Every applicant interviewed, scored, and shortlisted.

A recruitment agency lives or dies on speed, and a busy role buries it. Roughly a thousand applicants land on one opening, and turning that pile into a hire is brutal arithmetic. The realistic version: two days cutting it down to a shortlist of 50, then 50 first-round interviews at 30 minutes each, then 20 minutes scoring and writing up each one. That is close to 58 hours — seven working days of one recruiter — before a single second-round conversation.

And that is the fast version, because it interviews almost no one. Speaking to all 1,000 applicants by hand would take a recruiter more than a hundred working days, so it never happens — 95 in every 100 are cut on a ten-second CV skim, never actually heard. The applicant who reads poorly on paper but interviews brilliantly is lost before anyone notices.

What does get done is inconsistent. The candidate interviewed first thing Monday gets a sharper recruiter than the one called at five on Friday; questions drift and scoring wanders. The agency wanted the first round rebuilt — every applicant genuinely interviewed, every one scored the same way, and a ranked shortlist of the people worth a recruiter’s time, without the seven days of manual work in front of it.

We built an AI voice agent on Vapi that calls each applicant and runs a real one-to-one interview — not a yes/no qualification script. It works from the role’s interview guide, asks a follow-up when an answer is thin, and holds a natural conversation at sub-second latency, so the candidate is genuinely interviewed rather than read a form.

Every interview is transcribed and evaluated against a rubric built for the role — the competencies that matter, weighted the way the agency weights them. The agent turns a free-form conversation into structured signal: a score per competency, the evidence behind it, and an overall rating. Because every candidate answers the same questions against the same rubric, the scores are comparable in a way human phone screens never are.

When applications for a role close, the agency does not get a spreadsheet of calls to wade through. It gets a ranked shortlist — the top five candidates, each with their score, the reasoning, and a transcript a recruiter can read in two minutes. The recruiter starts at the shortlist instead of working their way toward it.

The agent interviews the whole pool in parallel and around the clock, so a candidate in another time zone is interviewed at nine in the evening their time, not skipped. Borderline scores and off-script questions are flagged for a human rather than guessed, and every interview is recorded and monitored so the rubric keeps improving against real hiring outcomes.

Every applicant now gets a real interview — all 1,000, not the 50 a recruiter had hours for. Each one is consistent, scored against the same rubric, and available to take at any hour. The seven days of CV triage and first-round interviews in front of every role are simply gone — and the agent has run 800+ interviews in production doing it.

Instead of working through a list, recruiters now open a ranked top-five shortlist with the scoring and transcripts already attached — usually the morning after a role closes. The agency moves on its strongest candidates in hours instead of weeks, and stops losing them to faster competitors.