentic evaluation — the scoring loop, not the assessment loop, is working.

GAIA conducts each conversation in real time. If a candidate gives a non-specific answer, GAIA probes deeper; if a concrete example is provided, GAIA follows it and asks for the outcome and learnings. This adaptation is constrained by the rubric frame: follow-ups do not drift away from assessment criteria.

3. Are transcripts and scoring rationale exportable for audit?

A strong answer shows that the full transcript, the rationale for every rubric score, and any human overrides are exportable in plain-text or structured formats. Platforms that offer only a summary or aggregate score do not provide sufficient evidence for a legal or regulatory review.

Intrvio produces transcript-anchored rationale for every interview: each rubric criterion is scored with a link to the relevant transcript passage. All outcomes and human overrides are written to a signed audit log. Employers can export the full package as PDF or JSON and use it in EU AI Act compliance documentation.

4. How is human oversight enforced?

A strong answer makes explicit that AI output is advisory and that the final decision always belongs to a human. It also shows that every decision action is logged and that the human review path is a mandatory workflow step rather than an optional feature.

In Intrvio, GAIA scores are overridable: a recruiter can change any rubric score with a justification. All changes are written to a signed audit log. The hiring decision does not auto-generate an offer or rejection — it requires an explicit action from the recruiter panel. This design aligns with EU AI Act Article 14 — human oversight.

5. What happens when a candidate requests accommodation or human review?

A strong answer describes a candidate-reachable accommodation request flow and a documented recourse path through which a candidate can request an explanation or human review via the candidate notice mechanism.

Candidates can declare accommodation needs before the interview; the platform supports adjustments such as extended time or alternative format. During or after the interview, candidates can request human review via the candidate notice portal. Requests are routed to the employer and logged in the audit trail. Under EU AI Act Article 86, rejected candidates can request an explanation of the elements that contributed to the decision.

6. Which languages are supported?

A strong answer separates language support into two layers: in which languages an active agent is available for the live conversational flow, and what language coverage exists for transcription and text-to-speech. Broad transcription support alone does not imply full interview delivery.

Intrvio currently has dedicated live voice agents for Turkish, English, and Hindi. Transcription and text-to-speech cover 42-plus languages via ElevenLabs and Deepgram, enabling additional language packs for custom-deployment scenarios.

7. How does it handle fraud and proctoring?

A strong answer is explicit about which signals are captured and how they are used. Proctoring systems should provide context for a human reviewer, not function as automatic disqualification triggers.

Intrvio's optional proctoring layer monitors tab-switch events, browser visibility changes, and multiple audio source signals. These signals do not affect rubric scores — they are presented as context for the human reviewer only. Decisions about which candidates to flag remain with the human.

8. Does it write back to our ATS?

A strong answer clarifies which ATS platforms are supported, how upsert or update logic works, and the data mapping policy — which fields are written to the ATS.

Intrvio's integrations layer writes back to Greenhouse, Lever, Ashby, Workday, Workable, Recruitee, SmartRecruiters, TeamTailor, and JazzHR. Field mapping is configurable per integration: scorecard summary, rubric note, video link, and audit package URL are included. Bidirectional webhook flows synchronise candidate status updates in both directions.

9. Is it EU AI Act / GDPR / KVKK compliant?

A strong answer goes beyond compliance claims and lists specific mechanisms: technical documentation, a record of risk assessment, candidate notices, human oversight workflow, and data retention policy.

Intrvio provides the EU AI Act Annex IV technical documentation pack, a NIST AI RMF crosswalk guide ready for employer deployment, and comprehensive compliance documentation. Candidate notice flows cover GDPR and KVKK rights including deletion-on-request and portability. Technical documents are downloadable via the Annex IV document.

10. How is pricing structured?

A strong answer shows that pricing scales with usage volume and has a predictable growth curve that does not gate platform access. Per-seat fees and hidden activation charges are a red flag for HR teams budgeting tools.

Intrvio prices by interview volume: pay for interviews conducted, not seats. The first five interviews are free with no credit card. Pricing growth is predictable and linear as volumes and roles scale. See the pricing page for full details.

Red Flags to Watch For

If you notice any of the following, the vendor deserves closer scrutiny:

• No transcript export. Providing only a summary or aggregate score leaves indefensible records in an audit or challenge.
• Black-box scores. If a score has no transcript passage explaining where it came from, you do not know how it was derived.
• No human override path. Systems that do not allow AI output to be changed cannot satisfy the EU AI Act human oversight requirement.
• No compliance documentation on request. The EU AI Act requires ready technical documentation for high-risk systems; vendors who cannot produce it are shifting the compliance burden to you.