Everyone Suddenly Wants AI Training. Here's How to Buy It Without Wasting the Budget

AI training is the most consistently in-demand AI service in Ireland right now, and the 2 August 2026 start of AI literacy enforcement is about to pour fuel on it. That surge will produce a wave of compliance-theatre training — a webinar, a certificate, no behaviour change. Here is how to tell training that supports your Article 4 measures and actually changes how people work from training that merely generates an attendance record.

Regulatory update, 4 October 2026. The Commission’s current timetable separates the applicable 2026 duties from the relevant high-risk requirements in December 2027 and August 2028. The current Article 4 guidance does not mandate a particular course or certificate. The AI Office of Ireland is now established. Earlier forecasts and proposed-law descriptions below are historical commentary, not current deadlines. Assess the intended use, organisational role and transition provisions before acting.

Update, October 2026. Regulation (EU) 2026/1744 (the Digital Omnibus on AI) amended Article 4 with effect from 27 July 2026. Providers and deployers must now take measures to support the development of AI literacy of their staff and others dealing with AI systems on their behalf. The amended text does not require any specific level of AI literacy of any individual, and national market surveillance authorities supervise the duty from 2 August 2026. Descriptions of the earlier wording below reflect the law when this post was written.

Across every signal we track — search demand, ad engagement, inbound conversations — AI training is the most consistently sought-after AI service in the Irish market this year. And the demand is about to spike, because from 2 August 2026 the supervision and enforcement of the EU AI Act's AI literacy obligation begins, and "we should get the team trained" becomes "we are required to have done something about this."

That combination — genuine demand plus a compliance deadline — reliably produces a specific market failure: a flood of suppliers selling attendance certificates, and a flood of buyers who discover a year later that the training satisfied the paperwork and changed nothing. We wrote about the underlying pattern in Why Most AI Training Doesn't Change Anything; this piece is the buyer's guide version, updated for the deadline.

What Article 4 actually requires — and doesn't

The obligation, as rewritten by the Digital Omnibus, is to take measures to support the development of AI literacy among staff dealing with AI on your behalf, taking account of their role, the systems in use and the context. The amended text does not require any specific level of AI literacy of any individual. Three practical consequences:

• It applies to nearly everyone. The obligation is not tied to high-risk systems. If your people use AI at work — and they do, whether sanctioned or as shadow AI — you are in scope.
• What is appropriate is role-relative. The board needs to govern AI; the marketing team needs to use it well and safely; nobody needs a machine-learning syllabus. One generic course for everyone is mis-fit by design.
• Evidence matters more than hours. What you can show a regulator — who was trained on what, mapped to how they use AI, with a named owner and a refresh cycle — beats seat-time. This is one of the four questions in what still lands on 2 August.

The five tells of compliance-theatre training

1. Same content for every role. If the finance team and the warehouse team get the same deck, it is a broadcast, not training.
2. No connection to your actual tools and policies. Training that never mentions which AI your organisation uses, or what your policy allows, produces knowledge nobody can apply on Monday.
3. No before/after measure. If nothing is assessed, "it went well" means people enjoyed it.
4. A certificate as the deliverable. Certificates evidence attendance. Regulators — and more importantly, outcomes — care about capability.
5. No governance layer. Literacy for staff without AI governance training for the board leaves the accountable people the least equipped — exactly backwards, as we argued in what separates AI advisory from AI awareness.

What good looks like

The training that works starts from your AI inventory and your policies, splits by role, uses your live tools, and leaves behind three artefacts: a capability baseline, a documented training record fit for a regulator's question, and a set of working practices people actually adopted. Done properly, the literacy obligation stops being a cost of compliance and becomes the adoption programme you needed anyway — the organisations getting measurably better results from AI are the ones that trained for use, not for attendance.

Our AI training for Irish organisations is built on that structure, with role-specific tracks including board programmes and legal-team training, and tool-specific tracks for Copilot and Claude. For teams already working in Claude, AI at Work's Claude training covers skills, artifacts and Claude Code in more depth. If you are comparing options, the market overview on AI training providers in Ireland sets out what to ask any supplier — including us.

The deadline is real, the demand is rational, and the budget will be spent either way. The only question is whether it buys capability or a folder of certificates.