Screening & intelligence
AI Screening: criteria, match % and the ranked pool
How candidates are scored against your criteria, what match percentage means, and how to slice, filter and act on the ranked pool.
6 min read
What AI Screening is
Every search gets an AI Screening view: the entire sourced pool, scored candidate by candidate against your ranking criteria, sorted by match. It replaces the first two rounds of manual resume review, and unlike a resume screen, it shows its reasoning on every score.
Criteria: what candidates are graded on
- Criteria come from your outcomes ("can build a partner channel from zero"), your query, and role context. You can add, remove and reorder them.
- Each criterion is graded HIGH, MEDIUM or LOW per candidate by an AI evaluator that must cite the evidence behind every call.
- Priority order matters: your first criterion weighs more than your fourth. Drag to reorder and the pool re-ranks.
- Criteria are evidence-gradable by design: we grade what a career record can actually show, and say No Data when it cannot.
Match percentage, honestly computed
Match % is the weighted fit across your criteria, adjusted by two honesty rules that most tools skip:
Abstention
A criterion with no evidence scores No Data and is excluded rather than counted as a failure. Coverage discounts the final number, so a profile graded on one lucky criterion cannot outrank a fully-evidenced one.
Constraint fit
A candidate who contradicts one of your hard filters (wrong city, outside the experience band) is demoted below comparable full fits, visibly, instead of being silently dropped or silently kept.
Reachability
Move likelihood nudges the ranking so people you can actually hire rank above equally-qualified people who will never take the call.
The full methodology behind every signal on the profile lives on the talent intelligence page.
Working the pool
- Talia’s read: one-tap slices for Active job seekers, Ready to join soon, Low flight risk, High impact and In budget.
- Column filters: Excel-style multi-select on Job Title and Company, with per-value counts, composable with the slices.
- Select rows (or select all matching the current filters) and bulk shortlist or bulk add to an outreach sequence.
- Build shortlist and plan: hand your selection (or the whole pool) to Talia and she picks channels and drafts outreach per candidate.
- Evaluate more profiles: extend the same search deeper into the market when you want a bigger pool.
Two candidates with identical criteria grades can differ a few points in match %. That is reachability at work: the one more likely to move ranks higher, because a shortlist is a plan, not a leaderboard.
Frequently asked
Why does a candidate show No Data on a criterion?
Because their profile carries no evidence either way. We deliberately abstain instead of guessing, and the match percentage accounts for coverage. A No Data grade is a prompt to ask in the first call, not a hidden penalty.
Can I change criteria after a search has run?
Yes. Add, remove or reorder criteria and the pool re-scores against the new bar. Already-built shortlists stay stable until you rebuild them, so your plan does not reshuffle under you.
How many candidates get scored?
The whole working pool, typically 200 to 300 candidates per search, not just the first page. Scoring streams in live while sourcing runs.