Adopt AI 2025 showed Europe shifting from AI theory to real deployment — in health, industry, climate and beyond. Even President Macron made a light nod to the Irish presence, as the Grand Palais turned into Europe’s most compelling AI showcase yet.
1. A summit that felt different
Adopt AI 2025 could have been just another enormous tech event. The numbers were impressive: the nave of the Grand Palais filled for two days with thousands of participants, hundreds of CEOs and C-level leaders, and more than 250 exhibitors spread across “AI for Finance”, “AI for Health”, “AI for the Planet”, “AI for Industry”, “AI for Travel”, “AI for Sport”, “AI for Society” and “AI for Tech”. (Adopt AI Grand Palais)
But the atmosphere was not that of a typical conference. The organisers liked to say: “this is where real-world AI meets the people who are building its future”. For once, that didn’t sound like marketing. The programme was packed with people who actually run banks, energy systems, rail networks, hospitals, airports, retail chains, data centres and climate services – and they came with concrete use cases, not prototypes on slides.
This was not the AI Action Summit in February, where heads of state wrestled with global AI politics and headline investment figures. (Le Monde.fr) Adopt AI was its business sequel: same location, but a different focal length. Less about grand declarations; more about deployment, governance and execution.
2. Macron’s visit: from rhetoric to positioning
The organisers were explicit: hosting Adopt AI under the patronage of the French presidency was a way of signalling that AI adoption is not just a tech theme, but a pillar of France’s and Europe’s economic strategy.
When Emmanuel Macron walked into the Grand Palais, the symbolism was deliberate. His presence was framed as “a celebration of innovation and the transformative power of AI”, but the underlying message was sharper. In his earlier interventions at the AI Action Summit, he had already insisted that Europe must “stop being naïve”, support its own AI systems and be willing to match US and Chinese ambitions with serious investments in data centres and compute. (Science Business)
At Adopt AI, the same line reappeared, but grounded in concrete examples. With CEOs of TotalEnergies, Airbus, Carrefour and others on stage, the president’s message was not about a hypothetical AI future. It was about making sure the AI that runs European hospitals, planes, grids, factories and administrations is built in Europe or at least anchored in European infrastructure and rules.
In this context, Macron also offered a brief nod to several international delegations, including Ireland. It was a light but meaningful reminder that Europe’s AI momentum is being built through a constellation of interconnected ecosystems, not by any single country alone.
3. Finance: when AI becomes core infrastructure
The finance tracks were among the most mature. There was no talk of experimental proofs of concept tucked away in innovation labs. Executives from Société Générale, Groupe BPCE, BNP Paribas, Caisse des Dépôts, Munich Re and others described AI systems already deployed in the nervous system of financial institutions: fraud detection engines, credit decisioning pipelines, real-time risk monitoring, and machine-learning tools that support long-term savings and pension schemes.
One session on “Building the next generation of banking with AI” made the shift obvious. The discussion was not about whether AI would transform banking, but about how fast leadership teams can re-architect organisations around it – in terms of data foundations, talent, governance and customer relationships. Another session on “From data to decision” explored the full value chain, from raw data to model design and operational decisions, emphasising that AI value only appears when institutions have resolved issues of data quality, infrastructure and model governance.
What emerged clearly is that finance is already living in a world where AI is not a nice-to-have analytics layer, but part of the institution’s critical infrastructure. The open question is no longer “if” but “how safely” and “how far”.
4. Health: between regulatory hard law and soft agents
The health sessions showed AI adoption at a different kind of frontier. On the one hand, major pharma players and healthtech innovators demonstrated what can be done: from Novartis explaining how data and AI strategies can be genuinely centred on patients in need, to biotech efforts using generative models and multimodal data for drug discovery and clinical insights.
On the other hand, regulators, legal experts and policy thinkers pulled the discussion back toward evaluation, safety and legitimacy. A striking cross-continental debate brought together experts from Europe and Japan to compare strict regulatory “hard law” frameworks with more flexible “soft law” approaches. The conversation turned on risk classification, the role of launch-and-learn regulatory sandboxes, and the future European Health Data Space as an infrastructure for trustworthy AI in medicine.
Another session focused on AI agents in healthcare: not just static models, but systems capable of taking actions, coordinating tasks and following up. The speakers were careful not to oversell autonomy, but their message was clear: if AI is to ease the burden on clinicians and improve outcomes at scale, it will have to move from being a diagnostic “second opinion” to becoming an operational partner embedded in hospital workflows.
Here, the tension was existential: can Europe’s regulatory model protect patients and rights without slowing down innovation to the point that real-world benefits drift elsewhere?
5. Industry and energy: managing complexity, not chasing hype
If any track at Adopt AI captured the specificity of the European industrial model, it was “AI for Industry”. Industrial leaders from ENGIE, Renault, SNCF, Siemens, Veolia, ITER, Airbus, Naval Group and others unpacked in concrete terms what it means to scale AI in sectors defined by long-cycle assets, heavy regulation and entrenched engineering cultures.
A roundtable with ENGIE, Renault and SNCF offered an unusually honest inside view. AI at scale, they argued, is less about magical algorithms than about clean data, coherent platforms, internal “enabler” teams, and sustained investment in people. The challenge is to move beyond stalled pilots toward environments where thousands of engineers, operators and field workers trust and use AI systems daily.
Another session on digital twins in industry showed how AI is migrating into the heart of complex systems. Digital twins are no longer static models of plants or processes; with AI they become dynamic entities, constantly fed by sensor data, capable of optimising entire value chains in real time. The most provocative element was the discussion of “agentic systems” – AI systems that can reason, plan and act across the lifecycle of an asset, not just provide isolated predictions.
Knowledge-intensive industries were given their own wake-up call. ITER, Airbus and others explained how they are using AI to tackle one of their most dangerous problems: the silent erosion of institutional memory as experts retire and projects span decades. If industrial leaders don’t invest in open ecosystems, clean documentation and AI-supported knowledge management, they argued, they are already late.
Add to this a defence-oriented session on sovereignty, where former defence chiefs and AI entrepreneurs insisted that Europe must “own its data science and AI stack or risk losing sovereignty”, and a clear thread appears: AI for industry in Europe is not primarily about cool demos, but about control over critical systems.
6. Retail, travel, sport, longevity: the edges of the AI economy
While finance, health and industry dominated the core, other tracks at Adopt AI hinted at where the next frontiers may lie.
Retail and commerce were framed through the lens of “agentic commerce”. Sessions explored how AI agents will become the main interface between consumers and brands, negotiating, filtering and purchasing on behalf of users. If bots do the shopping, the panelists asked, how do brands remain visible and relevant? The discussion introduced concepts such as “generative engine optimisation” and new key performance indicators that track how often a brand is recommended or selected by AI agents, rather than by human search. Physical retail was not left behind: another session showed how computer vision, shelf analytics and task automation are already changing stores by reducing friction for shoppers and workload for staff.
Travel had its own version of this story. Executives from airport operators, hotel groups and travel platforms described hyper-personalisation built on connected data and AI – from seamless airport journeys orchestrated across systems to dynamic travel experiences built in real time. A recurring phrase was “AI that amplifies human judgment”: the aim is not to erase staff, but to give them better information at the moment decisions are taken.
Sport and longevity might have looked peripheral on the programme, but they pointed to deeper shifts. A panel led by AWS with executives from the NBA, top cycling teams and European football clubs showed how AI and cloud technologies are filtering into performance analytics, strategy, fan engagement and stadium operations. Meanwhile, a session on longevity used data platforms to map biotech innovation against the “hallmarks of ageing”, illustrating how AI is already guiding investments and research into new therapies.
Taken together, these tracks suggested that AI is becoming an invisible layer of optimisation and personalisation in sectors that Europeans experience daily: shopping, travel, entertainment, healthspan.
7. Climate, planet and trust: Europe’s distinctive angle
If there was one area where Europe sounded particularly confident, it was climate and the environment. A panel on “Greening AI and greening with AI” brought together actors from SAP, UNESCO and the French Ministry for Ecological Transition to explore the dual agenda: making AI models themselves more efficient and low-carbon, and using AI to accelerate climate solutions. (Coalition for Sustainable AI)
Another roundtable, centred on weather and climate intelligence, connected AI to Destination Earth, EuroHPC and the emerging AI Factories. Participants from ECMWF, Météo-France, the European Commission and supercomputing centres described how high-resolution models, AI-assisted forecasts and new data products are being built not as research toys, but as operational tools for adaptation, emergency management and resilience.
Yet the climate sessions did not ignore risk. Several speakers acknowledged the paradox of deploying massively energy-intensive AI systems to fight climate change. The broader debate echoes analyses from earlier AI events in Paris, where NGOs and researchers have warned of the environmental and geopolitical costs of hyperscale AI.
The “AI for Society” track, meanwhile, took on trust head-on. A session curated by experts from the World Bank, OECD, WIPO and the ICC explored what it takes to create an AI ecosystem that is both innovative and legitimate: robust policy, inclusive governance, education, IP clarity, and protections for those most exposed to automation. This was one of the places where Europe’s regulatory reflexes felt less like a constraint and more like an asset.
8. Tech, agents and sovereignty: the meta-layer
Beyond sectors, the “AI for Tech” content hinted at where the underlying stack is going. A keynote on cognitive autonomy in aerial systems described how agentic AI enables drones and other platforms to interpret missions, reason about constraints and act with verifiable intent – not just follow pre-programmed routines.
Another well-attended conversation between Mirakl and OpenAI looked at “business in the age of agents”. Here again, the focus was on AI systems that don’t just answer questions but orchestrate workflows: updating catalogues, handling multilingual support, automating back-office processes and, ultimately, becoming the default interface through which a large share of business gets done.
This is not just a technical frontier; it is a strategic one. Agentic AI raises new questions about liability, explainability and control. Combined with the earlier sovereignty sessions – arguing that Europe must own key parts of its data, model and infrastructure stack – it painted a picture of an AI landscape where the real competition is not over who has the largest models, but over who controls the architecture in which agents operate.
9. AI Factory France, NVIDIA and the Irish AI Antenna: the quiet infrastructure story
If the sessions described the visible surface of AI adoption, a quieter layer of the summit concerned infrastructure. On the French side, one of the most significant announcements was the acceleration of AI Factory France (AI2F), led by GENCI, built on Jean Zay, Adastra, Joliot-Curie and the future Alice Recoque exascale supercomputer. AI2F positions itself as a one-stop shop aligning public HPC resources, training and industrial use, and has now entered into a landmark collaboration with NVIDIA’s Inception and Connect programmes to support European start-ups and software vendors. (genci.fr)
At the European level, EuroHPC has been quietly building a network of AI Factories and AI Factory Antennas across member states, designed to give SMEs, researchers and public bodies access to AI-ready compute and expertise. (eurohpc-ju.europa.eu)
Ireland’s recently selected AI Factory Antenna (AIF IRL-Antenna) sits squarely within this framework. Implemented by a consortium led by ICHEC and CeADAR, co-funded by the EU and Ireland’s Department of Further and Higher Education, Research, Innovation and Science, it will connect Irish users directly to leading AI hubs in France and Luxembourg, among others. (gov.ie)
For Ireland, this is more than an infrastructure project. It is a bridge: between Dublin and Paris, between national ecosystems and European platforms. At Adopt AI, in formal sessions and hallway conversations alike, that network logic was increasingly visible. AI Factory France is not meant to be a closed French fortress; it is a node in a European grid where antennas like the Irish one will plug in, test workloads, train models and send back applications into their own sectors – health, climate, fintech, space and beyond.
10. What Adopt AI really told us about Europe
Stepping back from the overload of content, three things stand out from Adopt AI 2025.
First, Europe has moved on from talking about AI in the conditional tense. The summit’s tracks made it obvious: AI is already inside Europe’s hospitals, banks, factories, airports, shops and climate services. The questions are increasingly operational: how to scale, how to govern, how to align with regulation, how to manage talent.
Second, the continent is beginning to understand AI as infrastructure. The conversation about AI Factory France, EuroHPC AI Factories, the Irish antenna and other nodes is not a side note; it is the prerequisite for everything else. Without shared compute, data spaces and training pipelines, the best intentions on health, climate or industry will remain stuck at the level of pilots.
Third, Europe is trying – not always perfectly, but seriously – to reconcile three imperatives: competitiveness, sovereignty and trust. The sessions on defence and sovereignty, the exchanges on regulation in health, the debates on climate footprints and inequality, all showed a region that is aware of the risks but refuses to retreat into paralysis.
For Ireland more broadly, the momentum around Adopt AI underscored how the country fits into Europe’s wider AI puzzle. As the continent accelerates deployment across sectors, new opportunities are emerging for countries with strong digital and deep-tech foundations to contribute to, and benefit from, this evolving landscape.
Adopt AI 2025 was not the end point of this story. In many ways, it felt like the beginning – the moment when Europe’s AI conversation finally became concrete.
