An AI recruiter that shows its work.
Meet Noa AI-Agent, the NoaMetrics Recruiting Agent. Give it a one-line brief and get back a ranked shortlist. Every score quotes the resume or interview answer behind it. A second agent pokes holes in the claims that matter. You make every call that counts.
Built for heads of talent acquisition at companies that hire remotely. We run it on one of your own open roles, not a canned demo.

Most AI screening hands you a score and says “trust me.”
Works great, until one of these happens.
Your hiring manager asks “why this person?” and you've got nothing.
A bare number is hard to defend when someone senior pushes back.
You're now babysitting a black box.
If you can't see how it got there, you recheck everything by hand. That isn't automation. That's a second job.
Legal shows up with a bias claim or an AI Act question.
Hiring is a regulated use of AI in the EU. Legal and the works council will want to know how decisions get made and who signs off.
Someone who isn't who they say they are slips through.
Remote hiring means deepfaked candidates, coached answers, and resumes written to trick the tools that read them.
We built Noa AI-Agent around exactly these four moments.
Five rules we don't bend
You can check every one of them in the product.
Evidence, not scores
Every score links to the words behind it.
Quotes from the resume and the interview transcript sit right next to each score. No evidence for a requirement? The report says so. Noa AI-Agent doesn't guess.
Verified, not just claimed
A skeptic fact-checks before you see anything.
A separate Skeptic Agent goes back to the source material and challenges every claim that decides whether someone qualifies. The report keeps four buckets apart: claimed, corroborated, verified and demonstrated in the interview.
Fraud-resistant by design
Integrity is part of the interview, not a plugin.
NoaMetrics integrity signals run on the session and the answers. Resumes are treated as untrusted input. The signals are indicators with reasons attached, and a human reviews them. Noa AI-Agent never decides someone lied.
Humans decide
Noa AI-Agent recommends. You decide.
Approval gates are mandatory, and we list every one on this page. Noa AI-Agent never rejects anyone on its own.
Built for scrutiny
Ready for your legal and works-council review.
Designed for GDPR, the EU AI Act and the German AGG. Personal data is masked before scoring, every action lands in an audit log, hosting is in the EU, and there's no emotion recognition.
One line in. Shortlist out. Humans at the checkpoints.
Ten specialised agents, one orchestrator. It keeps things moving and stops at every gate until a named person approves.
Write one line. Noa AI-Agent writes the vacancy.
Your brief becomes a full vacancy, checked for inclusive language and pay transparency.
You decide: approve the vacancy
Noa AI-Agent builds the interview.
Questions plus behaviourally anchored rubrics, so every answer gets scored against the same yardstick.
You decide: approve the scenario
The role goes live.
Noa AI-Agent posts to job portals through their official APIs or feeds.
Resumes get read, masked and scored.
Identity data is masked first. Then every requirement is scored with a quoted evidence trail.
Top candidates get picked and scheduled.
Noa AI-Agent selects the strongest and books the interviews.
You decide: keep or swap any pick
An AI interviewer runs a structured interview.
It says upfront that it's an AI. Anyone who'd rather talk to a human can ask for one.
Answers get scored and integrity-checked.
Multiple independent judge passes score each answer. NoaMetrics checks the session and answers for integrity indicators.
You decide: review any indicator before it touches a candidate
The Skeptic Agent pushes back.
It re-reads the primary evidence and challenges every claim that decides qualification.
Candidates get ranked in confidence tiers.
Each one comes with a Why? report showing the reasoning.
You make the call.
Nothing is final until a human has looked at the shortlist.
You decide: approve the shortlist and every rejection
The Why? report: what you hand the hiring manager
The sample up top shows one candidate. Here's how to read it, and how an auditor would.
- Tier
- How strong and how well-confirmed the evidence is. Not just the score.
- Requirement
- One line per requirement from the vacancy, each with its own evidence level.
- Quote and source
- The exact words, and where they came from: the resume or a numbered interview answer.
- Skeptic outcome
- Did the claim survive when the Skeptic Agent went back to the source?
- Integrity
- Any indicators, why each was raised, and whether a human has reviewed them.
- Decision
- Who approved, who overrode, and when.
The four evidence levels
- Claimed
- Candidate says so.
- Corroborated
- More than one place in the file backs it up.
- Verified
- Checked against an outside source.
- Demonstrated
- Shown in the candidate's own interview answer.
Request the full sample report
Same report, every section filled in. Leave your details and we'll get in touch to walk you through it.
Five places a human decides
Not a limitation. This is the product.
| Gate | What you decide | What Noa AI-Agent does |
|---|---|---|
| The vacancy | HR expertApprove, edit or reject the text before anything goes live. | Drafts it and checks inclusive language and pay transparency. |
| The interview scenario | HR expertApprove or change the questions and the rubrics. | Writes the scenario with behaviourally anchored rubrics. |
| Who gets interviewed | HR expertKeep Noa AI-Agent's picks or swap any of them. | Selects the top candidates and schedules the interviews. |
| Integrity indicators | HR expertConfirm or dismiss each one before it touches a candidate's outcome. | Raises the indicator and says why. |
| The shortlist and every rejection | HR expertApprove, reorder or remove. Reverse any rejection recommendation. | Ranks in confidence tiers and writes a Why? report for each candidate. |
Noa AI-Agent never sends a rejection on its own. Every override is logged: who, when and why.
What your security and legal folks will ask for
Better they see it early. Here's the list.
For security
- A resume is data, not instructions. Candidate files are untrusted input, so text hidden in a resume can't steer the agents.
- Integrity indicators on every session. NoaMetrics checks the session and answers, and a human reviews anything it raises.
- An audit log. Every agent action, approval and override is recorded.
- EU hosting.
For legal, the DPO and the works council
- Masked personal data. Identity details are masked before scoring, so scores rest on the requirements.
- Candidates are told. The interviewer says it's an AI, and a human alternative is there on request.
- No emotion recognition.
- Language and pay checks. Every vacancy gets checked for inclusive wording and pay transparency before a human approves it.
On wording. We say Noa AI-Agent is designed for GDPR and the EU AI Act, and we're picky about that. We'll say compliant or certified once counsel and an audit for your setup say we can.
Request the controls overview
A short doc on GDPR, the AI Act, fairness and security, written for the people who sign off. Leave your details and we'll get in touch.
Jump to what matters for your job
- Head of talent acquisitionTime saved, shortlist quality, and a Why? report you can show the business.The Why? report
- Recruiting operationsThe workflow, the gates, and how it plugs into job portals.How it works
- HR director or CHROFairness, the audit trail, and the legal exposure that comes with AI in hiring.Human gates
- DPO, legal or works councilWhat's masked, what's logged, and what we will and won't claim.Security and legal
- Security leadHow fake candidates and manipulated resumes get handled.Security and legal
Questions people ask before booking
Does Noa AI-Agent reject candidates on its own?
No. Noa AI-Agent recommends and explains. A human approves the shortlist and every rejection, and nothing gets sent without that approval.
Is Noa AI-Agent compliant with GDPR and the EU AI Act?
It's designed for both, and we use that wording on purpose. The AI Act treats hiring tools as high-risk, so human oversight, an audit log, candidate disclosure and masked personal data were in from day one. We'll say compliant or certified once counsel and an audit for your setup allow it.
Can Noa AI-Agent tell if a candidate is lying?
No, and it doesn't try. NoaMetrics integrity signals flag things worth a second look, like a session or answer that seems off. They're indicators with a reason attached. A human decides what they mean.
Do candidates know they're talking to an AI?
Yes. The interviewer says it's an AI right at the start, and any candidate can ask for a human instead.
What happens to candidate data?
Identity data is masked before any scoring, so scores reflect the requirements, not who someone is. Hosting is in the EU, and every action, approval and override goes in an audit log.
What if I disagree with a ranking?
Override it. Swap any candidate Noa AI-Agent picks for interview, and change or reverse the shortlist and any rejection recommendation. Every override is logged with who, when and why.
How does Noa AI-Agent post jobs?
Through the official APIs or feeds of the job portals. No scraping, no working around platform terms.
How is Noa AI-Agent different from the AI in our ATS or other AI interviewers?
Most suite add-ons and AI interviewers hand you a score. Noa AI-Agent shows the evidence behind it, has a second agent challenge the claims that matter, and writes a Why? report that a candidate, hiring manager or auditor can read.
What does a pilot look like?
We start with a walkthrough on one of your own open roles, so you see real output for a job you know. We sort out the details after you book.
Not ready to book? Request the checklist.
The Remote Interview Fraud Red Flags Checklist: what to watch for when you can't meet a candidate in person.
Try it on a role you're hiring for right now
Book a pilot and we'll run Noa AI-Agent on one of your own open roles, with someone from your team at every gate.
- Real brief in, real Why? report out
- Your HR experts approve every step
- Bring your legal and security questions
Book a pilot
We start with a walkthrough on one of your own open roles, so you see real output for a job you know.