Hi everyone,
Most AI hiring tools are trained to find candidates who look like the people you already hired. That is what the training data is. Past hires, past decisions, past definitions of a strong profile. The model gets very good at recognising the pattern it was shown.
For a hiring team that creates a quiet problem. The roles that are hardest to fill are usually the ones where your existing pattern is the weakest guide, because the candidates who can actually do the work do not resemble your previous hires closely enough to score well. So the tool is most confident exactly where you most need it to surprise you, and least useful for the searches that are genuinely difficult.
The teams handling this well seem to treat AI screening as a way to widen the top of the funnel rather than narrow it. They use it to surface people they would have missed, then keep human judgment on who advances, instead of letting the model's confidence stand in for a decision. It is a smaller shift in how the tool is positioned, but it changes what comes out the other end.
How are your teams thinking about this? Do you trust AI to rank candidates, or only to surface them, and where did you land on that line?