Science1 publisher2 min readPublished
Trinity-led survey of Irish workplaces finds AI adding verification work to existing jobs
The Right-Sizing AI at Work study reports almost half of respondents using AI regularly and describes review and judgment duties spreading into ordinary jobs. The published account carries no sample size.
The Scientist · Science desk
What happened
- The Right-Sizing AI at Work study reports that AI's most significant changes in Irish workplaces are happening at the level of tasks and workflows, not through the elimination of jobs.
- Almost half of respondents reported regular use of AI tools across a wide range of workplace activities, according to the study.
- The study also found confidence in AI skills running ahead of participation in formal training, and argues for more structured, role-specific skills development.
Compiled by The ScientistSomething wrong?How this is made
Why it matters
- constraint An employer treating this as evidence that AI does not cut jobs is leaning on an instrument that questions current employees, so a role closed quietly after a resignation sits outside what the survey can see.
- cost Reviewing and verifying model output is work, and it falls on people already employed, so the cost of adoption appears as workload inside existing roles instead of in a hiring line.
- decision The lever the study hands a manager is training design rather than tool selection, and the body that paid for it builds training programmes with industry.
- contradiction Regular use alongside a weak sense of dependency undercuts any reading in which AI has already become load-bearing in these jobs, and the design never tests what happens when the tool is taken away.
A workforce survey reaches the people who are still working. The phys.org account of Right-Sizing AI at Work reports no sample size, no sampling frame and no fieldwork dates [16], so the finding that almost half of respondents use AI tools regularly is a share with no count behind it [7][18]. It matters more for the claim about jobs, because a post left unfilled after someone resigned has nobody to answer a survey question about it.
Yasseri keeps his qualifiers. "Our research finds that AI's impact on work in Ireland is currently best understood as a process of task and workflow transformation rather than immediate large-scale job displacement," he said [10]. The words doing the work there are "currently" and "immediate". The phys.org write-up drops both, headlines the study as AI reshaping the workplace and not replacing jobs, and opens by saying AI is changing work in Ireland for the better [15]. None of the findings it enumerates is a measured productivity or job-quality outcome [17].
The part that does not depend on the denominator is the description of what the work now consists of. Workers are increasingly required to review, verify, coordinate and exercise judgment over AI-supported outputs [6]. Checking a model's output is a task with a duration and an error rate, done by someone already on the payroll, and no headcount series records it. The study's own framing is that competitive advantage comes from how organisations integrate AI into jobs, workflows and workforce development, and not from the technology alone [14].
Respondents reported widespread use without a strong sense of dependency; many said AI supported their work without being essential to completing it [8]. The test for substitutability is removal, and a questionnaire removes nothing. The same limit applies to the skills result: the study found confidence in AI ability sitting alongside weaker participation in formal training [9].
The commissioner and the recommendation point the same way. Technology Ireland DIGITAL Skillnet commissioned the work [2], which the Centre for Sociology of Humans and Machines, a joint Trinity College Dublin and TU Dublin centre, carried out [3]. Its director, Maire Hunt, said: "We commissioned this research to give employers a robust evidence base for those decisions and to ensure the programs we develop with industry remain practical, relevant and aligned to the challenges businesses are facing today" [19]. Hunt also said that "using AI and generating real business value from AI are two very different things" [11]. Mark Jordan, chief executive of Skillnet Ireland, said the technology's "value will ultimately be determined by how we prepare people and organizations to harness it effectively" [12].
What to watch
- Publication of the full report with respondent count, sampling frame and fieldwork dates, which would make the usage share interpretable.
- Whether the training gap is backed by enrolment or completion records, and not just respondents' self-reported confidence.
- A second wave of the same survey, or Irish vacancy data, showing whether roles are being absorbed through attrition.