Step 1: Invitation and scheduling
The flow usually starts with an email. When the employer advances your application to an interview stage in their ATS, Intrvio creates an interview for you and sends a unique link. The link is single-use — no one else can join in your place, and it cannot be reused.
Most interviews are async: you get a link to complete within 72 hours, on your own schedule. Some employers use a fixed time slot for qualification rounds; in that case you see a real appointment with a calendar add link. The invitation tells you the interview duration, language options, and which transparency and human review rights you have.
Step 2: Setup — camera, mic, environment check
When you click the link, your browser walks you through a setup screen. This checks three things: that your microphone works (with a level meter), that your connection is stable enough, and optionally that your camera is in frame. All of this is browser-based — you do not need to install anything.
The same screen shows a candidate notice: a plain-language statement of what data is collected and how it is processed. The EU AI Act gives candidates the right to an explanation of the main elements of decisions made by AI systems, so you see this in advance. You can download the notice or request deletion before the interview begins.
Step 3: GAIA conducts the conversation
When you are ready, the interview begins. GAIA — our voice AI interviewer — introduces itself, explains what to expect, and then asks the first question. From that point on, it is a normal conversation.
Three things distinguish GAIA from a script reader. First, real-time turn-taking: GAIA predicts when you have finished and does not interrupt your answer. It allows seconds of silence mid-sentence because it does not assume a pause means you are done. Second, when you barge in mid-sentence to clarify, GAIA stops and listens. Third, follow-ups are real, not canned: a model picks up your 'principled debate' answer and goes deeper.
One technical nuance: latency is measured continuously. Between your answer and GAIA's next utterance, there is usually less than 800 ms. The human brain perceives speech as continuous around 300 ms; good voice AI interviewers target that range. The rhythm is what makes the interview feel like a conversation, not a survey.
As the interview progresses, GAIA does two things quietly: it builds a transcript that captures exactly what you said, and it scores it against the pre-defined rubric. You do not see this; you are just having a conversation.
Step 4: Scoring and proctoring
When the interview ends, GAIA produces a scorecard in seconds, not minutes. The scorecard breaks down each rubric criterion — say, problem solving, communication, ownership, systems thinking — and attaches a score plus a rationale that quotes from your actual answers. This evidence-anchored design is deliberate: a human reviewer can click into any score to see exactly where it came from.
Simultaneously, an optional proctoring layer runs. If the employer enabled it — typically for regulated industries or high-stakes roles — the system checks the environment: multiple people present, screen-share indicators, or signals of cheat-tool overlays. Proctoring signals do not affect rubric scores; they are context for the human reviewer.
Step 5: Employer review
On the employer side, the hiring manager or recruiter sees the interview within an hour. The review screen is a three-pane layout: per-question playback on the left (they can click any question and hear your answer), the full transcript in the middle, and the AI summary plus rubric scorecard on the right.
One important design choice: rubric scores are overridable. If a hiring manager believes the 'systems thinking' score is wrong and they have a justification, they can override it. The override is logged into a signed audit log and usually forms part of the EU AI Act deployer evidence pack. This is what makes the AI an instrument for evidence, not a decision maker.
Step 6: Outcome and rights
The hiring decision rests with the human, not GAIA. Rejected candidates can use the channel under EU AI Act Article 86 to request an explanation of which elements went into the decision. In most cases, the ATS contacts you with next steps. If you ask for extra transparency, the employer can share your rubric summary and the human review actions.