"I just get a good feeling about him": what gut feeling really costs you

Confidence is not competence

Sales hiring has a built-in trap: the trait that impresses in an interview — smooth, high-confidence delivery — is also the trait that masks weak discovery. Confident talkers get hired, and then discover that prospects don't care how confident you sound if you never ask a useful question.

Meanwhile candidates with an accent, a career gap, or a non-traditional background get filtered out long before anyone hears them handle an objection.

Structure is the cheapest fairness tool available

Decades of selection research point the same way: structured, job-relevant work samples predict performance far better than unstructured interviews, and they shrink the gap caused by interviewer bias.

A simulation is a work sample. Same prospect, same objections, same rubric, every candidate. When two people score differently, you can read the transcript and see exactly why.

Make the scorecard boring and explicit

Define four or five criteria before the first call and weight them. Score each one with a quote as evidence. If a criterion can't be evidenced with a line from the transcript, it doesn't belong in the scorecard.

This also protects you: when a rejected candidate or a works council asks how the decision was made, you have an answer that isn't "vibes".

Keep a human in the loop

AI scoring should rank and evidence, not decide. Set a pass threshold, review the borderline band manually, and always let a human make the reject call on anyone close to the line.