Rollout sequencing

All at once, or phased by risk?

Your L&D lead has the training approved and now has to decide who goes first. Everyone at once feels fair. Starting with the recruiter running a ranking tool and the claims handler closest to automated decisions is where the exposure actually sits.

Everyone at once

One announcement, one kickoff. Every team hears about the training the same week, from the same message.
Momentum. A single launch date creates a shared sense of urgency that a staggered start can lose.
No one feels skipped. Nobody asks why their department went last.
Reporting is simple: one start date, one completion deadline, one number to report up.

Phased by risk

Start with the roles closest to AI decisions: the recruiter running a ranking tool, the claims handler, the case worker screening applications. That is where the exposure sits.
The AI register shows which systems touch which roles, so the order is not a guess: it follows where the AI is actually used.
Depth follows risk, not job title, so the first wave already gets sector-specific scenarios instead of a generic overview.
Every assignment is recorded with the answers given and the rule version applied, so the sequence itself can be replayed and explained, not just remembered.

See the public log of legal changes

When one wave is the right call

A team small enough to fit around one table does not need a phased order: everyone is exposed to roughly the same decisions, so one wave covers the real risk anyway. And when a customer, regulator, or a signed agreement sets one hard date for everyone, phasing by risk just adds coordination overhead without adding protection. In both cases, column A is not a compromise: it is the correct read of the situation.

What the risk order actually buys you

The recruiter running a ranking tool and the case worker screening applications carry more exposure than most of the organisation combined, and the AI register is what shows that, connecting each system to the roles that use it. /start turns that into an order: sector, systems in use, and roles feed a role-based route, and every assignment is recorded with the answers given and the rule version applied, so the sequence can be replayed and explained later, not just remembered. The first wave gets scenarios built for its own sector, because depth follows risk, not job title, so the case worker is not sitting through a generic module while waiting for the real one.

Article 4 asks for measures matched to context

Article 4 has applied since 2 February 2025. Since 27 July 2026, under Regulation (EU) 2026/1744, it asks providers and deployers to take measures that support the development of AI literacy, taking context into account. It does not require guaranteeing a specific individual level. That holds whether training rolls out all at once or in phases. A risk-ordered rollout is that context-matching made visible: the register shows which roles sit closest to AI-assisted decisions, and starting there is a measure that fits the context, not a compliance shortcut.

See what this looks like for your organisation

The 5-minute scan maps your AI use to the roles around it and shows where your evidence is thin.

AI Act evidence routes

Pick the route that matches your AI Act evidence question.

Focused pages per question: from mandatory Article 4 literacy and training records to sector evidence and an audit-ready dossier.

HR AI Act compliance

Training evidence for recruitment, people analytics, workforce decisions, and HR governance.

View route

AI literacy for HR

AI literacy for HR teams with role-based learning paths, recruitment and people analytics scenarios, and evidence reporting.

View route

AI literacy program

AI literacy program for organizations: training, course routes, masterclasses, certificates, team reporting, and Article 4 evidence.

View route

AI literacy evidence dossier

The evidence chain from role and use case to scenario practice, scores, certificates, and reporting.

View route

Everyone at once

One announcement, one kickoff. Every team hears about the training the same week, from the same message.

Momentum. A single launch date creates a shared sense of urgency that a staggered start can lose.

No one feels skipped. Nobody asks why their department went last.

Reporting is simple: one start date, one completion deadline, one number to report up.

Phased by risk

Start with the roles closest to AI decisions: the recruiter running a ranking tool, the claims handler, the case worker screening applications. That is where the exposure sits.

The AI register shows which systems touch which roles, so the order is not a guess: it follows where the AI is actually used.

Depth follows risk, not job title, so the first wave already gets sector-specific scenarios instead of a generic overview.

Every assignment is recorded with the answers given and the rule version applied, so the sequence itself can be replayed and explained, not just remembered.

See the public log of legal changes

When one wave is the right call

What the risk order actually buys you

Article 4 asks for measures matched to context