AI resume screening and candidate ranking

What is AI resume screening?

AI resume screening is the automated reading of every CV in an applicant pool into a comparable structured form, then scoring each candidate against the requirements of a specific role. In OpusHire, a CV is parsed once into skills, roles, tenure, education and achievements, and that structured profile — not the raw document — is what gets scored, stored and re-used.

The distinction matters. Keyword matching on raw text rewards candidates who happened to use your vocabulary and penalises the ones who described the same work differently. Parsing into a structured profile first means "led a team of six" and "managed six direct reports" resolve to the same fact, and it means the same parse is reused everywhere the candidate appears rather than being redone per role.

Because the profile is stored structurally, screening also stops being a one-shot operation. When a requisition's priorities change mid-search — the team decides depth in one skill matters more than breadth — the existing pool can be re-scored against the new weighting without re-reading a single CV.

How does OpusHire rank candidates against a role?

Each requirement in the requisition contributes to a score computed in code from defined rules, not produced by a language model. Recruiters see the per-dimension breakdown behind every ranking — which requirements a candidate met, which they missed, and how much each one counted — and can change the weighting to see the pool re-rank against a different definition of "best fit".

A ranking you cannot interrogate is a ranking you have to take on trust, which is no use when a hiring manager asks why their preferred candidate is eighth. Every OpusHire ranking decomposes into the requirements that produced it.

• Per-dimension scores. Skills, experience, education and role-relevance are scored separately, so a strong candidate with one gap is visibly that, rather than a middling overall number.
• Adjustable weights. Change what the role actually values and the shortlist recomputes immediately across the whole pool.
• Advisory minimums. A screening minimum raises a flag for a human to look at. It never auto-rejects, and no timer moves a candidate into a rejected state.

Does the AI decide who gets rejected?

No. Scores and recommendations are computed in code from defined rules, and every one of them is advisory. A language model may draft a narrative summary of a candidate or help word a question, but it never picks an outcome and never produces the number that decides one. Rejection is always a person clicking a button.

This is an architectural constraint rather than a configuration option: there is no setting that turns OpusHire into an auto-rejecting filter, and there is deliberately no timer that ages an untouched candidate into a terminal state. Every AI contribution to a score or a written summary is labelled as AI-assisted and human-reviewable wherever it is displayed, in the recruiter view and the candidate view alike.

For teams hiring in jurisdictions that restrict decisions made about individuals by automated means alone, that constraint is the point. It is also simply better screening practice — the value of a ranked pool is that it directs attention, not that it replaces judgement.

What happens to a CV after it is screened?

The structured profile persists with the candidate, so a person who applies to a second role is not re-parsed and does not re-enter their details. Recruiters get one record per person carrying application history, assessment results and interview records, instead of the same candidate appearing as three unrelated CVs in three requisitions.

Candidates keep control of that record. They can review and correct their profile, export it, and request deletion; access is scoped to the workspace that owns the requisition they applied to, and privileged views of candidate data are written to a tamper-evident audit trail.

Frequently asked questions