TL;DR
The largest population in scope of Article 4 is staff who use AI tools without configuring or building them. They need a curriculum that is short, role-anchored, and evidence-led: how the tool they use can fail, how to spot it, what to do next, and what records to keep. Five to seven hours of role-specific training per non-technical role tier, refreshed annually, is the defensible baseline. The point is not to make sales reps into ML engineers. It is to make sure they can recognise an AI failure mode in their own workflow and escalate it without breaking the audit trail.
Why non-technical roles are in scope
The literal text of Article 4 of Regulation (EU) 2024/1689 requires providers and deployers to ensure “a sufficient level of AI literacy of their staff and other persons dealing with the operation and use of AI systems on their behalf”. The European Commission’s AI Act service desk makes the scope explicit: any staff who configure, supervise, or rely on AI outputs are in scope, whether or not they have a technical role.
In practice, that means a sales rep who relies on an AI-generated proposal draft, a finance analyst who uses an AI categoriser for transactions, a recruiter who uses an AI-driven CV screen, and a customer-support agent who uses an AI summary all fall in scope. The Government of Portugal’s digital portal explicitly flags non-technical roles as the dominant exposure area for SMEs.
What “literacy” means for a non-technical role
A non-technical curriculum has four parts:
- Capability and limit. What the tool the team uses can and cannot do. Not the abstract “what is GenAI” deck. The actual tool, named, with its known failure modes.
- Failure-spotting drills. Worked examples of bad AI outputs in the team’s own domain, with the cues that flagged them.
- Escalation paths. When to stop relying on the AI output, who to tell, where to log it.
- Personal-data and audit hygiene. What can and cannot be pasted into a tool, and what record to keep when it is.
Nova IMS’s executive AI programme and applied work from Instituto Superior Técnico both run cohorts in this exact shape for Portuguese corporate clients.
Curriculum length and cadence
The defensible baseline for non-technical operators:
- 5 to 7 hours of initial training, split across two or three sessions over four weeks, role-specific cohorts (sales, marketing, ops, support, finance, legal each get their own).
- A 30-minute refresh every time a new AI tool enters service for that team.
- An annual recap of one to two hours, with new failure-mode examples drawn from the year’s incident log.
This adds up to roughly seven to nine hours per role per year. That is enough to defend a literacy claim when a regulator asks. It is also short enough that the business will actually run it.
What it is not
It is not a generic 30-minute “Intro to AI” video. It is not a one-off lunch-and-learn. It is not a pop quiz on neural network architectures. The European Commission’s Digital Strategy page is explicit that proportionate, role-specific training is what Article 4 asks for. Generic awareness sessions fail the proportionality test on their face.
How AISO Learn helps
Our team-engagement programme runs role-tier cohorts for non-technical teams: sales, marketing, ops, support, finance, legal, and HR each get their own curriculum, drawn from the tools they actually use day to day.
Run Article 4 training for non-technical teams