"Will AI take my job?" That's the question everyone seems to be asking, and it's a fair concern given how fast artificial intelligence is advancing. But honestly, I think that's the wrong question, especially for us here in Ireland. The impact of AI on the Irish job market, and globally, is much more nuanced than just a simple count of jobs lost.

When I look at the immediate economic effects of AI, one of the most striking things is the sheer amount of money needed to build the underlying infrastructure. Big tech companies, many with major operations in Ireland, have historically been incredibly profitable. But they've shifted from funding this kind of massive investment with their own cash to borrowing on an enormous scale. Some estimates suggest that AI-related borrowing by these major US companies could hit around $200 billion this year alone.

This level of borrowing is creating a completely new dynamic in financial markets. Traditionally, high government borrowing would push up interest rates, which would then reduce private sector investment, what we call "crowding out." Now, it seems the reverse might be happening. AI investment is competing for the world's savings, and that's actually pushing up government borrowing costs. This isn't just a US issue; government debt markets are global, so the cost of UK and European government debt is also higher as a result. More money spent servicing government debt means higher taxes or less money for public services here in Ireland. While many factors influence government debt costs, long-term UK gilts recently reached levels not seen since 1998.

Taking a longer view, the real impact of AI on Irish jobs will ultimately depend on how quickly we adopt these technologies. If you've been regularly using the latest AI models, you'll know how capable they are. But just because a better technology comes along doesn't mean everything changes overnight. History has a way of repeating itself, and we've seen this before. Electric motors were available in factories from the 1880s, but the significant productivity gains didn't really start appearing until the 1920s. The technology wasn't the constraint; factories had been designed around central steam engines, and it took decades for owners to completely redesign their production processes. People and organizations often need many years to fully embrace change, and the technology itself is rarely the bottleneck.

AI seems to be following a similar path. The Office for National Statistics (ONS) in the UK reports that the share of businesses using at least one AI technology has almost tripled since late 2023, jumping from around 12% to 35%. This sounds impressive, but the adoption remains quite shallow. Only one in ten businesses describe their use as extensive, and about half report no impact on headcount. In fact, firms have been much more likely to retrain existing staff than to replace them. This really helps to dispel the myth of a widespread "jobs apocalypse" for AI and Irish jobs.

This relatively low level of deep adoption primarily comes down to a lack of expertise. The skills needed aren't just about coding anymore; they're increasingly about management. It's about identifying how to effectively use the technology, redesigning processes around it, and crucially, training staff on how to integrate it into their daily work. This suggests a shift in demand within the Irish job market, focusing on skills that facilitate AI integration rather than just technical development.

Another interesting issue linked to AI's capability is starting to emerge. Rather than the technology not being good enough, it might be becoming too good. This leads to what some are calling a "deployment ceiling," where the most capable systems become too powerful or too risky to release openly. We've already seen this happen. In April, Anthropic, a leading AI company, decided its latest model was too capable of finding and exploiting software vulnerabilities to make publicly available. Access was restricted to a select group of technology firms and critical infrastructure operators. Then, in June, the US Government used export controls to order Anthropic to suspend access to its latest models to foreign nationals worldwide.

While access to the Anthropic model may have been restored in July, the issue continues at Google, where its latest, most capable model is only accessible to a select group of companies. The rationale for this staged release is based on safety and the security of IT infrastructure, which many would consider prudent. However, a different issue arises if a permanent "moat" is created around a small group of organizations with privileged access to the best technology indefinitely. This could have significant implications for the competitive landscape and the equitable distribution of AI's benefits, impacting the development of AI and Irish jobs.

History again offers a parallel. After 1945, nuclear power, the most powerful technology of its era, was confined to a small group of countries. Access to it shaped economic and political power for decades. Frontier AI may follow a similar path, with a small group using the most capable models while everyone else works with safer but less effective versions. If this happens, the productivity gains and profits will be concentrated in those organizations and countries.

Unfortunately for us in Ireland, AI model development is largely a two-horse race between the US and China. This means our access to potentially transformative technology, and in turn, our long-term economic prosperity and the future of AI and Irish jobs, is dependent on governments over whom we have very limited influence.

In short, concerns about AI taking our jobs are largely misplaced. It's a technology we absolutely need to embrace. The people and firms in Ireland that adopt it first, and adapt their processes and workforce to it, are the ones who will thrive. The real challenge isn't fending off AI, but learning how to effectively integrate it into our lives and economy to ensure Ireland remains competitive and prosperous.