AI video interviews you can defend, not just trust

No. Truffle is a decision-support platform. AI Match surfaces evidence against the rubric you defined. A human reviewer makes every hire and no-hire call, captures their reasoning, and can override any AI score. NYC Local Law 144 specifically regulates tools that 'substantially assist or replace discretionary decision-making.' Truffle is built around the architecture that keeps the discretion with the human. That said, regulatory interpretation varies, and you should confirm your own compliance posture with counsel for the jurisdictions where you hire.

A black-box scoring tool gives you a number with no defensible reasoning. Truffle gives you a per-criterion match breakdown, the model's reasoning for each criterion in plain language, and a link to the exact transcript moment the AI cited. Click any criterion, jump to the source moment. Disagree with a score, override it. The whole audit trail archives per candidate. The number is the start of the conversation, not the verdict.

Yes, in three layers. First, the overall match percentage breaks down into per-criterion alignment scores. Second, each criterion has a written reasoning paragraph explaining what the AI weighted and why. Third, every reasoning paragraph cites a timestamp in the transcript so you can hear the candidate's exact words. If you disagree with the score, you override it and the override stores alongside the AI output for the audit record.

AI Check flags responses with patterns consistent with AI-assisted phrasing, off-screen reading, or copy-pasted output. It is a context signal, not a verdict. The model can produce false positives, especially for candidates whose natural style happens to match generated text patterns. Use it to ask a sharper follow-up question, weight the response with caution, or schedule a quick live verification. Do not use it as a basis for automatic disqualification.

Generic AI hiring scores rank candidates against a model trained on aggregated data, then output a single ranking number. The model decides what matters. Per-criterion scoring inverts that: you write the rubric, the AI scores against your criteria, and the output is alignment-per-criterion plus reasoning. A great fit for one role's rubric can be a poor fit for another. The criteria you defined drive the math, not someone else's median hire.

Yes. Generate a read-only candidate link, send it via email, the hiring manager reviews the transcript, the rubric, the per-criterion score breakdown, and the AI reasoning, then leaves a structured rating and notes. The rating aggregates into the consensus view alongside other reviewers. No login, no provisioning, no IT ticket. Independent ratings reduce groupthink and the consensus view shows where reviewers agree, disagree, and converge.

The Illinois AI Video Interview Act requires that you notify candidates that AI may be used to evaluate their responses, get consent, explain how the technology works, and limit who sees the recordings. Truffle ships consent and notification flows you can configure per position. Colorado SB 205, the EU AI Act high-risk classification, and proposed federal rules push the industry toward similar transparency requirements. Truffle's per-criterion reasoning, transcript citations, and audit trail are built for this regulatory direction. As always, confirm your own approach with counsel.

No. Truffle never auto-rejects. AI Match surfaces alignment evidence, ranks responses against the rubric, and generates summaries and Candidate Shorts. A human reviewer advances, holds, or rejects every candidate. There is no setting for 'auto-reject below threshold' because the architecture is designed around the assumption that a human owns every decision.

Three layers. First, the rubric is yours, not a generic hiring model. The criteria you write down drive the scoring math. Second, every score is per-criterion with reasoning, so you can spot a model output that doesn't match the response and override it. Third, divergence reports show where human reviewers consistently disagree with AI on a role family, so you see model drift before it becomes a pattern. We do not claim AI eliminates bias. No tool can. We claim AI surfaces evidence in a way that human reviewers can audit, override, and adjust.

Three minutes for the first one. Paste the job description, accept or edit the AI-suggested questions, write or edit the rubric, add your branding, share the Position Link. Subsequent positions reuse the rubric library so the second one takes about a minute.

Plans start at $49 per month (Starter), credit-based. One shared monthly pool of credits across resume screening (1 credit), assessed candidates (2 credits), and one-way interviews (5 credits), with credits rolling over on Core and up. Unlimited positions, unlimited team members, every AI feature on every plan, no per-seat fee. 7-day free trial, 30 credits, no card.