Human oversight.
Name competent people with the authority to oversee each high-risk system.
AI register and EU AI Act
You can only govern the AI you have found.
MetroMinds analyses your application portfolio to find where AI is used, including AI built into vendor software, and records each system in an AI register aligned with the EU AI Act.
Every system in the register records what the EU AI Act asks you to know:
- Its purpose and the business capability it supports
- Your role as provider or deployer
- Risk category and Annex III area
- Human oversight and an owner by role, never by name
GovernanceAI registerDemo environmentAI register
AI systems in use
Northstar Bank · aligned with the EU AI Act
Add AI system
All systems8
High risk3
Limited risk1
Minimal risk3
Under assessment1
Risk categoryAll categoriesAnnex III areaAll areasRoleProvider and deployerSearch AI systems
| AI system | Application / vendor | Purpose | Role | Risk category | Annex III area | Owner | Status | Last review |
|---|---|---|---|---|---|---|---|---|
| Credit decline reason codes | LoanDesk Origination Fenwick Systems | Explain credit declines to applicants | Deployer | High risk | Creditworthiness | Head of Credit Risk | In review | 12 Sep 2026 |
| Candidate ranking in recruitment platform | TalentFlow Recruit Orvia Software | Rank applicants for open vacancies | Deployer | High risk | Employment | HR Director | Approved | 3 Sep 2026 |
| Mortgage affordability pre-check | HomeLend Portal In-house | Pre-assess affordability before a mortgage offer | Provider | High risk | Creditworthiness | Head of Mortgages | Draft | 28 Aug 2026 |
| Shift allocation assistant | WorkPlan Scheduler Orvia Software | Allocate contact centre shifts and tasks | Deployer | Under assessment | Employment | Head of Customer Operations | In review | 19 Sep 2026 |
The deadlines moved. The work did not.
- Prohibited practices and AI literacy apply.
- Obligations for general-purpose AI models apply.
- 22 October 2026
- Obligations for high-risk systems in Annex III apply, after the Digital Omnibus.
- Obligations for high-risk AI in regulated products (Annex I) apply.
The extra time is best used for the inventory. Classification depends on what each system is used for, and that starts with knowing where it is.
Find the AI you already use.
Much AI arrives without a project: a vendor release adds a ranking feature, a business tool switches on an assistant. MetroMinds analyses your application portfolio and flags where AI is present, so nothing stays outside the register.
- Applications and vendor modules with AI features
- The business capability and process each one supports
- A first indication of the Annex III area, for review by your experts
From application list to register entry
- Analyse. Your application portfolio is checked for AI features, including those that arrived with a vendor release.
- Review. Your experts confirm the purpose and the first Annex III indication.
- Register. Each finding becomes a register entry with an owner, ready for oversight and monitoring.
The Annex III indication is a starting point for your experts, not a legal classification. Whether a system is high-risk depends on its intended purpose and how it is used.
Check one of your use casesGovernancePortfolio analysisDemo environmentPortfolio analysis
Where AI sits in your applications
48 applications analysed · 9 with AI found · 21 AI features
- TalentFlow RecruitOrvia Software
- LoanDesk OriginationFenwick Systems
- ServiceHubCalderon Labs
- Kestrel Ledger ERPKestrel Business Software
- WorkPlan SchedulerOrvia Software
Application
TalentFlow Recruit
Vendor: Orvia Software · release 8.2
- Business capability
- Human Capital and Culture › Talent Acquisition › Recruitment and Selection
- Process
- Hire to retire · screening and selection
- Application owner
- HR Director
AI features found (3)
| Feature | AI domain | First Annex III indication |
|---|---|---|
| Candidate ranking Ranks applicants against the vacancy profile | Prediction | Employment |
| CV parsing Extracts work history and skills from CVs | Perception | Employment |
| Interview scheduling assistant Proposes interview slots to candidates | Conversational AI | — |
Annex III indication for review by your expertsAdd to AI register
High-risk AI does not look different. Its purpose does.
A first screen of the MetroMinds Library flags 333 of 4,980 use cases, about one in fifteen, that touch an area listed in Annex III. Most employment-related ones sit outside HR. Technically they look like the rest; what decides the classification is the intended purpose.
- 122
- of the 333 sit in operations and delivery.
- 56
- sit in Human Capital and Culture, the capability HR owns.
- 21
- sit in Technology, Data and AI.
So the people who own high-risk AI are often operations managers who may never think of it as an HR or IT system.
From register entry to daily practice.
Monitoring and logs.
Monitor operation and keep automatically generated logs for at least six months.
Informing people.
Inform workers before first use, and inform people affected by decisions the system supports.
Impact assessment.
Run a fundamental rights impact assessment where the Act requires it, for example for credit scoring and public services.
GovernanceAI registerEntryDemo environmentAI register entry
Candidate ranking in recruitment platform
TalentFlow Recruit · Deployer · Employment
High risk Approved
OverviewClassificationOversightLogs and monitoringImpact assessmentHistory
Human oversight
| Role | Oversight task | Authority |
|---|---|---|
| Recruitment Lead | Reviews every ranked shortlist before candidates are rejected | Can override or stop |
| HR Director | Accountable owner; approves changes to ranking criteria | Can suspend use |
| HR Analytics Specialist | Checks monthly for drift and adverse impact | Escalates to owner |
Training on the systemCompleted for 3 of 3 roles
Workers informed before first useYes
Log retention6 months minimum
MetroMinds supports your compliance work. It does not replace legal advice.
CIO
Why EU AI Act Compliance Starts With Finding Your AI
You cannot classify or govern AI you have not found. One in fifteen Library use cases touches a high-risk area, and most of them sit outside HR and IT.
Risk and compliance
What the EU AI Act Asks of You Once Your AI Register Is Ready
An AI register shows what you have. The EU AI Act deployer obligations are about how you run it: oversight, logs, monitoring, impact assessments and informing people.