Seven in ten employees in Ireland believe their unique human capabilities cannot be replaced by AI, and only 7% expect AI to take on the majority of their current tasks. The finding comes from a survey of 1,000 workers and more than 300 employers in Ireland, part of wider research by payroll firm SD Worx, published alongside research showing workers trust their organisation to use AI fairly and ethically at higher rates than elsewhere in Europe.
Calm is easy to mistake for complacency. The same dataset contains a far sharper demand: three-quarters of Irish employees (75%) believe AI systems should be transparent about how they make decisions, against 64% across Europe. Confidence in keeping one's job and insistence on being able to interrogate the machine are not contradictory — together they describe a workforce that expects to be told, not replaced.
What the survey found
Adoption is concentrated among supervisors. Almost half of employees who supervise one or more people say they are increasingly using AI tools in their daily tasks, compared with 26% of employees who supervise nobody. Overall, 37% of employees in Ireland say their use of AI at work is increasing.
That gap is the most actionable finding in the research. People with responsibility for others' work already use these tools more heavily, which means they are the population most likely to be applying AI to assessments of those others — hiring screens, performance summaries, scheduling, anomaly detection in payroll.
The trust finding completes the picture: Irish employees are more likely than European peers to trust their organisation to use AI fairly and ethically. For employers, that is an asset rather than a comfort. Trust expressed as an expectation is conditional, and it is withdrawn quickly when decisions cannot be explained.
The regulatory timing
The survey coincides with new transparency obligations under Article 50 of the EU AI Act taking effect. Under those rules, people must be informed when interacting directly with certain AI systems, and certain AI-generated or manipulated content must be marked or disclosed.
Eimear Byrne, Managing Director, SD Worx Ireland, framed the obligation around explainability rather than disclosure alone: "AI is influencing high-impact decisions and this is particularly important for those working in HR and payroll to consider. Whether screening candidates, identifying payroll anomalies, or analysing workforce data, it is shaping real outcomes for real people. The new EU legislation means it is no longer enough for these systems to simply deliver results: organisations must be able to explain those results, justify them, and demonstrate that they are fair. This isn't just a legislative obligation, however. Our research shows that employees seek this clarity, too."
The convergence she describes is what makes this moment different. Transparency obligations arrive at the same time as employees asking for the same thing, which removes the usual argument that compliance is a cost with no internal beneficiary.
Why payroll is the sensitive frontier
Payroll and HR are where AI consequence is most immediate and least visible. Three features distinguish them.
Decisions are individual and legally consequential. A misclassified expense or a mispriced lead is recoverable; a wrongful deduction, incorrect overtime interpretation or biased screening outcome affects someone's income and statutory rights, and may be unlawful regardless of intent.
Errors compound silently. Payroll runs continuously, so a misinterpreted rule does not fail once — it fails every cycle for years, surfacing only when someone reconstructs history. This is precisely the pattern seen in large underpayment cases.
Correcting intent is difficult after the fact. Once a model has contributed to a hiring or pay decision, reconstructing how much weight it carried is hard, even with good logs. "The system recommended it" has never been a defence, but it is becoming a much-repeated explanation.
What transparency should mean in practice
Four operational commitments translate Article 50's principles into practice for anything touching pay or employment decisions:
- Disclose when AI is involved in a decision affecting an individual — including third-party tools embedded in HR and payroll platforms, where usage is frequently undisclosed and unmeasured.
- Record inputs and logic well enough to reconstruct a decision months later, rather than relying on the output.
- Keep a human able to explain the reasoning, not merely to approve it, since escalation to someone without visibility of the logic changes nothing.
- Test outcomes periodically for disparate impact, because fairness cannot be inferred from aggregate accuracy.
The competitive read
The research points to supervisors as the effective adoption channel. Training effort directed at people who manage others' work is more consequential than general workforce upskilling, because that is where AI is already being applied to consequential decisions and where explainability obligations will be tested first.
It also suggests employee trust is currently an advantage worth defending. Organisations that can explain their automated decisions will retain that goodwill; those that cannot will find the same 75% asking harder questions of auditors, unions and regulators.
Irish workers appear unconcerned about being replaced. That should not be mistaken for permission to deploy without explanation. The demand for transparency is the more durable signal, and it now carries statutory weight behind it.
Sources
- SD Worx research cited in this article: survey of 1,000 employees and more than 300 employers in Ireland, as part of a wider European study — for the finding that seven in ten employees believe their human capabilities cannot be replaced by AI (7% expecting AI to take most of their tasks), the supervision gap in AI usage (almost half of supervisors versus 26% of non-supervisors, 37% overall), the higher Irish trust in fair and ethical use, and the 75% versus 64% transparency comparison.
- EU AI Act, Article 50 transparency obligations, in force from the date noted — requirement to inform people interacting with certain AI systems and to mark or disclose certain AI-generated or manipulated content.
- Eimear Byrne, Managing Director, SD Worx Ireland — quoted remarks on AI's influence on high-impact decisions in HR and payroll and on explainability as both legal obligation and employee expectation.
- Note: analysis of payroll-specific risk, recommended transparency practices and the competitive implications is the author's.
