The Irish care provider’s guide to AI readiness

If an AI compliance plan ends with a policy and a course-completion report, it is not finished. Ireland’s emerging framework points providers towards turning governance into day-to-day behaviour.

The short version: three instruments, three different jobs

The EU AI Act, Ireland's AI for Care Strategy and HIQA's July 2026 national guidance are closely related, but they don’t have the same legal status. A defensible plan starts by separating them clearly.

The EU AI Act: the binding obligation

Article 4 applies to providers and deployers of AI systems. It requires organisations to take measures that support the development of AI literacy among staff and other people operating or using AI on their behalf. Those measures must take account of technical knowledge, experience, education and training, the context in which AI is used, and the people or groups affected. The Act doesn’t prescribe a specific level of literacy for each individual, but the duty on organisations to support its development remains. The organisational obligation remains, but no specific level is mandated. National market surveillance authorities began supervising and enforcing Article 4 on 2 August 2026.

AI for Care: the national direction

Ireland's 2026-2030 strategy is sets the national direction for AI in healthcare. It calls robust change management a prerequisite for effective and sustainable adoption, says workforce readiness should include structured training, hands-on workshops and mentorship, and immediately prioritises AI literacy across the workforce, patients and service users.

HIQA guidance: care-sector good practice

HIQA's final National Guidance for the Responsible and Safe Use of Artificial Intelligence in Health and Social Care Services supports good practice across public, private and voluntary services. It is non-statutory guidance rather than standalone AI legislation. It shows how responsible AI can be translated into proportionate care-sector practice across governance, human rights, safety, workforce capability and the full AI lifecycle.

Put simply: the Act sets the duty, the strategy sets the direction, HIQA shows what it looks like in practice. 

The policy and process checklist

HIQA doesn't prescribe one universal policy pack, either. The response should be proportionate to the provider, the tool, the intended use and the care context. Some organisations will update existing policies; others will need new ones. Either way, the following questions need clear answers:

1. Which AI tools are authorised?

Maintain an inventory of approved AI tools, their owners, intended purposes, approved users and risk status. Make the list easy for staff to find. State which tools or uses are prohibited, what information may be entered, and how new tools are requested or approved.

2. Who is accountable?

Name the executive or governance function responsible for AI. Define responsibility at service, team and individual level. Management owns procurement, governance, monitoring and audit. Staff remain responsible for interpreting outputs and acting within their professional obligations. Clinical staff remain responsible for clinical decisions.

3. How does human oversight work in practice?

Specify who reviews an output, what information they need, when independent verification is required, and how they can question, override, pause or escalate. A statement that a human remains involved is not enough if the workflow gives that person no time, authority or practical route to intervene.

4. How are rights, safety and data protected?

Connect AI use to the policies that already cover this ground: privacy and consent, information governance, cybersecurity, data quality, accessibility, equality, risk management and professional practice. Pay particular attention to people who may be vulnerable or under-represented in the data.

5. How is the full lifecycle controlled?

Build checks in at every stage:

• Planning and procurement: supplier evidence, testing in the intended setting, risk assessment
• Deployment and operation: incident and complaint routes, performance monitoring, business continuity
• Retirement and decommissioning: safe withdrawal, data handling at exit

6. How will people be supported through the change?

Define how staff will be involved, informed, trained and supported before and after implementation. Update job guidance and workflows, brief managers, create visible feedback routes and explain what changes for each role. Both HIQA and AI for Care explicitly call for robust change management rather than technology being introduced around the workforce.

The people requirements, role by role

Article 4 doesn't demand the same learning for everybody. HIQA adds care-specific detail by calling for a training-needs analysis and ongoing education adapted to role, technical knowledge and context. The real question isn't who's completed a course. It's what each person needs to be able to do differently.

Boards and accountable executives

Set the organisation's position on AI, approve risk appetite, assign accountability, resource the change and ask for evidence of safe adoption. Leaders need enough understanding to challenge value, risk, rights, workforce readiness and the organisation's ability to sustain oversight.

Procurement, AI, data and governance teams

These teams create the conditions in which everyone else will use the tool, so their depth needs to go furthest: intended purpose, evidence and supplier claims, data provenance, privacy and cybersecurity, bias, interoperability, regulatory classification, monitoring and exit plans.

Operational managers and change leads

Translate policy into the local workflow. Managers need to explain what is changing, support staff through uncertainty, make escalation routes real, identify inconsistent practice and bring feedback from the frontline into governance and improvement.

Frontline and care staff

Know which tools are approved, what information can be used, and who remains responsible for the decision. Frontline staff need to be able to:

• Assess outputs critically and recognise misleading information or bias
• Maintain data quality
• Override or escalate when something looks wrong
• Explain relevant AI use to people receiving care

Learning, HR and workforce teams

Undertake the training-needs analysis, define role groups, assign suitable pathways, support onboarding and role changes, plan refreshers, and retain evidence. Their success measure should include applied understanding and changed practice rather than a simple tick on a training quiz.

What role-relevant learning should cover

HIQA's examples provide a useful practical curriculum. Depending on role and use case, staff may need to:

• Understand how the tool functions and the evidence behind it
• Evaluate outputs critically, and recognise bias or poor-quality data
• Comply with privacy requirements and local policy
• Override an output or escalate a concern
• Explain benefits, risks and limitations to people using services
• Support safe withdrawal when a tool is retired

A shared foundation can establish common language and organisational rules. It should then branch into deeper pathways. A frontline worker needs practical decisions and boundaries. A manager needs workflow, supervision and change-leadership capability. A procurement or governance specialist needs much greater depth on evidence, risk and supplier assurance.

Use scenarios wherever possible. Ask a frontline colleague what they would do with a plausible but flawed output. Ask a manager how they would respond to repeated workarounds. Ask a procurement lead what evidence would change their recommendation. Those exercises test judgement and expose process gaps that a knowledge quiz will miss.

Why completion is not the end point

A completion dashboard proves that learning was assigned and accessed. It doesn't prove that an approved tool is being used correctly, that staff feel able to challenge an output, or that managers know how to respond when practice diverges from policy.

Treat learning as one part of a change programme. Pair it with:

• An accessible authorised-tool list and workflow guidance at the point of use
• Manager briefings and scenario practice
• Local champions
• Communications that explain why the change matters
• Feedback routes and clear triggers for reinforcement

Use operational signals to decide what happens next. Repeated overrides, recurring questions, incidents, complaints, low adoption or workarounds may point to a learning need, a poorly designed workflow, an unsuitable tool, or all three. Feed that evidence back into policy, education and implementation.

The test is whether people can explain what changed in their role and act safely when reality doesn't match the happy path.

A practical 90-day starting plan

Days 1-30: understand

Appoint an accountable sponsor. Build or validate the AI inventory. Map current and planned use cases, affected groups and responsible roles. Conduct a training-needs analysis. Review policies, workflows and existing learning. Establish a baseline for confidence, behaviour and adoption, not only awareness.

Days 31-60: equip

Update the priority policies and approval routes. Define a common foundation and role-based learning paths. Launch the relevant EU AI Act education. Give managers practical briefing materials and frontline teams scenario-based guidance. Involve staff in implementation and open clear routes for questions, feedback and concerns.

Days 61-90: embed

Reinforce the change through workflow prompts, manager conversations and applied assessments. Review questions, overrides, incidents, complaints and adoption patterns. Correct unclear policies or workflow friction. Agree refresher triggers and report both learning evidence and operational evidence to the accountable governance group.

What evidence should the organisation retain?

Article 4 doesn't prescribe one certificate or mandate a specific individual literacy level. That makes the reasoning behind the organisation's measures important. Keep a practical evidence pack across four areas:

Governance evidence
Named accountability, the authorised-tool inventory, risk assessments, impact assessments, approval records, lifecycle controls.

People evidence
The training-needs analysis, role and audience map, assigned pathways, participation and assessment data, scenario outcomes, accessibility considerations, onboarding and refresher records.

Change evidence
Staff involvement, communications, manager briefings, updated workflow guidance, questions raised, feedback received and actions taken in response.

Operational evidence
Monitoring results, audit findings, overrides, incidents, complaints, workarounds, adoption patterns, and evidence that policy or learning was updated when conditions changed.

How the Data & AI Literacy Academy helps

The Data & AI Literacy Academy helps organisations move from broad requirements to a practical workforce and change management plan. Our EU AI Act OnDemand learning path provides a scalable foundation and can be tailored by job role, including practical pathways for frontline workers. 

We can combine it with deeper learning for leaders, managers, procurement, governance and specialist teams, supported by workshops and scenarios where applied judgement matters.

The outcome is a workforce that knows what changed, what good practice looks like and what to do when an AI tool is wrong.

‍

This article is informative content, not legal advice. Sources below:

• 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.
• European Commission, AI talent, skills and literacy, updated 27 July 2026. Article 4 enforcement by national market surveillance authorities began on 2 August 2026.

‍

‍