AI Inventory for HR Deployers — What to Document Under EU AI Act
AI Inventory for HR Deployers
An AI inventory — a structured record of every AI system your organisation deploys or uses — is not an explicit standalone obligation in Regulation (EU) 2024/1689 ("EU AI Act"), but it is a practical prerequisite for meeting the obligations that are.
Before you can determine whether a system is high-risk, apply Article 26 deployer obligations, retain the right logs or train staff under Article 4, you need to know what AI systems are in use across your HR function.
Why an AI Inventory Matters for HR
HR functions often deploy AI at multiple points in the employee lifecycle without a unified view of the whole picture:
- Applicant tracking systems with AI-assisted screening.
- Video interview platforms with automated evaluation features.
- Performance management tools with predictive analytics.
- Workforce planning tools that model headcount and skills gaps.
- Learning and development platforms that personalise content.
- Employee engagement survey tools that use NLP to analyse open text.
Any of these may fall under Annex III, point 4 — particularly where they affect decisions about individual employees or candidates. An inventory is the starting point for understanding your regulatory exposure.
What the Inventory Should Contain
A minimum AI inventory entry per system includes:
| Field | What to record |
|---|---|
| System name and vendor | Identify the tool and whether it is external SaaS, internal build or a model embedded in a broader platform |
| Deployment purpose | What HR decision or process does it support? |
| Annex III classification | Does it fall under point 4? If not, which category, if any? |
| Provider documentation received | Has the vendor provided Article 13 technical documentation and Article 14 oversight specifications? |
| Data processed | Categories of personal data; whether profiling of individuals is performed |
| Human oversight mechanism | Who oversees outputs; what is the override process |
| Risk status | High-risk / limited risk / minimal risk, with brief rationale |
| Date of classification review | When was this last verified? |
Linking the Inventory to Compliance Action
Once you have a complete inventory, each high-risk system in the HR category triggers a predictable set of actions:
- Obtain or verify provider documentation — Article 13 requires the vendor to provide a user manual with the system's performance characteristics, known limitations, and oversight requirements. If you do not have this, request it explicitly.
- Implement human oversight — per Article 14 and your inventory's oversight column.
- Train relevant staff — Article 4 AI literacy.
- Retain logs — per Article 26(6).
- Inform workers — per Article 26(7) before deployment or changes.
AI Literacy Obligation (Article 4)
The inventory process itself is an AI literacy exercise. Organisations that go through the process of identifying and classifying AI systems across their HR function will have materially better AI literacy than those that do not. Article 4 (applicable since 2 February 2025) requires documented evidence that staff who work with AI have sufficient knowledge — not just formal certifications, but a verifiable record of structured learning.
Timeline
- 2 February 2025 — Article 4 (AI literacy) applicable now.
- 2 August 2026 — Full EU AI Act applicability for high-risk AI deployers.
- 2 December 2027 — Digital Omnibus (adopted EP 16 June 2026, Council 29 June 2026) defers Annex III obligations. Awaiting Official Journal publication; plan for 2 August 2026 until in force.
Getting Started
PowerQuant Module 1 includes a structured AI inventory template alongside Article 4 literacy documentation and Article 26 compliance records — delivered as a source-cited evidence pack within five working days.