Ibec found 36% of Irish organisations have not modelled AI's effect on their workforce, while entry-level hiring in AI-augmented roles fell 16.2%. The change is arriving through hiring decisions.
On 21 October Ibec published its Talent Trends survey of 319 HR professionals at its annual HR Leadership Summit. The headline was reassuring: most Irish HR leaders do not expect AI to reduce headcount in 2027. Ibec's Maeve McElwee said the job displacement narrative does not reflect what HR leaders are seeing.
One line further down the release, the same survey found that 36% of organisations have not yet formally modelled the impact of AI on their workforce structure. So more than a third of the people predicting no headcount effect have not done the analysis that would tell them.
A week earlier, IDA Ireland, Microsoft and LinkedIn released the Labour Market Pulse at the International AI Summit in Dublin. It reported that entry-level hiring in AI-augmented occupations fell 16.2% year on year between April and June. Hiring across those occupations overall fell 10.8%.
Put those two reports side by side and the picture is fairly clear. In Irish organisations, AI is showing up as the graduate role that never gets advertised. Very few firms are running redundancy programmes because of it.
Headcount is the wrong measure
HR leaders are right about one thing. Very few Irish firms are planning to let people go because of AI, and the ones that announce it tend to be using AI as cover for a cost decision they had already made.
But headcount is a stock. Workforce change happens in the flows: who you hire, who you replace when they leave, which roles get backfilled and which quietly get absorbed. A professional services firm that takes on twelve trainees this year instead of twenty has not reduced headcount. It has made a decision about its 2031 partner pipeline, and in most firms nobody at board level signed off on that decision. It happened one requisition at a time.
The IDA report is careful here, and it deserves credit for that. It says there is no clear evidence that the overall hiring slowdown is concentrated in the jobs most exposed to AI. But it also flags a sharper pullback in junior recruitment in software engineering and data analysis. That is exactly where you would expect to see the first effect, and exactly where a headcount survey will not see it.
We wrote earlier this year about what is happening to the bottom of the career ladder. The new data says it has not stopped.
What "not modelled" looks like in practice
When we ask leadership teams how they have assessed AI's workforce impact, the answers fall into three groups.
The vendor deck. Someone presented a productivity figure from a supplier's case study, say 30% time saved on drafting, and it was noted. Nobody translated it into roles, grades or hiring plans, because the figure was never measured inside the organisation.
The HR estimate. HR was asked for a view and gave one, usually based on conversations with line managers. Line managers are the people least likely to volunteer that their team could be smaller.
Nothing. The question has not been asked, because AI sits with IT or innovation and the workforce plan sits with HR, and the two documents have never been in the same meeting.
All three end in the same place. The board approves an AI budget on productivity grounds and approves a hiring plan on growth grounds, and the two plans contradict each other. If the AI investment works, the hiring plan is too big. If the hiring plan is right, the AI investment is not delivering. One of those two papers is wrong, and nobody has been asked to say which.
Four things to do before the 2027 budget
Model by task, not by role. Pick the five roles with the most people in them. For each one, list the tasks that take up the week and mark which ones current AI tools are actually being used for, based on usage data rather than licence counts. The useful output is a percentage of hours, by grade. Most firms can do it in two weeks.
Put the AI business case and the hiring plan in one paper. If a business case claims time savings, the hiring plan has to show where that time goes. Redeployed to client work, absorbed as capacity for growth, or reflected in fewer hires. Any of the three is a legitimate answer. A blank is the one answer the board should refuse.
Decide the graduate intake deliberately. If you are going to take fewer trainees because AI handles first-draft work, make that a board decision with a view on where your senior people come from in eight years. If you are going to keep the intake and change what trainees do, write down what the new training looks like. Firms that cut the intake by default will find out in 2032.
Watch the HR risk as well as the plan. AI used in recruitment, screening or performance management falls into the AI Act's high-risk category for employment, and the high-risk obligations now apply from December 2027. Employers will also need to inform workers' representatives before high-risk AI is put to use in the workplace. A workforce plan that relies on AI-driven screening to handle more applications with fewer recruiters has a compliance dependency built into it.
Where the board comes in
HR cannot solve this on its own. The Ibec figure tells you something about governance: in more than a third of organisations, a strategic workforce question with a five-to-ten-year tail is being answered by assumption.
Directors should ask management one plain question before signing off next year's budget: what does our AI investment assume about the number and mix of people we will need, and where is that written down? If the answer takes more than a page, or does not exist, that is the finding.
The Acuity AI position
We start workforce questions the same way we start every AI question: with evidence from inside the organisation, not a forecast from outside it. That means actual usage data, task-level time analysis, and a straight read of where AI is already changing the work, before anyone writes a headcount number or a hiring plan. The output is a single view of AI investment and workforce plan that a board can test and an HR director can defend. It is vendor-neutral, and in most firms it takes weeks. If your 2027 plans were built on the assumption that AI will not change your workforce, a diagnostic conversation is a sensible place to check it.