What AI evaluation criteria are

The AI proposes criteria for scoring the candidates in your list; you rank them by relevance and decide which ones count.

Updated

Step 3 of creating a list is AI Variables, and what lives inside it are evaluation criteria: how you tell the AI what to look for in each candidate in order to score them.

From the role description you wrote in step 1, the platform proposes the criteria itself. You review them, rank them and add whatever's missing.

What a criterion looks like

Each criterion is a short statement ( the thing you want validated ) plus three traffic-light levels that define how it's scored:

  • Green: what has to be true for this to count as a yes.
  • Amber: the half-way case, when the evidence isn't clear.
  • Red: what rules the candidate out.

For a data engineering role, for instance, the AI proposes something like:

4+ years of relevant Data Engineering experience

  • Green: has 4+ years of professional data engineering experience clearly shown in work history
  • Amber: has relevant experience but dates are unclear or within about one year of four years
  • Red: clearly has less than three years of relevant professional experience

That rubric is what makes scoring consistent between candidates: the AI doesn't decide on the fly what counts as "enough experience": it reads it off your criterion.

Order matters

The screen says it outright: drag to rank them by relevance, the top one weighs most. It isn't decoration: the order changes the candidate's final score.

Put what actually decides a hire at the top, and leave the nice-to-haves at the bottom.

What you can do with each criterion

  • Untick the checkbox to leave it out of this list without deleting it.
  • Edit it if the definition isn't yours (the AI's thresholds are a proposal, not a law).
  • Remove it with the ×.
  • Add your own in plain language, using the field below → how to add your own criteria.

When they get applied

When you hit “Confirm and generate”. The AI scores the candidates in the list against the active criteria, and the result lands as columns in the list, ready to sort and filter.

Always read the proposed criteria before confirming. They come from your role description, so if you pasted a generic job description in step 1, the criteria will be generic too.

How to write them so the scoring is actually useful is in best practices for evaluation criteria.

Related articles