AI in training, under the AI Act’s gaze.
The EU AI Act classifies AI systems by risk and attaches duties accordingly — and employment-related systems draw particular attention. Here’s how AI in workplace learning tends to sit within that structure, and the questions worth putting to any vendor claiming AI capability.
Where learning AI sits, and where it starts to matter
The Act’s structure is risk-tiered, with employment and worker-management systems attracting heightened obligations. A useful way to think about learning AI is by what it determines: AI that drafts practice questions for human approval, or selects which concept a learner reviews next, sits far from consequential employment decisions. AI that gates promotion, allocates work, or produces scores used for termination sits much closer.
This isn’t legal advice — your counsel classifies your deployment, and the regulatory picture is still settling. But the architectural implication is stable and worth acting on now: keep AI in the drafting, scheduling and measuring layers, keep consequential decisions with humans holding evidence, and document which is which. That posture ages well under any interpretation.
Documented AI roles
Which tasks use AI, what gates them, and what a human approves — written down per capability rather than summarised as ‘AI-powered’, because governance questions ask per function.
Transparency to learners
Learners can know when AI generated a question or a hint, and how the coaching works. Explainability was already a pedagogical commitment here; regulation makes it a compliance asset.
Records that support your assessment
Drafting, review, approval and coaching interactions are logged — the artefacts a conformity or DPIA-style assessment will want, retained rather than reconstructed.
The AI, documented like it matters
Model purpose, human oversight, decision logs — the transparency posture the AI Act’s spirit asks of learning AI.
| Function | Control | Applies | State |
|---|---|---|---|
| Question drafts | Human-approved | always | Gated |
| Scheduling | Explainable rules | per item | Inspectable |
| Risk flags | Human-in-loop | per flag | Reviewed |
| Logs | Decision trail | retained | Exportable |
Interface shown as an illustration with representative numbers, not a screenshot — the layout is the product’s.
Bring the governance questions early.
AI roles, gates, logging and data flows — walked through for your DPIA rather than deferred to legal review.
The evidence this page stands on
Questions buyers ask
Is this legal advice on AI Act compliance?
No. Classification and obligations for your deployment are your counsel’s determination; this page describes the architecture and documentation that make that determination easier.
Does the platform make employment decisions?
No — it measures knowledge and produces evidence. What an organisation does with that evidence is its own decision, and we’d advise keeping consequential determinations human-held with the evidence as input.
What AI models are used, and where?
A curated set of commercial models routed per task, behind an abstraction that lets us change them. The functions they serve — drafting, hinting, clustering — are documented per capability.
Can we opt out of AI features entirely?
Question banks can be human-authored and coaching features configured down, if your governance posture requires it. You lose drafting speed, not the retention engine.
How do you handle AI-generated content errors?
Two gates before learners see anything — automated checks then human approval — plus production performance monitoring that flags weak items. The review gate is not optional in the workflow.
See it on your own content.
Bring one course. We’ll show you the retention curve your current training leaves behind — and what scheduled review does to it.
- 30 minutes, on your calendar — pick a slot here
- Run on your own content wherever possible, not a canned deck
- You see the dashboards, the learner surface and the evidence exports
- No commitment — and pilot data stays yours either way