When a candidate submits a take-home case study, you're making an assumption: that the work represents their thinking. That assumption is increasingly untenable. A large share of unsupervised submissions now show clear signs of AI generation.
This isn't a moral judgment about AI use. It's a measurement problem. If you're trying to evaluate a candidate's reasoning and communication ability, and the work was generated by a model, you're evaluating the model, not the candidate.
Why we proctor instead of guess
The most reliable way to know the work is the candidate's own is to watch it happen. Chesky runs the case study as a proctored, recorded session. The candidate solves a real problem live on a work canvas while screen, camera, and audio are captured. There is no take-home window to outsource the thinking.
On top of the recording, Chesky weighs converging integrity signals: how the solution took shape on the canvas over time, the candidate's spoken reasoning as they work, and behavioral patterns across the session. No single signal is conclusive. The system looks for convergence across all of them before it raises an integrity flag.
A read, not a verdict
Chesky reports its integrity finding as a qualitative read with the evidence attached, never as a number and never as an automatic pass or fail. A weaker signal prompts a follow-up conversation. A strong signal is a clear flag your team can review against the recording itself.
The decision always remains with your team. Chesky surfaces what it observed and points you to the exact moment in the session that supports it. You decide what it means.
Why this matters for culture
Companies building high-performance remote cultures are selecting for people who do original thinking under pressure, not people who are good at prompting AI tools. The case study is one of the few evaluation moments where you can directly observe a candidate's reasoning process. Protecting the integrity of that signal matters.