Making Sense of a Changing Economy - Speech by Governor Gabriel Makhlouf at Technological University Dublin

Thank you for the invitation to speak today.

It is a pleasure to be here at Technological University Dublin to speak about Ireland’s changing economy, and I look forward to your questions afterwards.

Whatever you go on to do, you will spend your working lives making decisions shaped by the economy, and many of you will help shape it. Alfred Marshall called economics the study of people in the ‘ordinary business of life’. Looking across an audience of young people starting out on their studies or careers, two things come to mind when I think of the ordinary business of your lives: what you can expect to earn and what things will cost.

I am, of course, talking about living standards: how wages evolve with prices. Central banks aim to keep inflation – the rate of change of prices – at a sustainable level. I will start with the sources of recent inflation, what we have done about it, and what published statistics miss but people feel. Over the longer run, real wage growth depends on productivity. And the biggest question about future productivity is artificial intelligence, which I turn to in the second half.

Inflation in the 2020s

I expect at least some of you were in secondary school when inflation returned. The reopening of economies after the pandemic, and then Russia’s invasion of Ukraine, pushed it to levels not seen for decades. Irish inflation peaked at almost 10% in 2022, led by energy and food (Chart 1). Core inflation, which strips out those volatile components, was lower but still high, at around 6%.

The ECB responded by raising interest rates at the fastest pace in its history (Chart 2). That was painful for many borrowers, and we did not take it lightly. The rise in interest rates, combined with easing of supply bottlenecks, contributed to a moderation of inflationary pressures with the annual rate falling to around 2 per cent by February 2026. Core inflation dynamics, at 2.4% at the time, were also increasingly in-line with the inflation target.

This year has brought a new, but sadly familiar, test. The conflict involving Iran has delivered another energy shock, and headline inflation in Ireland and the euro area rose to 4% in October. We have raised interest rates twice to prevent higher energy prices from becoming entrenched as broader inflation.

So far, price pressures appear largely confined to energy – euro area core inflation in October was 2.5% – wage growth is contained and inflation expectations remain anchored.  But indirect effects take time to work through supply chains, one reason inflation is still projected to rise in early 2027. And, as I set out in my recent blog, wage-setting adjusts with a lag, so second-round effects will only show gradually.

Despite expectations of more price increases in the near term, I supported keeping interest rates unchanged in October. Two things argue for patience. The 50 basis points – half a percentage point – of tightening since June are still working through the economy. And financing conditions have tightened independently, as long-term government bond yields have risen. Both should dampen demand. But upside risks to inflation remain. These mixed signals warrant waiting for further data and staff December projections, including our own for Ireland in the Quarterly Bulletin, to see how these factors play out.

Behind the numbers

Good policy depends on accurate data. But published inflation data do not always give the full picture. I want to highlight two things: what the average hides, and the gap between measured and perceived inflation.

Living outside the average

Households buy very different baskets of goods and services, so few, if any, experience the ‘average’ inflation rate that the Harmonised Index of Consumer Prices (HICP) is designed to capture.

Taking account of these differences, Central Bank research (PDF 237.71KB) showed that post-pandemic inflation was higher for older, rural and younger households, largely because they spend relatively more on energy. More recently, CSO data show relatively higher inflation in 2026 for younger, renting, rural and lower-income households.

So short-run real income shocks can differ a lot. But the time-frame also matters: over the longer run, our research shows smoother impacts (Chart 3), as fiscal policy, changes in behaviour and, in some cases, lower energy prices take effect.

It is not just about energy. Housing costs also weigh heavily on some groups, including younger workers, who are more likely to rent.¹ Rents are 7% of the Irish HICP basket (6% in the euro area), but that is an average across renters and owners. For renters, the CSO estimates rent at 25–30% of disposable income. That matters for their real income growth. Nominal graduate pay has largely kept pace with average pay in the economy (Chart 4). However, accounting for larger shares of spending on rent for younger workers and accounting for rental inflation since 2019 – 41% in Ireland, versus 17% in the euro area – the gap in real incomes for renters compared with the average household would be significantly wider.²

These differences matter for policy. For central banks, they inform how we communicate. Diversity in our research and engagement strategy reflects the diversity of economic experiences in Irish society. For fiscal policy, two recurring themes in my pre-budget letters are that cost-of-living supports should be targeted, tailored and temporary, to avoid adding to inflation; and that closing infrastructure deficits – in housing, transport and utilities – is one of Ireland’s most critical challenges.

What we measure and what we perceive

When economists say inflation has come down, we mean prices are rising more slowly. It does not mean prices have fallen.

Consumer prices in Ireland were around 26% higher this October than in October 2019, and around 28% higher across the euro area. For many items, they are not going back. Energy is the partial exception: its prices are set on world markets and can fall as well as rise (Chart 1).

Research shows that big shifts in the price level can linger long in the memory, far beyond the twelve-month window we use to measure annual inflation against our 2% target.³ So when some people hear that inflation is near target, as it was before this year, it can seem at odds with their own experience. Both can be true. The statistics measure the rate of change, but people live with the level.

The evidence bears this out. In August, with headline inflation at 3.4%, the ECB’s Consumer Expectations Survey found median perceived inflation in Ireland two percentage points higher than actual inflation, and average perceived inflation five points higher (Chart 5). Why? Differences in what households buy help explain why perceptions vary between people, but they tend to cancel out in the median or average. One answer, suggested in the literature, is that people judge prices against what they remember paying a few years ago, not just last year. Salience is another factor: people tend to anchor their inflation perceptions and expectations to items they buy often, such as food and fuel (Chart 6).⁴

This is not just a matter of how people feel. Perceptions influence how people spend, save and bargain over wages, and they can shape expectations of future inflation. Robert Shiller has argued – in my view, persuasively – that economic narratives can spread quickly and widely, driving real behaviour, sometimes as powerfully as the data.

So what should central banks do? First, meet our mandate: inflation sustainably at 2% over the medium term. There is no substitute for that. Second, earn trust, which I have previously described as foundational to effective monetary policy. People need to know that we understand their experience. That means explaining inflation simply and widely: its drivers, its likely path, and what we are doing about it. Good data are necessary, but they are not sufficient; good communication is also required.

Productivity, AI and your careers

If inflation is about what things cost, productivity is about what you can earn. It is the only lasting way for pay to rise faster than prices. On paper, Irish output per hour is among the highest in the world. But large multinationals skew the statistics, and much of the value they create is not paid out as Irish wages. In the domestic economy, productivity has grown more slowly in recent years.⁵

Your generation will be the first to start your careers alongside a technology that can draft, code, summarise and, increasingly, carry out complex tasks on its own. Two questions matter for your pay. Will AI raise productivity where you work? And will the gains reach the people doing the work? Let me look at what history suggests, and what we can already see in Irish data.

What history suggests

AI is now widely regarded as a general-purpose technology that will eventually change how almost every sector works. Its historical counterparts – steam, electricity and computing – offer two lessons.

The first is that they take time. Electricity did not raise factory productivity until the 1920s, four decades after the first central power station opened. Factories had to be redesigned around the new technology, not simply have it bolted onto the old way of working.⁶ The same lag explains why Robert Solow observed you could "see the computer age everywhere but in the productivity statistics".⁷

Some argue that the wait may be shorter this time. Bundesbank President Joachim Nagel points out that AI is being taken up faster than any previous general-purpose technology, including the internet, the smartphone and the personal computer; it builds on digital infrastructure already in place; and AI itself may speed up innovation. And so, while the AI diffusion will almost certainly be quicker than these other technologies, using a tool is not the same as reorganising a business around it, and that is where the lags have always been.

The second lesson is that the gains go to those who use a technology, not just those who invent it. When the computer age did reach the statistics, from the mid-1990s, productivity growth almost doubled in the United States but slowed in the EU. The gap opened not in making computers but in using them, in sectors such as wholesale and retail trade and financial services.⁸ Policymakers have made this point strongly in relation to productivity in Europe: we might not be frontier technology developers, but we should aim to be frontier adopters.⁹

How big might the gains be this time? The honest answer is that it is too early to tell, but many are trying to estimate what the impact might eventually be. ECB staff estimate that AI could add 0.2 to 0.4 percentage points a year to euro area productivity growth over the coming decade, depending largely on how quickly firms adopt it.¹⁰

So, measuring adoption will be a key task for economists and statisticians. How many of you used an AI tool this week? Most of you, I suspect. Across the euro area, half of workers now say they use AI in their jobs, double the share two years ago.¹¹ But firms are moving more slowly: only one in five Irish enterprises used AI last year, in line with the EU average and less than half the rate in Denmark.¹² And foreign-owned firms in Ireland, which tend to be larger, are more than two and a half times as likely to use AI as Irish-owned firms.¹³ Adoption is weakest in our domestic firms, exactly where productivity has lagged though use of AI by individuals in Ireland is comparatively high.  These survey-based metrics provide a useful high-level gauge of take up of AI technologies but tell us less about how the technology is being used, so we should be careful about drawing firm conclusions from these early trends.

There is also a monetary policy angle. In the short run, AI adds to demand before it adds to supply: our recent Quarterly Bulletin points to AI-driven investment and power use. In the long run, a genuine productivity boom should ease cost pressures, but it would also tend to raise the neutral interest rate, the rate that neither stimulates nor restrains the economy. Telling a spending boom from a productivity boom in real time is hard, which is one reason we need to invest in new sources of data.

What is happening to labour demand?

Official statistics are slow to catch up with fast change, so we have turned to a timelier source: job adverts on Indeed, the global hiring platform.

One in seven Irish job adverts now mentions AI (Chart 7), almost four times the share in 2019 and the highest of the six European countries in Indeed’s published tracker, ahead of the United Kingdom, Germany and France. It took off only from mid-2024 and is still rising, which suggests that new tools take time to find their way into jobs.

AI is also spreading across occupations. In 2019 in Ireland, it featured in at least one advert in ten in just three of 41 occupations. Today it is 21, including law, accounting, marketing, human resources, finance and engineering (Chart 8). Most of the recent increase has come from outside technology jobs: AI is no longer just for specialists.

What we do not see is a consistent shift in labour demand in AI-heavy occupations. Job openings in the most AI-heavy ones, mainly in technology, did fall sharply in 2023. But that was part of a global hiring correction, and came before AI mentions in job adverts took off. From early 2024 to March this year, as those mentions climbed, we see no consistent link between how often an occupation’s adverts mention AI and how its job postings changed (Chart 9). Pay tells a similar story. Since 2021, advertised pay in the occupations most exposed to AI has risen broadly in line with the rest of the market (Chart 10).

What about graduates and young people? Recent Department of Finance and ECB work finds that employment of younger workers in ICT fell by around a fifth between 2023 and 2025. Neither attributes this directly to AI, and it does coincide with the hiring correction in technology. But it is also consistent with international evidence that the share of junior jobs falls in firms that adopt AI.¹⁴ More positively, in the most recent data graduate roles in 2026 make up a larger share of Irish job adverts than at any point since Indeed’s data begin in 2018 (Chart 11), although their number is still around 40% below 2022, as hiring overall has cooled.

This is where Robert Shiller’s point about narratives comes in. ‘AI is wiping out junior jobs’ is one such narrative. So far, the Irish data do not clearly support it. My reading, which echoes recent IMF research, is that AI is changing what employers ask for faster than it is changing how many people they hire.¹⁵ History offers examples, such as automation in boot making: the mix of tasks and skills in an industry can change while the overall number of jobs stays the same, perhaps because demand rises as productivity increases and prices fall.   However, as AI technology itself and how it is being used is changing rapidly, its effects on workers and the broader labour market is something that warrants continuous monitoring.

The first rung of the ladder

In my recent Letter to Students of Economics, I noted that the tasks AI does well – summarising documents, drafting standard analysis and writing routine code – were skills junior workers have traditionally learned on the job, on the first rung of the ladder. It would be wrong, in my view, to stop hiring juniors because AI can do that work: where would tomorrow’s experienced staff come from? The learning curve can be shorter, as some of the literature suggests, but it should still exist. My message to employers is this: keep the bottom rung of the ladder in place, because today’s juniors are tomorrow’s experts.

AI also differs from earlier technologies, which tended to raise the value of education. Central Bank research (PDF 1.8MB) shows that the more educated you are, the more likely your job is both to be exposed to AI and to be helped by it. In other words, graduates could both gain and lose.

So, how best to prepare? First, learn to use these tools well. Our research (PDF 1.8MB) finds that missing digital skills significantly lower a worker’s chance of moving into a new occupation – and many of tomorrow’s occupations do not yet exist. Second, invest in what AI does less well: judgement, context, working with people, and taking responsibility for a decision. Together, these give you a strong bundle of skills that are valuable and not easily replaced.

Conclusion

Let me bring the two halves together. Your living standard depends on what you earn relative to what things cost. On prices, our job is clear: return inflation to 2% and keep it there, so that what you earn holds its value. On earnings, the long-run answer is productivity, and AI could be the biggest boost to productivity in a generation, though how big and how soon remains uncertain.

But, like good data, a powerful technology is necessary, not sufficient. The payoff depends on whether firms, and especially Irish-owned firms, use it. It depends on whether employers keep hiring and training young people. And it depends on people learning to use it well. History suggests the gains go to those who use a technology well. That means investing in yourselves. And it starts here.

Thank you.