Crystal Method builds hiring AI to assist, not replace, interviewers

Crystal Method's post is an implementer's account rather than a product launch. The company, which develops virtual humans, frames hiring as having three problems — interviewer workload, inconsistent evaluations, and candidate experience — and argues the realistic first step is supporting interviewers rather than having AI run interviews end to end. That choice is presented as a risk judgment: handing pass/fail decisions to AI is said to carry large risk on accountability and fairness.

The design follows from that: AI assists with questions and with organizing evaluation criteria, while the decision stays with people. On the candidate side, the company draws on technology from its interview-practice work that shows a speaker's expression, emotion and tension along a speech timeline, and says the candidate-side and employer-side tools can share the same dialogue and multimodal foundation. The company explicitly avoids asserting figures such as emotion counts or accuracy as product specifications.

A recurring point in the post is that the more AI is used in hiring, the more important it becomes to keep a state where people can explain why an evaluation was reached — an argument the company makes from its own build experience, not from published results.