Evidence screeningPremium
AI candidate screening that shows its evidence
Evidence screening reads every application against a rubric your team approves, quotes the resume line behind each verdict, and suggests a next step. People make the decisions. The AI never rejects anyone, and every decision lands in a ledger you can audit.
Evidence screening is a Premium feature. Premium is the set of paid AI features that sit on top of the Core hiring workspace, such as timed interviews and AI screening.
Backend engineer, 2026 graduates · Rubric v2
Candidate 07Blind
4 of 5 must-haves shown
1 of 2 nice-to-haves shown
AI suggests
Ask
MustBuilt web services in production
ShownDesigned and shipped 14 endpoints for the orders service, serving 2 million requests a day.
MustDesigned relational database schemas
ShownNormalised the orders schema and cut report queries from 9 seconds to 400 ms.
MustWrites automated tests
ShownRaised unit test coverage on the billing module from 40% to 85%.
Follow-up question
Which cloud platform have you deployed a service to, and how?
Your decision
Why screening broke
More applications, fewer recruiters, and scores nobody can explain
Applications per recruiter rose about five times between 2022 and 2025, while recruiting teams shrank.
Screening is now the slowest step in most hiring loops. The usual answer is a 0 to 100 fit score, and that is the part that fails. A recruiter cannot tell a hiring manager why one applicant scored 72 and another 68. A rejected candidate learns nothing. An auditor asking how a decision was made gets a number, not a reason.
Hiring rules are moving the same way. Recruitment AI is treated as high-risk under the EU AI Act, and several US state and city rules now ask for bias audits, candidate notices, records, and explanations after adverse decisions. Teams need screening that shows its reasons and keeps a person in charge.
How it works
A rubric you approve, evidence you can read, decisions you own
- 01
Approve the rubric first
The AI drafts must-have and nice-to-have criteria from the job. Your team edits and publishes them. A bias check flags risky criteria such as age, gender, marital status, religion, caste, nationality, photos, pincodes, college tier, or culture fit.
- 02
Evidence quotes per criterion
Each application gets a verdict per criterion: shown, partial, not shown, or unclear, with a short quote from the resume. Quotes are checked against the resume text, so a quote that cannot be found becomes unclear. You read "4 of 5 must-haves shown", never a percentage.
- 03
AI suggests, never rejects
The AI suggests advance, ask with follow-up questions, or manual review for unreadable resumes. Only a person can record a not-advance decision, and it needs a reason, as does any decision that differs from the suggestion.
- 04
Blind by default
Names, emails, phone numbers, links, ages, gender words, and pincodes are removed before a resume is evaluated. The board shows a label such as Candidate 07 until a person records a decision.
- 05
Decision ledger and audit export
Every decision is saved with the AI suggestion, whether the reviewer agreed, who decided, and why. Decisions are append-only, and the ledger exports as a CSV file for audits.
- 06
Fairness monitor
For each job, compare AI and human advance rates by source, department, or batch year. Cohorts below four-fifths of the highest rate are flagged, small cohorts show counts without a flag, and the agreement rate shows how often people follow the AI.
For candidates and students
Second look: feedback a candidate can act on
Most screening tools stop at the rejection email. When a recruiter chooses to share feedback, evidence screening turns that moment into a next step for the applicant.
See what was missing
Candidates and students see which criteria their application showed and which it did not, in plain language, without any score.
Practice for each gap
Each gap comes with practice links from ClavePrep, so the applicant knows what to work on before the next application.
Submit proof once
The applicant can send one written proof note per application. The AI re-checks only the gap criteria against it, and the recruiter sees that a Second look was requested. Nothing changes until a person reviews it.
Second look re-checks are free for the candidate and for the organization.
For universities
Cohort gap maps for placement teams
Placement officers screen roster students against the rubric for each campus drive, grouped by department and batch year. The cohort gap map shows which criteria students most often fail to show, across one job or every drive in the organization.
That map becomes a training plan for the placement cell. Students who receive feedback get gap notes and practice links before the next drive, so the same gap does not cost them twice.
For agencies and companies
One rubric for every reviewer and every client
Recruitment agencies
Screen applied and sourced candidates against a rubric the client signs off. Share evidence quotes instead of gut feel, and keep a ledger that shows how each shortlist was built.
Startups and companies
Give every hiring manager the same first-pass review. Follow-up questions arrive with each application that needs them, and the fairness monitor shows whether one source or department is being left behind.
Pricing
Premium, 1 Clave per evaluated application
Evidence screening is part of Premium, not Core. Claves are AI usage credits purchased from billing and distributed to team members. Each evaluated application uses 1 Clave from the balance of the team member who runs the screening.
Drafting and editing the rubric is free. Applications without a readable resume go to manual review without a charge, and Second look re-checks cost nothing. If a balance runs out mid-batch, the batch stops and reports how many applications were evaluated; finished evaluations are kept.
Evidence screening questions
Does the AI reject candidates?
No. The AI only suggests one of three next steps: advance, ask follow-up questions, or send to manual review. Only a person on your team can record a not-advance decision, and that decision needs a written reason. Applications without a readable resume go to manual review instead of the bottom of the list.
Is this compliant with the EU AI Act?
No tool makes a hiring process compliant on its own, and the employer using it stays responsible for its duties. Recruitment AI is treated as high-risk under the EU AI Act. Evidence screening is built to support two deployer duties: human oversight, because a person makes every decision, and record keeping, because each decision is logged with the AI suggestion and can be exported. Take legal advice for your own situation.
Why evidence quotes instead of a match score?
A single score cannot tell a reviewer, a hiring manager, or a candidate why someone was or was not moved forward. A verdict per criterion with a quote from the resume can. Every quote is checked against the resume text, and a quote that cannot be found there is shown as unclear rather than trusted.
What does blind by default hide?
Before a resume is evaluated, names, emails, phone numbers, links, dates of birth, ages, gender and marital words, and pincodes are removed. On the board each applicant appears as a stable label such as Candidate 07, and the name and email show only after a person records a decision.
How much does evidence screening cost?
Evidence screening is part of Premium. Each evaluated application uses 1 Clave from the acting team member's balance. Drafting and editing the rubric is free, applications without a readable resume go to manual review without a charge, and Second look re-checks are free for the candidate and the organization. See Pricing plans for Claves packs.
What can a candidate or student do with Second look?
When a recruiter shares feedback, the candidate sees which criteria were shown and which were missing, with practice links for each gap. They can submit one written proof note per application. The AI re-checks only the gap criteria against that note, and a recruiter reviews the result. Second look never changes the application status by itself.
How does the fairness monitor work?
For each job, the monitor compares AI and human advance rates by source, department, or batch year. A cohort whose rate falls below four-fifths of the highest cohort is flagged for review. Cohorts with fewer than five screened applications show counts without a flag, and the monitor also shows how often reviewers agree with the AI suggestion.
Can universities use evidence screening for campus drives?
Yes. Placement teams screen roster students against the rubric for each drive, and the cohort gap map shows which criteria students most often fail to show across a job or the whole organization. Students who receive feedback get gap notes and practice links before the next drive.
Start with a clear hiring workflow
Build your hiring workspace
Claves are the AI usage credits purchased from billing and distributed to team members for timed interviews, AI screening, and related AI hiring work. Premium sourcing is the entitlement-gated, AI-assisted candidate discovery that imports candidates into the organization's pipeline. Start with the AI Recruiter workspace for jobs, candidates, applications, and team access, then add premium AI when your hiring work needs it.