Care needs more than a quiz to test AI judgement

The simplest AI training plan is also the most tempting: choose one course, enrol everyone and use completion rates as proof that the workforce is ready.

Equal access does not mean identical learning

There should be a common foundation of data and AI literacy across a care organisation. Everyone needs to understand what AI is, the organisation’s rules for using it, the importance of privacy and data quality, the limits of AI-generated outputs and how to raise a concern.

But people do not carry equal responsibility for AI. A care worker using an approved summarisation feature, a home manager reviewing a scheduling recommendation, a procurement lead assessing a supplier and an executive approving investment face different decisions and consequences.

Giving them identical learning may be administratively neat, but it ignores the context that both HIQA and the EU AI Act say matters.

Start with a training-needs analysis

HIQA’s July 2026 guidance is clear and direct: services have a responsibility to undertake a training-needs analysis and provide education, training and career support so staff have the knowledge to use AI safely and responsibly in their role.

A useful analysis does more than asking how technical someone is. It looks at the decisions they influence, the tools they use, the information they handle, the people affected by their work and what they are expected to notice, explain or escalate.

That creates a defensible link between the learning offered and the real context of use.

Different roles need different depth

Executives and accountable leaders

Leadership needs enough understanding to challenge the business case, define risk appetite, assign accountability and judge whether the organisation has the capability to scale. The key question is whether value, safety, rights and operational readiness have been considered together. Without this understanding and buy-in throughout leaders in the business, the roll out will fail.

Procurement, governance and technical teams

These groups need deeper capability around intended purpose, evidence, data provenance, supplier assurance, privacy, cybersecurity, bias, interoperability, monitoring and exit planning. Their decisions shape the conditions under which everyone else will use the tool.

Care managers and operational leaders

Managers need to understand how AI changes workflows and responsibilities. They may need to review outputs, support staff, spot inconsistent practice, respond to incidents and explain how the tool affects the people receiving care.

Frontline staff

Frontline learning must be practical. Which tools are approved? What information can be entered? What does the output mean? What are the warning signs? When should it be ignored, overridden or escalated? How should its use be explained to the person receiving care?

Train for judgement, not recall

Knowledge checks can confirm that someone remembers a policy. They are less useful for testing whether that person can act under pressure or ambiguity.

Applied scenarios are stronger. Show a plausible AI output with an obvious flaw and ask what the person would do. Present a public AI tool and a piece of identifiable information and ask what can be entered. Give a manager a case where a recommendation conflicts with professional judgement and ask who owns the decision and what should be recorded.

Those exercises reveal whether the person can transfer knowledge into practice. They also expose gaps in the surrounding workflow that training alone can’t fix.

Ongoing means ongoing

HIQA says training should keep pace with new developments, roles, tools and context. That rules out treating AI literacy as a once-a-year exercise.

Capability should be reviewed when a new tool is introduced, a role changes, an incident occurs, monitoring identifies a pattern, regulation develops or the organisation learns something new about performance. Short refreshers tied to real changes will be more useful than repeatedly sending everyone through the same broad course.

The goal is a workforce that remains competent as the environment and AI landscape changes. 

Education as an architecture

The scalable answer is a learning architecture: a shared foundation, role-based pathways, use-case-specific practice and ongoing evidence of understanding. That gives the organisation consistency without pretending every person needs the same depth.

Data & AI Literacy Academy customises their education around your people, decisions and AI estate. In care, the quality of AI adoption will depend as much on human judgement as it does on the technology itself.

This article is informative content, not legal advice.

• HIQA, National Guidance for the Responsible and Safe Use of Artificial Intelligence in Health and Social Care Services, July 2026, especially pp. 17, 25–26, 44–47 and 63–68.
• Department of Health and HSE, AI for Care: The Artificial Intelligence Strategy for Healthcare in Ireland 2026–2030, especially pp. 16 and 38–40.
• Regulation (EU) 2024/1689, Article 4 (AI literacy), consolidated text dated 27 July 2026.

‍