AI Interviews · Proctoring · Technical Screening · Hiring Process · 100Networks Workspace

How AI Interviews Work, and What Proctoring Actually Checks For

A live AI interview can cover a spoken conversation, live coding where the code actually runs, and a system-design question in one proctored session. What it tests, what proctoring watches for, why a flag should never fail a candidate on its own, and where a person decides.

Kushal Agarwal· Co-founder, 100Networks··5 min read

An AI interview is a first-round interview run by an AI interviewer: questions built from the job description, follow-ups based on the answers, and a written report with scores and reasons that a person reviews before deciding anything. One session can cover a spoken conversation, live coding where the code actually runs, and a system-design question, so it can stand in for what usually takes three separate first-stage rounds, each on its own calendar.

Proctoring is what lets a team trust that report. Done badly, it fails good candidates over a frozen webcam. Done well, it gives the reviewer context and leaves the judgment to them. Here is what happens inside an AI interview, what proctoring watches for, and where the line is.

What a live AI interview covers

A single session can evaluate competencies that teams usually split across rounds:

  • Conversation. A spoken interview about the candidate's experience, decisions and judgment, with follow-up questions that go deeper when an answer is thin.
  • Live coding. Real problems in a proper code editor, where the candidate's code is run against test cases. They write, run and fix code rather than describing it in the abstract.
  • System design. An open-ended architecture question worked on a shared whiteboard, so the diagram and the reasoning behind it are both part of the evaluation.

Not every role needs all three. A fifteen-minute screening interview can skip coding entirely; an engineering deep dive can run for forty-five minutes with harder follow-ups. Questions are generated from the role and pitched at its seniority, not pulled from a generic bank.

Because it is one continuous session, the candidate experiences one interview, not a three-week relay of scheduling emails.

Why one session can replace three rounds

The traditional first stage exists for good reasons: you want signal on problem-solving, on design thinking and on how someone works. Splitting those into three rounds is mostly a logistics constraint: different interviewers, different calendars, different days.

An AI interview removes the logistics, not the rigour. It covers the same ground in one sitting and returns a structured report a person can act on the same day. The senior-engineer hours that used to go on first-round screens go where they matter: final-stage conversations with candidates who have already shown they can do the work.

What it does not replace is that final stage. Whether someone will thrive on your team, and whether they want to join it, is still a conversation between people.

What proctoring checks for

Proctoring exists to give the reviewer integrity context: did anything unusual happen during the session? In 100Networks AI interviews, these moments are recorded with a timestamp:

  • leaving full screen
  • switching tabs
  • pasting text
  • no face, or more than one face, in view
  • looking away from the screen

Each event appears in the report next to the scores, at the point in the session where it happened.

Why a flag should never fail a candidate on its own

There is a wrong way to proctor: end the session, or mark the candidate as failed, the moment the system sees something it does not like. That punishes people for a dropped Wi-Fi connection, a noisy flat or a webcam that froze, and it turns a skills assessment into an anxiety test.

So integrity events are logged for the reviewer, not acted on automatically. Concretely, proctoring:

  • does not end the interview when an event is recorded;
  • does not reject anyone on a flag;
  • does not treat every anomaly as cheating. Technical hiccups are common and expected.

A momentary glitch does not sink a strong candidate. A genuine pattern, such as repeated pasting during a coding question, is there in the report for a person to weigh.

The AI evaluates; a person decides

This is the principle that makes AI interviews trustworthy. The interview produces scores, the evidence behind them and the integrity events, and a recruiter or hiring manager makes the call with all of it in front of them. For why asking every candidate the same way matters, see structured interviews with AI; for the written alternative to a live round, see AI take-home assessments.

Where 100Networks uses AI interviews

On a managed search. When you send us a role, specialist recruiters who have filled it before work the search with our in-house sourcing desk, and a bespoke, proctored first round built on your rubric is part of how the shortlist is assessed. You interview the shortlist and pay 8.33% of first-year salary only when the person joins.

In the 100Networks Workspace. For the roles your team fills itself, the 100Networks Workspace is an AI hiring team, and its Interviewer runs AI interviews and take-homes built from your job description, scores them against your bar and writes them up after every round. The candidate receives an invite link and opens the interview when ready; booking the live rounds that follow is the Scheduler's job. Your team keeps a library of interview templates, from a quick no-coding screen to an engineering deep dive, and sees each report with its scores, reasons and integrity events. Nothing reaches a candidate, including the invite, until your team approves it. For live rounds run by your own interviewers, the Workspace's Notetaker takes the notes.

What it means for your funnel

  • Candidates get a faster, single-session first round, a skills-based evaluation instead of a keyword filter on their CV, and no penalty for a technical glitch.
  • Teams get consistent, structured evidence on every candidate, the integrity context to trust it, and their senior engineers' time back.

If your funnel is losing time between rounds, an AI interview is one of the fastest fixes; see how to reduce time to hire with AI. If the problem is that the right people never enter the funnel at all, that is a reach problem, and what is 100Networks explains how the marketplace solves it.

Frequently asked questions

What is an AI interview?
An AI interview is a first-round interview run by an AI interviewer instead of a person. The candidate opens an invite link when ready, the interviewer asks questions built from the job description and follows up on the answers, and the session ends in a written report with scores and the reasons behind them, which a recruiter or hiring manager reviews before deciding anything.
What does an AI interview actually test?
Depending on the role, one session can include a spoken conversation about experience and judgment, live coding in an editor where the candidate's code is actually run against tests, and a system-design question worked on a shared whiteboard. Short screening interviews can skip coding entirely. Questions are generated from the role and pitched at its seniority.
Can one AI interview really replace three rounds?
It can replace three first-stage rounds, typically a phone screen, a technical screen and a behavioural screen, because it covers those competencies in one sitting and removes the scheduling between them. It does not replace final-stage conversations with the team, which should still be run by people with already-qualified candidates.
What does proctoring check for in an AI interview?
In 100Networks AI interviews, proctoring records integrity events with a timestamp: leaving full screen, switching tabs, pasting text, no face or more than one face in view, and looking away. The events are shown to the reviewer next to the scores. They are evidence for a person to weigh, not an automatic verdict.
Will a flaky webcam or a dropped connection fail the candidate?
No. Integrity events are logged for the reviewer rather than ending the session, so a frozen webcam or a brief disconnect is recorded, not punished. A strong candidate is not failed over a technical problem outside their control, and a genuine pattern is still there for a person to see.
Does a person still make the hiring decision after an AI interview?
Yes. The AI interview produces scores, the evidence behind them and any integrity events; a recruiter or hiring manager reviews the report and decides. The AI evaluates, it does not hire. In the 100Networks Workspace nothing reaches a candidate, including an interview invite, until the team approves it.
How does 100Networks use AI interviews?
In two ways. On a managed search, 100Networks runs a bespoke, proctored first round built on the client's rubric as part of the shortlist it delivers. In the 100Networks Workspace, an AI hiring team for the roles a company fills itself, the Interviewer agent runs AI interviews and take-homes built from the job description, scores them against the team's bar and writes them up after every round. The candidate gets an invite link and opens the interview when ready.