That question ran through the International AI Summit, held at the RDS in Dublin on 14 October 2026. Bringing together leaders from government, research and industry, the event explored how Europe can translate AI capability into competitiveness, innovation and tangible benefits for society.
For us at Creme Global, the most valuable discussions connected technological ambition with practical questions. What data do we need? Which problems are worth solving? How do we validate the output? And how do we equip people to use these systems effectively?
Our central takeaway: Europe’s AI opportunity lies in connecting scientific expertise and valuable data with applications that deliver reliable, measurable outcomes.
1. The opportunity is in the application
The opening addresses set an ambitious direction. Taoiseach Micheál Martin announced Intelligence Applied, a programme intended to position Ireland as a leading applied AI economy, supported by investment in skills, research and industry collaboration.
The industry discussions made that ambition more concrete. Siemens’ Dr. Peter Koerte highlighted a part of the AI conversation that can receive less attention than infrastructure and frontier models:
“The application layer is really the layer where the value is being generated for the customer.”
He emphasised the importance of context, previous cases and industry-specific knowledge in making AI useful. The model alone is not the whole solution.
For science-led businesses, this points to a better starting question than “Which AI should we use?”
Which decision do we need to improve, and what would a better outcome look like?
That might mean improving a formulation, identifying a risk earlier, reducing development time or making a complex supply-chain decision with greater confidence. The technology should follow the problem—not the other way around.
2. AI can accelerate science, but evidence still matters
In AI and the Frontiers of Science, Sir Demis Hassabis of Google DeepMind offered an important distinction between solving problems in verifiable domains and working with the complexity of the natural sciences.
AI can help generate hypotheses and identify promising directions. But biology, chemistry and other real-world disciplines still require experimental testing, new data and iteration.
“You still need real-world loops.”
The healthcare discussion reinforced this point. Johnson & Johnson’s Belén Fraile stressed the importance of robust validation, transparency and human oversight throughout a model’s life cycle.
Nuritas founder and CEO Dr. Nora Khaldi brought a particularly relevant perspective for food and nutrition. She described using AI to discover peptides in nature, with predictive work extending beyond potential efficacy to oral availability and stability under food-processing conditions.
A promising discovery must also work in the product and environment in which it will be used.
Our takeaway is straightforward: faster prediction is valuable, but it is not a substitute for validation. The opportunity is to make the scientific process more effective, not to bypass it.
3. Food systems made the potential tangible
One of the day’s clearest demonstrations of applied AI came from Kerry Group CEO Edmond Scanlon during Feeding the Future: How AI is Driving Innovation in the Food System.
He described applications across customer engagement, supply-chain management and manufacturing.
Concept-development work that previously took days or weeks could now be completed in minutes. Supply-chain tools were helping teams handle change requests faster, with an additional reported benefit: a 25% reduction in email traffic within the supply-chain team. Digital twins of selected manufacturing facilities or production lines had also supported improvements of around 20% in frontline-worker productivity.
These were useful examples because they connected AI with recognisable business outcomes: less time coordinating, faster decisions and more productive operations.
Scanlon also looked beyond efficiency. He described early-stage work combining AI, biotechnology and food science to explore alternative ingredients with similar taste or functional properties to raw materials affected by supply pressures.
His explanation of the underlying opportunity was particularly memorable:
“AI without the data is not of value.”
The combination he highlighted was AI, accumulated scientific knowledge and human capability—not technology in isolation.
For us, this was a compelling expression of what applied AI can contribute to food and nutrition: using existing knowledge more effectively while opening new avenues for innovation.
4. Data and know-how must remain strategic assets
Access to data was a recurring theme, but so was control over it.
Microsoft Vice Chair and President Brad Smith raised a concern with direct relevance to knowledge-intensive businesses: organisations can risk paying twice for AI—once financially, and again through the knowledge and expertise they contribute.
He argued that companies and economies need to retain the ability to protect and control their accumulated know-how.
Scanlon made a related point from the food industry’s perspective. Kerry’s scientific intellectual property has been developed over decades, and protecting that knowledge remains central to how it works with technology partners.
Our interpretation is that an AI strategy needs to address more than access to a model. It should also establish what data is being used, where it comes from, who can access it, how it is protected and how outputs can be examined.
The goal should be to extract more value from an organisation’s knowledge without losing control of the knowledge itself.
5. Scaling AI means redesigning the work
OpenAI CFO Sarah Friar described a shift from asking AI questions towards using agents to carry out work. The afternoon enterprise discussion explored what that transition means inside an organisation.
Fenergo’s Niall Twomey offered a revealing example. As AI accelerated engineering, bottlenecks appeared elsewhere: product requirements, review and release processes, and customer support. Improving one part of the business exposed the need to adapt the others.
SAP CTO Dr. Philipp Herzig similarly emphasised working backwards from business outcomes, such as cash flow and inventory optimisation, rather than simply encouraging more AI usage. He also highlighted the importance of using models efficiently as organisations move beyond proofs of concept.
The lesson for businesses is to look beyond the individual task. A successful pilot is a starting point. Sustained value depends on how the technology fits into the wider process, how responsibilities change and whether the organisation can act on the increased flow of information.
Measure the improvement in the workflow—not just the activity in the AI tool.
6. People need to be part of the change, not an afterthought
The workforce sessions brought an essential balance to the discussion.
Una Fitzpatrick of Technology Ireland emphasised investment in people, upskilling and reskilling. Fionnuala Ní Bhrógáin of the Communications Workers’ Union argued for meaningful worker participation and collectively bargained discussions about how technology is introduced.
Ní Bhrógáin also asked a question that should sit alongside every productivity forecast:
“Who benefits from those productivity gains?”
She raised the longer-term implications of automating entry-level work: how do people develop the experience needed to become the senior professionals of the future?
There were encouraging examples too. Tines co-founder and CEO Eoin Hinchy pointed to the company’s recently announced plan for 100 new jobs in Ireland, with half intended to be entry-level. His emphasis was on curious, talented people who can understand customers, interpret information and use the technology effectively.
For organisations adopting AI, the implication is clear: technical capability and human capability need to develop together.
Training should help people question outputs, recognise limitations and apply their expertise—not simply operate a new interface.
7. Europe needs capacity—and customers
The later sessions brought attention back to the foundations of AI: energy, chips, computing infrastructure and the people needed to build and operate it.
Mercury Engineering CEO Eoin Vaughan highlighted factors that can be overlooked in infrastructure debates: community trust, early engagement and the development of local skills. AMD’s Keith Strier stressed the complexity and long-term planning involved in building substantial computing capacity.
The startup discussions identified another requirement for a strong European AI economy: customers willing to buy.
Jentic co-founder and COO Dorothy Creaven called on established businesses to give innovative companies opportunities to compete for meaningful commercial work, rather than stopping at small pilots.
Her argument was not for charity, but for recognising the value these businesses can deliver.
That is a useful challenge for the wider ecosystem. Supporting innovation means more than funding its development. It also means creating pathways to adoption, revenue and scale.
8. Trust is part of the route to adoption
The day did not produce a single view on regulation.
Some industrial speakers pressed for innovation to move faster and questioned whether governance could keep pace. Healthcare speakers called for clearer, more consistent requirements that work alongside existing sector-specific rules.
Smith, meanwhile, argued for stronger safety measures, genuinely independent evaluation and transparency about what those evaluations reveal.
Closing the summit, European Commissioner Michael McGrath captured the connection between trust and adoption:
“If people do not trust AI, they simply will not use it.”
For us, the practical response is to make trust something that can be demonstrated: through the quality of the data, the relevance of the testing, the transparency of the outputs and the expertise of the people responsible for using them.
From ambition to better decisions
Our conclusion from the day is not that every organisation needs to do everything at once. It is that organisations need a credible route from experimentation to useful, dependable applications.
Start with an important question. Understand the available data and its limitations. Build and validate an appropriate model. Keep domain experts involved. Measure the difference it makes.
At Creme Global, our work brings together scientific expertise, data engineering, mathematical modelling and applied AI to help organisations take that journey.
The summit offered plenty of ambition. The next step is to turn that ambition into evidence, practical applications and better decisions.
Interested in turning your organisation’s data and scientific expertise into measurable value? Talk to the Creme Global team about how applied AI and predictive modelling can help you make better, evidence-based decisions.
