Automated Candidate Screening Calls
First-round phone screens are the most repetitive part of recruitment: the same six questions, hundreds of times, mostly to candidates who will not move forward. Veytrix AI runs that round as a real phone conversation, then returns a transcript, a summary and a score for every candidate — so your recruiters spend their day on the shortlist instead of building it.
The problem with the first round
A recruiter screening 400 applicants at fifteen minutes each — dialling, waiting, leaving voicemails, rescheduling — spends the better part of two weeks to eliminate most of them. The work is necessary but almost entirely mechanical, and it is done worst when the recruiter is tired: the hundredth call of the week is not asked the same way as the first.
How the automated screen runs
- Upload the list. A CSV of candidates with names and phone numbers, plus any fields you want the agent to reference.
- Write the screen in plain language. The opening line and the questions you want asked — notice period, current CTC, location preference, a couple of role-specific checks.
- The agent calls. Multiple calls run in parallel, within the calling window you set. Candidates who do not pick up are retried according to your rules.
- It has an actual conversation. It listens, follows up on vague answers, and handles the candidate asking questions back.
- You get structured results. Transcript, summary, extracted fields and a rating against your criteria — sortable, so the shortlist builds itself.
What to ask on an automated screen
The screens that work are the ones checking facts, not judging nuance. Good candidates for automation:
- Notice period and earliest joining date
- Current and expected compensation
- Location, relocation willingness, shift acceptance
- Specific tool or language experience, and for how long
- Whether they are still actively looking, and their other offers
Leave to humans: culture judgement, negotiation, anything requiring persuasion, and any conversation where a wrong answer damages the relationship with a strong candidate.
Screening in the candidate's language
For volume hiring outside metros, English-only screening quietly filters out capable people for the wrong reason. Running the screen in Hindi, Telugu, Tamil or another Indian language usually raises completion rates — candidates finish the call instead of abandoning it. See the India guide for the language and compliance detail.
How the rating works
After the call ends, the transcript is scored against criteria you define — not a generic "good candidate" score. If notice period matters more than years of experience for this role, weight it that way. The output is a number you can sort by plus the reasoning behind it, so a recruiter can check why someone ranked where they did rather than trusting a black box. There is a longer explanation in how AI call ratings work.
Being straight with candidates
Tell candidates it is an automated call. In practice it costs you very little — most people are fine with it for a first round, and some prefer it because they can take the call when it suits them. What damages trust is pretending otherwise and being caught. Have the agent answer honestly when asked, and give candidates a route to a human.
A sensible rollout
Run one requisition through it first. Read every transcript from the first fifty calls — that is where you will find the questions that were ambiguous or the follow-ups that missed. Fix the prompt, then scale. Teams that scale before reading transcripts generally end up re-screening.