A proctored AI interview should evaluate skills fairly without failing good candidates over technical glitches. On 100Networks, the interview tests real competencies — coding, system design, and behavior — while proctoring runs in soft-enforced mode: potential integrity issues are logged for a human reviewer, never used to auto-abort the session. The result is integrity signals without the false positives that make automated proctoring feel hostile.
Here's exactly what's monitored, what's tested, and where the line is.
What's actually tested (the real signal)
The interview itself evaluates several competencies in one proctored session:
- Coding — real problems in a Monaco editor (the VS Code engine) with sandboxed execution across 15+ languages, so the candidate runs and tests their code, not just writes it.
- System design — open-ended architecture questions, with whiteboard analysis of the candidate's diagram and reasoning.
- Behavioral — structured conversation via real-time speech, scored against the role's competencies.
The output is a structured, JD-aligned score — the substance of the evaluation.
What proctoring checks for — and why "soft-enforced" matters
Proctoring exists to give the reviewer integrity context: was anything unusual during the session? But there's a right and a wrong way to do it.
The wrong way is hard enforcement — automatically failing or kicking out a candidate the moment the system sees something it doesn't like. That punishes people for a dropped Wi-Fi connection, a noisy room, or a webcam that froze. It turns a skills assessment into an anxiety test.
100Networks uses soft enforcement: potential violations are logged as flags for the human reviewer, not acted on automatically. The reviewer sees the flags alongside the score and the recording, and uses judgment. A momentary glitch doesn't sink a strong candidate; a genuine pattern is there for a human to weigh.
What it does not do
- It does not auto-reject on a flag.
- It does not replace human judgment — flags are evidence, not verdicts.
- It does not treat every anomaly as cheating; technical hiccups are common and expected.
The principle: AI evaluates, a human decides
This is the same control model that runs across 100Networks. The AI does the time-consuming work — running the interview, scoring the competencies, surfacing integrity flags — and a recruiter or hiring manager makes the final call with all of it in front of them. Combined with Pilot, which proposes every action for human confirmation, the candidate is always judged by a person, not silently filtered by a machine.
Why this produces better hiring
- Candidates get a fair, single-session, skills-based evaluation — not a proctoring gauntlet that fails them on technicalities.
- Teams get consistent, structured scores plus the integrity context to trust them — without the false-positive noise of hard-enforced proctoring.
See how interviews fit the rest of the platform in the product overview, or read the deeper mechanics in How AI Interviews Work.