Through this year’s graduate hiring season at the big banks, the first real filter was no longer a person. It was a scoring model.
Candidates sat short AI interviews — five to twenty minutes, a handful of recorded questions, no interviewer on the other end. Some finished and heard nothing at all. “After the AI interview, it just disappeared into the sea,” one candidate said of her elimination. She never learned how she had scored, or against what.
The banks describe the systems as efficient: thousands of interviews compressed into a week, at a fraction of the cost of a human panel. They say the scoring weighs what a candidate actually says — logic, clarity, fit — and not lighting or camera angle.
What candidates describe is closer to answering a wall. The answer is measured, the score is not shown, and when an application is dropped, the only reply is silence. Those who asked for their results were told, in effect, to wait for a notification that for many never came.
The black box cuts both ways: no feedback when it goes right — and none when it goes wrong.