Europe has plenty of AI ambition. Now comes the difficult part

 In the space of 24 hours this October, Dublin hosts two major conversations about artificial intelligence. On 13 October, the Analytics & AI Summit brings together the people trying to make data and AI work inside organisations. The following day, the International AI Summit at the RDS brings European and global leaders together to discuss competitiveness, investment, infrastructure and Europe's position in the next phase of technological change.

Both matter. But the space between them is where one of the biggest questions about AI now sits.

Europe does not have an AI ambition problem. Organisations are not short of strategies, pilots, licences or enthusiasm. The question is whether we can turn that activity into measurable organisational performance. AI activity has become easy to generate. AI value has not.

I spend much of my working life with organisations trying to move AI from experiment to something embedded in how the business operates. Increasingly, the constraint is not the technology. It is everything around it.

The data exists, but nobody is completely confident. The process being automated has never really been mapped. A pilot has a sponsor, but nobody owns the operational number that is supposed to move if it succeeds. The technology works, but the roles and ways of working around it remain exactly as they were.

These are not glamorous AI problems. They are, however, the problems that will determine whether the current investment cycle delivers.

One question I increasingly ask organisations is simple: what number are you using to tell me your AI programme is working?

Too often the answer is licences deployed, pilots underway or hours theoretically saved. Those numbers can be useful, but they are not value. Licences tell me about access, not about whether work has changed. A portfolio of eleven pilots is not eleven-elevenths of a transformation; sometimes it means nobody has decided to stop ten and industrialise the one that matters.

Even hours saved needs interrogation. Hours saved into what? Giving a team back four hours per person a week may be genuinely worthwhile. But unless we decide what that capacity is for and measure the outcome, we should be careful about presenting it as financial return.

This is why we need to stop treating AI strategy primarily as technology strategy. It is operating-model strategy.

A useful test is to ask whether you can name a process in your organisation that has materially changed because of AI. Has the sequence of work changed? Have hand-offs disappeared? Has the point at which a human intervenes moved? Has cost-to-serve moved? If you cannot point to something like that, you may have deployed AI. You have not adopted it.

That distinction matters for Europe as much as for individual organisations. There is an important debate underway about the conditions Europe needs to compete: infrastructure, compute, investment, regulation, skills and sovereignty. Those things create the conditions for adoption. They do not guarantee it. Europe will not become more competitive simply because European organisations have access to more AI. The prize comes when thousands of businesses and public bodies use it to redesign work, make better decisions and become materially more productive.

That is why Ireland has an interesting opportunity. We sit at the intersection of global technology, highly regulated industries, a significant public and semi-state sector and a workforce already adapting quickly. We also have much of what makes adoption difficult: legacy systems, regulation, governance requirements and established ways of working. That could be read as a disadvantage. I think it could become an advantage. If we can show how AI is implemented responsibly and at scale inside real organisations, Ireland can contribute something more valuable to the European conversation than another statement of ambition.

That requires less exciting work: mapping the processes underneath the technology, getting serious about data, establishing the baseline before claiming the improvement, deciding who owns the outcome. Most importantly, it requires distinguishing activity from value.

The next phase of AI will be less forgiving. Boards will ask what they got for the money. Pilots that were exciting in 2025 will need a path to production. That is healthy, because the organisations that win with AI will not be the ones that moved first or bought the most technology. They will be the ones that worked out how to make it change something that matters.

Europe has no shortage of AI ambition. Now comes the difficult part: proving it works.