CV Screening Agent
Ranks against the requirements, and shows its evidence for every score
2 of 5 candidates meet every must-have.
Top: Dana Rowe, 88 out of 100.
SQL is the scarcest must-have: 4 of 5 CVs show it.
CVs that do not state years of experience: 1; none was assumed.
Score = 60% must-haves + 25% nice-to-haves + 15% experience. It reads the words on the CV only.
Must-have coverage (CVs that show it)
- SQL4/5
- Python4/5
- stakeholder4/5
Outcome
- Shortlist40%
- Review40%
- Not matched20%
Candidate by candidate
| Candidate | Score | Must-haves | Nice-to-haves | Years | Outcome |
|---|---|---|---|---|---|
| Dana Rowe | 88 | 3/3 | 1 | 6 | Shortlist |
| Marcus Lee | 88 | 3/3 | 1 | 5 | Shortlist |
| Priya Nair | 64 | 2/3 | 1 | 3 | Review |
| Sofia Bianchi | 60 | 3/3 | 0 | not stated | Review |
| Owen Clarke | 28 | 1/3 | 0 | 2 | Not matched |
Evidence from the CVs
| Candidate | Skill | Where it appears |
|---|---|---|
| Dana Rowe | SQL | Built SQL pipelines and dashboards in Tableau. |
| Dana Rowe | Python | Wrote Python scripts for forecasting. |
| Marcus Lee | SQL | Models in dbt and SQL on a cloud warehouse. |
| Marcus Lee | Python | Python for data quality checks. |
| Priya Nair | SQL | Daily SQL and Tableau reporting. |
| Priya Nair | stakeholder | Supported stakeholder requests for ad-hoc analysis. |
| Sofia Bianchi | SQL | Python and SQL daily for modelling; led projects with executive stakeholder alignment. |
| Sofia Bianchi | Python | Python and SQL daily for modelling; led projects with executive stakeholder alignment. |
What to do this week
- Interview Dana Rowe, Marcus Lee first.
- Ask Priya Nair about Python: it is not in the CV text.
Do
- Use the same criteria for every candidate.
Don't
- Don't reject on this score alone: it reads words, not ability.
How to improve
- Write must-haves as skills that appear on a CV.
- Read the Review group by hand.
Demo
Overview
Most AI screening tools output a ranked list nobody can explain. That is a problem twice over: hiring managers cannot tell a good filter from a broken one, and an employer cannot answer a rejected candidate who asks why.
This agent scores each application against requirements you define, one criterion at a time, and cites the line of the application that supports each score. Nothing about the candidate's identity enters the assessment — not name, age, gender, nationality, photograph, marital status, or where they went to school as a proxy for any of those.
It does not reject anyone. It produces a ranking, the evidence behind it, and an explicit list of what it could not determine — and the hiring decision stays with a person. That record is what makes the shortlist usable when someone asks how it was produced.
Benefits
A defensible shortlist
You can show how it was produced, not just what it produced.
Consistent across a pile
The two-hundredth application is read against the same bar as the first.
Reviewable in an hour
Evidence attached, so a manager checks rather than re-reads.
What it does
Scores criterion by criterion
One score per stated requirement, never a single opaque number.
Evidence for every score
Each score cites the part of the application it rests on.
Identity plays no part
Name, age, gender, nationality, photo and marital status are excluded — and so are the common proxies for them.
Rejects nobody
It ranks and evidences. The decision stays with a person, by design.
States what it could not tell
An unclear requirement is reported as undetermined, not scored zero. Those are different, and only one is fair.
An auditable record
Every assessment reconstructable months later, when someone asks.
What this changes
What this changes about your week
- It takes over
- Reading every application against the stated criteria, consistently, at volume.
- Done today by
- A recruiter or the hiring manager, late in the evening.
- How often
- Every high-volume role.
- Typical manual effort
- Two hundred applications read properly is two days.
Why HR & People in particular. The two-hundredth application gets the same bar as the first, and every score carries the evidence behind it — which is what you need when a rejected candidate asks.
Runs unattended
Started by you or by an event, and it finishes on its own. Nothing waits for someone to be at a desk.
The same standard every time
The two-hundredth item is held to the bar the first one was. Consistency is the part people cannot sustain.
It cannot act on its own
CV Screening Agent has no path to sending, spending or committing. That limit is why its output is safe to act on.
How it runs
This agent runs server-side through the PROMIVO runtime. Each run is logged step by step and every tool call is permission-checked before it executes.
- 1InputValidated against the agent input schema
- 2Modelclaude-opus-5 with automatic fallback
- 3Toolsextract_structured_data
- 4PermissionsRuns READ actions automatically; anything higher waits for approval
- 5OutputStructured result matching the output schema
- 6LoggingTokens, duration and every step recorded
Requirements
Integrations
No integrations required.
Before you start
- The role's requirements, stated as criteria
- Applications as text
- Requires the All-Access plan or higher
Use cases
High-volume roles
Hundreds of applications assessed against one consistent bar.
Structured first pass
A ranked, evidenced longlist for a human to shortlist from.
Auditing your own process
See which requirements are actually filtering, and which exclude for no reason.
Pricing
All-Access
Billed monthly through your PROMIVO subscription. Cancel at any time.
Runs consume your plan allowance for agent executions and tokens. See plan limits
FAQ
Does it reject candidates automatically?
No, and it cannot. It ranks and evidences; a person decides. Automated rejection is what draws regulatory attention, and it is not a capability this agent has.
How do you prevent bias?
It scores only against requirements you state, and identity attributes are excluded along with their common proxies — school as a stand-in for background, career gaps, name origin. That is a meaningful reduction, not a guarantee: your requirements themselves can be biased, which is why every score is evidenced and reviewable.
Can I use it for employment decisions?
It supports the decision; it does not make it. Automated employment decision tools are regulated in several jurisdictions and may require notice, bias auditing or human review. Check what applies to you — the agent produces the record such rules generally expect.
Reviews
Reviews open once customers have run this agent.
Need this agent to work differently?
Tell us what to change and our team will scope a customised version for your business.
Customize this agent