Use case

TalentScreen AI.
From audit to verified documents.

A startup discovers DILAIG. It starts the free audit, receives a score, generates its documents — automatically verified article by article. Three weeks later, it runs a re-audit to measure the concrete impact of its improvements.

Step 01

TalentScreen AI.
Who are they?

A fictional startup that developed an AI-based CV screening tool for recruitment teams. TalentScreen AI automates the first filter, ranking candidates according to job criteria.

High-risk sector. Personal data. Automated decisions affecting employment. TalentScreen is a textbook case for the AI Act. Here are three of the 50 questions asked during the audit:

04

Does your system fall within one of the high-risk domains listed in Annex III of the AI Act?

Annex III lists 8 high-risk domains: biometrics, infrastructure, education, employment, essential services, law enforcement, migration, justice. A system operating in any of these domains is automatically classified as high-risk.

Yes — employment, recruitment or HR management
Yes — education or vocational training
Yes — access to essential private/public services
No — none of these domains apply
06

What is your role regarding this AI system?

The one who designs the AI has far more obligations than the one who simply uses it. If you modify a purchased system, you may become responsible as if you had developed it yourself.

We designed and developed the system — it is our product
We use a purchased or licensed system, without modification
We resell or import an AI system without modifying it
26

Is the final decision made automatically by the system, or validated by a human?

The AI Act requires effective human oversight of high-risk systems. A fully automated decision without right of appeal constitutes a major non-compliance.

A human always validates before any action
The system proposes, a human confirms
The system decides alone, without systematic human validation
Variable depending on the type of decision
+ 47 questions · approximately 20 min total

Audit result

38/100High Risk

Triggered obligations

Art. 9Risk management system — mandatory
Art. 10Data governance and quality
Art. 11Technical documentation (Annex IV)
Art. 13Transparency towards users
Art. 14Effective human oversight
Art. 27Fundamental Rights Impact Assessment (FRIA)
Art. 47EU Declaration of Conformity
Art. 72Registration in the EU AI Act database

Recommendation #1

Human oversight absent — blocking

The system makes screening decisions without systematic human validation. Art. 14 of the AI Act requires that a human operator can intervene, correct or cancel any decision. This gap constitutes a critical non-compliance.

Step 02

Score 38/100.
High Risk.

Seconds after submission, the deterministic scoring engine calculates the risk level and identifies regulatory obligations article by article.

For TalentScreen, two factors weigh on the score: the absence of human oversight (Art. 14) and undocumented training data (Art. 10). These points will be at the heart of the FRIA.

Reproducible score, identical if nothing has changed
Calculated deterministically, without generative AI
Defensible before a regulator or auditor
Step 03

Documents generated
and verified in 60 seconds.

From your audit answers, DILAIG drafts your documents in legal English, structured according to the exact requirements of each article. Each document is automatically verified: blocking fields detected, compliance status calculated article by article, placeholders resolved before delivery.

Generated document

FRIA — Fundamental Rights Impact Assessment

Art. 27

§ 2.1 — Identification of affected fundamental rights

The TalentScreen system is likely to affect the following rights protected by the Charter of Fundamental Rights of the European Union:

  • ·Art. 21 — Non-discrimination (risk of algorithmic bias on protected criteria)
  • ·Art. 15 — Freedom to choose an occupation and right to work
  • ·Art. 8 — Protection of personal data

§ 3.2 — Mitigation measures

Given the absence of systematic human oversight identified during the audit (answer Q-26), it is imperative to implement a human validation mechanism before any candidate exclusion decision. This mechanism must be documented and traceable in accordance with Art. 14 §4.

✓ Automatically verified: human oversight (Art. 14), personal data (Art. 10), fundamental rights (Art. 27) — status calculated article by article — to be validated by your lawyer before official submission.

Generated document

Transparency Notice

Art. 13 & 50

§ 1 — AI system identity

This document describes the operation of the artificial intelligence system TalentScreen AI, developed and operated by TalentScreen AI SAS (hereinafter 'the provider'), in accordance with Article 13 of Regulation (EU) 2024/1689 on artificial intelligence.

§ 3 — Capabilities and limitations

The system assigns a relevance score (0–100) to each submitted CV, based on a comparison with the job criteria. This score is a decision-support tool. The final decision on interview or rejection belongs exclusively to the human operator. The system cannot predict a candidate's future performance.

✓ Automatically verified: Art. 13 and Art. 50 obligations covered, no blocking field detected — to be validated by your lawyer before official submission.

5 document types available

FRIA, Declaration of Conformity, Transparency Notice, Technical Documentation, Data Governance Plan.

€0.99

/ document · Free plan

Included from Starter

Step 04

Three weeks later.
Score 74/100.

The TalentScreen team followed the recommendations. The re-audit measures the concrete impact of these changes, in minutes, with the same questionnaire.

What TalentScreen changed

Automatic decision without validation
Mandatory human review before exclusion
No decision log
Complete timestamped history — 3 years
Undocumented training data
Data governance register created
No appeals procedure
Challenge form integrated into the product

Initial audit

38/100High Risk
Human oversight: absent
Decision traceability: 0%
Data governance: undocumented
Appeals procedure: non-existent
3 weeks of compliance work

Re-audit — after changes

74/100Moderate Risk
Human oversight: operational
Decision traceability: 100%
Data governance: documented
Appeals procedure: accessible to candidates
+36 points in 3 weeks = High Risk → Moderate Risk
Re-audit cost (Free plan):€0.99
Summary

TalentScreen at the end of the journey

74/100

Compliance score

vs 38 at the start

5

Documents produced & verified

FRIA, Declaration, Notice, Technical, Governance

8

Articles covered

documented obligations

< 2h

Total time invested

audit + re-audit + docs

Important note

A score of 74/100 does not mean full compliance. It means the most critical risks are covered and the file is defensible before an auditor. The documents generated and verified by DILAIG are professional drafts. They must be validated by a lawyer or DPO before any official submission. That is precisely why they exist: you give your lawyer a structured and verified starting point, not a blank page.

Your first step
is free.

Like TalentScreen, start with the audit. 20 minutes. No credit card. You will know exactly where you stand before investing anything.

No credit card · Instant result · €0.99 / document if needed