How do consulting firms proctor live case-study interviews?

TL;DR: Case interviews are especially exposed to AI cheating because they're structured, predictable, and text-heavy — exactly what an AI copilot handles well when read off a second screen. Firms proctor them by verifying candidate identity up front and running continuous behavior and environment monitoring for the full case, not by adding friction to the case itself.

By Pinal Dave | Last updated: 2026-08-05

The claim

A case interview is a live, multi-step reasoning exercise delivered over video — structurally similar to a technical interview, and vulnerable to the same off-screen AI coaching Fabric found in 38.5% of interviews it analyzed.

The evidence

Karat's research found 80% of candidates use LLMs during banned technical assessments, and in-person interview requests jumped from 5% to 30% of roles between 2024 and 2025 specifically because remote formats became easier to game. Case interviews share the same live, structured, video-delivered format that makes coding tests exploitable — the incentive to run an AI copilot in a second window is just as strong when the "test" is a market-sizing or profitability case instead of a leetcode problem.

Step-by-step: proctoring a case interview

Verify candidate identity (ID + selfie match) before the case brief is shared.
Keep AI Meeting Proctor active for the full case, not just the opening intro — most coaching happens during the analysis, not the small talk.
Watch for gaze patterns consistent with reading a framework or numbers off a second screen instead of working them out live.
Flag unnaturally fast, fully-formed answers to open-ended market-sizing questions — genuine case reasoning has visible thinking, false starts, and self-correction.
Review the trust report alongside the interviewer's case scorecard before advancing the candidate.

FAQ

Is a case interview more vulnerable to AI cheating than a standard behavioral interview? Structurally, yes — its predictable format (frameworks, market-sizing math, structured recommendations) is exactly what LLMs handle well, similar to why coding tests are heavily targeted.

Does proctoring slow down the case delivery? No — the AI agent joins the call like any other participant; the case is delivered exactly as it would be otherwise.

Can proctoring distinguish a candidate using a legitimate framework from memory versus reading one off-screen? Gaze and environment signals combine to flag reading patterns; the interviewer still makes the final judgment call using the trust report as evidence, not as an automatic verdict.