ENBook a demo
← All products / SYS.04
SYS.04 · Recruitment ·Active

Cadence

AI-assisted CV screening for hiring teams. Every application is scored against a role-specific rubric the moment it arrives, then routed - clear passes and clear rejects handled automatically, the borderline band put in front of a recruiter with the model's reasoning attached.

Book a demo
A recruiter reviewing a candidate's criterion-by-criterion score before making the call

A recruiter reviewing a candidate's criterion-by-criterion score before making the call

From rubric to decision

04 steps
Step 01

A role gets a rubric before it gets a link

The recruiter defines the scoring criteria for the role, each with a weight and the sub-points it covers, plus the score above which a candidate is auto-shortlisted and below which they are auto-rejected. Cadence generates a public application link from it.

Rubric builder - weighted criteria and the advance / reject thresholds

Rubric builder - weighted criteria and the advance / reject thresholds

Step 02

The candidate applies against criteria they can see

The public page states plainly that a model scores the application first and a person reviews every borderline case. The candidate fills in their details, pastes their CV, and can open each criterion they'll be judged on.

Public application form with the evaluation criteria listed alongside

Public application form with the evaluation criteria listed alongside

Step 03

Every application is scored the moment it lands

A background job scores each criterion from 0 to 10 with a written justification, weights them into a single 0-100 result, and adds a short summary. Scoring is deterministic and the same rubric is applied to everyone. The score then routes the application: shortlisted, rejected, or held for human review.

Candidate drawer - model summary, then each criterion scored and explained

Candidate drawer - model summary, then each criterion scored and explained

Step 04

The recruiter decides - and every override is recorded

Shortlisted candidates get an interview-scheduling invite and HR gets an alert; clear rejects get a polite email after a short hold. Everything in the middle waits in the queue. When a recruiter overrides the model, they leave a one-line reason and it is logged against the model's accuracy.

Candidates table - stage, score and the recruiter's advance / reject controls

Candidates table - stage, score and the recruiter's advance / reject controls

What ships in Cadence

06 capabilities
F.01Weighted rubric per role

Named criteria with weights that sum to one, each carrying the sub-points it covers, plus a threshold for auto-advance and one for auto-reject.

F.02Criterion-level scoring

Every criterion is scored 0-10 with a written reason and a 2-3 sentence overall summary - not a single opaque number.

F.03Threshold-based routing

Above the advance line a candidate is shortlisted, below the reject line they are declined, and the band between is put in front of a person.

F.04Automated candidate & HR email

Shortlist alert to HR, interview-scheduling invite to the candidate, rejection after a configurable hold. Nothing is sent for the borderline band.

F.05Override log

Each human change is captured with a mandatory reason, an override rate, and a view of exactly where reviewers disagreed with the model.

F.06Configurable model

Gemini, OpenAI or Groq sit behind one scoring interface; the provider is a setting, not a rewrite.

Interface reference

Screens 01–04
Recruiter home - highest-scoring applications still waiting on a human
01Recruiter home - highest-scoring applications still waiting on a human
Candidates table with stage filters, score bands and bulk actions
02Candidates table with stage filters, score bands and bulk actions
Score distribution across every scored application
03Score distribution across every scored application
Workspace settings - AI provider, automation toggles and the rejection hold
04Workspace settings - AI provider, automation toggles and the rejection hold

Built for hiring teams

Cadence replaces the shared inbox, the spreadsheet of names and the first-pass CV read that never quite happens consistently. The model reads every application against the same published rubric; a person makes every decision.

Typical use: 50–1,000 applications per role

Use case: High-volume roleHundreds of applicants against one rubric. The obvious rejects and obvious advances are handled; the recruiter spends their time on the middle band.
Use case: Small hiring teamOne or two people covering every open req, who need a consistent first pass done before they open the inbox.
Use case: Auditable hiringEvery score, its reasoning and every human override kept on record for a role that may be reviewed later.
See SYS.04 on one of your own roles

Send us a job description and a batch of anonymised CVs. We'll build the rubric, score them in Cadence and show you the shortlist - and the reasoning behind each score -.

Book a demo