Skip to content

A second pair of eyes for every online interview.

Sentry follows the signals a recruiter can’t track while running a conversation, and marks the moments worth a second look.

Illustrative interface. Signals are prompts for a human review, not verdicts.

Want to see it in your interview flow?

Book a short call. We’ll walk through how Sentry fits your interviews and answer the hard questions about accuracy and privacy.

01

Why online interviews break

A video call shows a face in a box. What shapes the answers can sit just outside it.

  • Something just off-screen

    Notes, a search tab, a second monitor. The eyes drift to the same spot before every answer.

  • An AI that listens along

    Tools now transcribe the question and draft an answer in real time. The candidate reads it back a few seconds later.

  • Someone else in the room

    A helper off-camera feeds answers by voice or chat. Glances keep going to the same place.

  • A voice that doesn’t match the face

    Someone else answers while the candidate barely moves their lips — or the lips and the audio drift out of sync.

No recruiter can watch for all of this and run a good conversation at the same time. Sentry can.

02

What is Sentry?

An AI layer for online interviews. It notices unusual behavior on camera and hands the recruiter evidence, never a verdict.

  1. The interview

    The candidate joins the video interview as usual and is told Sentry is on.

  2. Signals

    Sentry reads gaze, answer timing, facial movement and head position from the camera.

  3. Moments

    Unusual patterns become moments: a timestamp, a short clip and a plain-language reason.

  4. Your decision

    A person reviews each moment in context. Sentry never rejects anyone.

03

How it works

Sentry reads four families of signals and weighs them together. One signal alone proves nothing, so Sentry looks for patterns that repeat and line up.

Gaze direction

Where the eyes go, how long they stay, and whether they return to the same off-screen spot before each answer.

On its own, it could be: thinking, recalling, glancing at notes the interviewer allowed.

Response timing

The gap between question and answer, and how delivery changes: a long silence followed by an unusually even, fluent pace.

On its own, it could be: careful thinking, a second language, a slow connection.

Face & lips

Facial movement, and whether lip movement matches the audio — for example, speech heard while the lips stay still.

On its own, it could be: glasses, low light, a beard, an odd camera angle, audio lag.

Head position

Turns and tilts that keep repeating toward the same direction, or a head held unusually still while the eyes move.

On its own, it could be: a second monitor for the call itself, a stiff neck, a habit.

04

Fairness & false positives

A wrong flag can cost an honest person a job. We’d rather miss a moment than mark the wrong one.

  1. What was seen, in plain words. No scores and no “cheating probability” — just the behavior, when it happened and how long it lasted.

  2. A pattern, not a glance. A moment needs duration or repetition. A single look away is never marked.

  3. Innocent explanations, up front. Every moment lists ordinary reasons for the same behavior, before anyone judges it.

  4. A person decides. Reviewers keep or dismiss each moment. Sentry never rejects, ranks or scores a candidate.

05

Privacy & ethics

Watching people is a serious thing to build. These are the rules Sentry is built by — including the lines it won’t cross.

Sentry reads, during the interview

  • Where the eyes look, and for how long
  • The timing between question and answer
  • Facial and lip movement
  • Head position and movement

Sentry never

  • Identify who someone is — no face recognition
  • Guess emotions, personality or honesty
  • Reject, rank or score candidates
  • Sell interview data, or train on it without permission

Candidates are told before the interview starts. There is no hidden mode.

Let’s look at your interview flow together.

A short call: your process, and an honest take on where Sentry helps and where it doesn’t.