AI first-round interview
An AI first-round interview is an initial candidate screen conducted by a voice or chat AI agent that asks job-relevant questions, follows up adaptively, and scores answers against a recruiter-defined rubric.
Direct answer
An AI first-round interview replaces the human-conducted initial screen with a voice-or-chat AI agent. The agent introduces the role, asks four to six job-relevant questions, follows up on weak or vague answers, and stops within a fixed time cap (usually 10–15 minutes). The output is a transcript plus a rubric-scored report the recruiter reviews before deciding who advances. It matters because first-rounds are where recruiter time goes to die — most of the conversations end in no-advance, and the questions are repetitive enough that a structured agent can handle them consistently. Who needs it: founders, in-house recruiters, talent scouts, and hiring managers currently doing first-rounds themselves.
In detail
How the mechanism works
A voice AI first-round runs three loops in real time: speech-to-text on the candidate channel, a large-language-model turn that decides the next question, and text-to-speech on the agent channel. The LLM is constrained by a system prompt (the role brief, the rubric, the time cap) and by the turn-by-turn transcript of what has been said. That constraint is what makes the agent stay on rubric and ask sensible follow-ups instead of drifting into general chat.
When it applies
The pattern fits roles where the first-round is a 10–15 minute fit-and-signal check — the kind of call that ends in "let's schedule a longer technical" or "not this time, thanks." It fits less well where the first-round is itself the deep evaluation (e.g. a take-home review walk-through, an executive-search assessment, a sensitive judgement call). For those, the AI screen is upstream of the human round, not a replacement.
Common misconceptions
Three patterns to watch. First, AI first-round is not a one-way video recording — the candidate is in a real-time conversation, not talking to a camera. Second, it is not a decision-maker — the agent scores against a rubric, the recruiter still chooses who advances. Third, it is not the same product category as the enterprise structured-video tools (HireVue, Spark Hire). Those handle volumes of 1,000+ hires a year with proctoring and an ATS-native workflow. An AI first-round agent is for the human currently doing the screen themselves.
Related concepts
A structured AI first-round is a particular implementation of a structured interview. The conversation surface is voice AI screening. The output is a rubric-scored report. In NYC, if the score "substantially assists" an employment decision, the system also qualifies as an AEDT under Local Law 144.
How Vettika handles it
Vettika runs a 12-minute voice first-round end to end: define the rubric, send the candidate one link, get back a transcript plus per-criterion scoring. The candidate keeps a copy of the transcript and their own score — no black-box rejection. See the voice agent product page for the technical surface.
Try a first-round interview, no card.
First three are free. Twelve-minute cap. The candidate keeps the transcript.
Run your first 3 free