Automated interviews: what they are and when to use them

What an automated interview is

"Automated interview" covers any screening interview where software asks the questions and the candidate answers without a live interviewer. The goal is the same in every case: take the repetitive first conversation off a recruiter's calendar, and give every applicant the same chance to speak.

The word "automated" can mislead. It describes who asks the questions. It does not have to describe who decides. A well-run automated interview still ends with a person reading the results and making the call.

Three common formats

The formats overlap. Our guide to one-way video interviews covers the recorded format in depth, and our AI interview software overview compares recorded video, AI voice and live conversational AI side by side.

When automated interviews make sense

• High applicant volume. When a hundred people apply for ten jobs, a screening call for each is the bottleneck. See high-volume hourly hiring.
• Hard-to-schedule candidates. People working shifts or in other time zones can answer at 10 pm.
• Roles where consistency matters. Asking everyone the same questions is the heart of a structured interview.
• Agencies and small teams. One recruiter cannot run twenty first-round calls a day for several clients. See staffing agencies.

They are a poor fit when:

• You have a handful of applicants and a quick call is faster.
• The role needs open conversation from the first minute, such as a senior hire where the relationship is the point.
• You cannot give candidates a clear notice and a way to ask for another route.

Fairness and compliance

Using software in hiring brings responsibilities, and they stay with the employer. Practical habits that help in any jurisdiction:

• Ask what is scored. Get a written list of what the tool evaluates and ignores. If a vendor scores faces, emotions or voice tone, ask how that relates to the job.
• Keep a person in the decision. Tools that only inform a reviewer carry less risk than tools that auto-reject.
• Be able to show your evidence. You should be able to go from any score to the answer that produced it.
• Tell candidates. What is recorded, whether AI is used, who sees it and for how long.
• Offer an accommodation route. Not everyone can use a camera or a microphone.

Some places write the rules down. New York City's Local Law 144 covers automated employment decision tools used to screen candidates for jobs in the city. According to the city's Department of Consumer and Worker Protection, a tool must be subject to a bias audit within one year of use, information about the audit must be publicly available, and candidates must get notice, which the department's guidance says is ten business days before use. The employer or employment agency carries these duties, not only the vendor. Our Local Law 144 guide goes through the primary sources. Other states and cities have their own rules, and they change, so confirm yours with an employment lawyer. This page is not legal advice.

Where Infyva stands
Infyva has not had an independent bias audit yet, and does not have a SOC 2 report. On every plan, company owners and admins can export impact-ratio data from optional demographic answers for an independent auditor. If you hire for roles where an audit is required, plan for it before you use any tool there. Our AI and fairness page explains exactly what is and is not evaluated.

How Infyva does it

Infyva's standard interview is an automated AI voice interview. Here is how it runs:

1. Questions from the job description. Infyva writes one set for the job, and every candidate gets that same set in the same order. You can edit any question. Resume tailoring is opt-in.
2. An email invite. The candidate opens the link in a browser. There is no app and no scheduling.
3. Checks and practice. A microphone and camera check, a notice of what is recorded, and a practice question.
4. Nova asks, the candidate answers. Each question is spoken aloud. The browser records each answer on camera by default, or audio only if the job is set that way. One take per answer.
5. Scoring. Each answer is transcribed and scored against the job's competencies, with a written explanation. Scores come from what is said. Appearance, facial expressions, accent and tone of voice are not scored, and video is never sent to an AI model.
6. Integrity notes. Flags are tied to the answer they happened on, for a person to review. They never change a score.
7. A person decides. Infyva never rejects anyone. Under three answered questions, the report says there is not enough evidence to recommend a next step.

The one automatic step is optional and off unless you turn it on: inviting candidates above a resume score you choose. It only invites.

Infyva also offers Live AI interviews, a booked real-time conversation where Nova asks follow-up questions. They are a different format with a different tradeoff: more adaptive, less identical from candidate to candidate.

Phone calls
Candidates can also interview by phone call: the AI interview by phone on Pro and Enterprise, and Live AI by phone on every paid plan (US and Canadian numbers). Nothing is recorded until the candidate agrees to the notice by keypress, and there is no replay or retake. Camera and browser checks do not exist on a call, and the report says so.

Questions to ask any vendor

• Where do the questions come from? Written by you, picked from a library or generated from the job description? Can you edit them, and does everyone get the same set?
• What is scored, and what is ignored? Ask for the list in writing.
• Can I see the evidence? A score should link to a recording or transcript.
• Does it reject people? Or does it inform a person?
• What are candidates told? Ask to see the notice and take the interview yourself.
• What does it cost when candidates do not finish? Per seat, per interview or unlimited? Are unfinished interviews charged?

Common mistakes

• Making it too long. Candidates give up. A handful of well-chosen questions beats twenty.
• Treating the score as the decision. The score is a pointer. Read the answers behind it.
• Different questions for different people. That defeats the point of a structured process and makes scores hard to defend.
• No plan for candidates who cannot use it. Decide in advance who handles accommodation requests.
• Skipping the human step. Automated interviews should shorten the path to a human conversation, not remove it.

How to tell whether it is working

Decide what you want before you start, and measure it against your own numbers from before. Useful measures are the time from application to first screen, how many recruiter hours the first round takes, the share of invited candidates who finish, and how often the people you advance succeed in the live round. Compare against how you hired before. Do not rely on a vendor's averages. Your roles and your candidates are your own.

Keeping the human step

The strongest argument for an automated interview is also its limit. It is good at asking everyone the same thing and at giving your team something to read. It is not good at judging whether someone will fit a team, or at weighing a surprising answer in the context of a life. That is why the output of an automated interview should be a better starting point for a person, not a verdict.

In practice, this means three things. Read the reasoning, not only the number. Treat a flag as a question to look into, not a finding. And give candidates a way to reach a human, whether that is a review request, an accommodation or a plain email address on the invite.

Pricing

Plans include credits, not seats. An AI interview uses 1 credit when a candidate answers. Every new company starts with a 14-day Pro trial with 20 credits. Starter is $39 a month for 30 credits, Pro is $149 a month for 150. Details are on pricing, and the demo lets you take an interview as a candidate.